CA1139884A - Half duplex integral vocoder modem system - Google Patents
Half duplex integral vocoder modem systemInfo
- Publication number
- CA1139884A CA1139884A CA000354955A CA354955A CA1139884A CA 1139884 A CA1139884 A CA 1139884A CA 000354955 A CA000354955 A CA 000354955A CA 354955 A CA354955 A CA 354955A CA 1139884 A CA1139884 A CA 1139884A
- Authority
- CA
- Canada
- Prior art keywords
- input
- signal
- modem
- speech
- fourier transform
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B1/00—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission
- H04B1/66—Details of transmission systems, not covered by a single one of groups H04B3/00 - H04B13/00; Details of transmission systems not characterised by the medium used for transmission for reducing bandwidth of signals; for improving efficiency of transmission
Abstract
ABSTRACT OF THE DISCLOSURE
Disclosed is a system whereby in a transmit mode analog speech is sampled and converted digitally to 12 bits of accuracy and then fed into a fast Fourier transform (FFT) processor which analyses the speech into spectral and pitch parameters. These parameters are then quantized into a data stream which acts as an input to a differen-tial phase shift keying modulator. The modulator constructs a multi-tone modem output signal from a 25 tone stack which is converted to an analog signal which is fed into a communications channel. In the received mode a modem input signal from the communications channel is fed through the same analog to digital converter which was used for input speech. A data stream is now provided which is coupled back into the same digital FFT processor which now operates to provide pitch spectral coefficients which are then separated and used by a synthesizer to reconstruct A speech waveform. The speech waveform is applied to a digital analog converter which is the same converter used to produce the line signal when the processor operates in the transmit mode. The FFT processor implements a single FFT algorithm which is used for both vocoder and modem processing in both the trans-mit and receive modes.
Disclosed is a system whereby in a transmit mode analog speech is sampled and converted digitally to 12 bits of accuracy and then fed into a fast Fourier transform (FFT) processor which analyses the speech into spectral and pitch parameters. These parameters are then quantized into a data stream which acts as an input to a differen-tial phase shift keying modulator. The modulator constructs a multi-tone modem output signal from a 25 tone stack which is converted to an analog signal which is fed into a communications channel. In the received mode a modem input signal from the communications channel is fed through the same analog to digital converter which was used for input speech. A data stream is now provided which is coupled back into the same digital FFT processor which now operates to provide pitch spectral coefficients which are then separated and used by a synthesizer to reconstruct A speech waveform. The speech waveform is applied to a digital analog converter which is the same converter used to produce the line signal when the processor operates in the transmit mode. The FFT processor implements a single FFT algorithm which is used for both vocoder and modem processing in both the trans-mit and receive modes.
Description
Background of the Invention This invention relates to speech compression systems and more particularly to a digital processor which combines the functions of both a vocoder and a modem in a half duplex mode of operation.
Vocoder systems function to transmit speech signals in a coded manner to reduce the transmission bandwidth which would other- I
wise be required if the speech was to be transmitted in an uncoded manner. Thus a vocoder system includes a transmit terminal to analyze the characteristics of the speech wave to be encoded and to encode the speech wave and a receive terminal which is used to synthesize, from the coded signal set to it, a reconstruction of the original speech wave. Dats modems on the other hand function to facilitate the transmission of data, for example data from a speech vocoder, over a transmission medium. Thus a modcm includes a trunsmil tcrminal to convert the encoded data into a modulating signal which is used to modulate a carrier frequency, and a receive terminal to demodulate the received signal and thereby recover the transmitted data. Both vocoder and modem equipment are therefore required for transmission of speech signal in an efficient manner.
A known system of particular interest is disclosed in U.S. Patent 3,681,530, entitled "Method and Apparatus for Signal Bandwidth Compression Utilizing The Fourier Transform Of The Logarithm Of The Frequency Spectrum Magnitude", issued to H.~. Manley, et al.
There a system is described wherein an input speech waveform i9 con-verted into an electrical signal which is then digitized by an analog to digital converter. Following this the digitized signal is directed through a device where magnitudes of the frequency spectrum of the input speech wave are obtained. These magnitudes are then directed to a logging circuit to obtain a logarithm of the frequency spectrum magnitudes of the input speech signal. The log magnitudes of the frequency spectrum MR/
~39884 are then directed to a computer where the discrete Fourier transform (DFT) of the log spectrum magnitudes are obtained to form what is commonly referred to as the "cepstrum" of the input speech signal. The system also includes an analysis section of the vocoder terminal which operates to decode received data and separate it into pitch data and vocal track impulse data. Also included is a computing device for computing the logarithm of the spectrum envelope of the vocal track impulse response function using FFT techniques. A convolution unit then convolves the pitch data with the impulse response data to yield the desired synthesized speech signal.
A digital modem adapted to be utilized in conjunction with the system described in U.S. Patent 3,681,530 is disclosed in U.S. Patent 3,617,941, entitled "Table Look Up Modulator", which issued to J. DeLellis, Jr., one of the subject inventors, and which discloses a differential phase shift keying (DPSK) system for generating in digital form a plurality of tones at a predetermined baud rate.
Additionally a digital processor which can use the same hardware to implement both the vocoder and modem functions is disclosed in U.S. Patent 3,706,929, entitled "Combined Modem and Vocoder Pipeline Processor", by J.L. Robinson, et al., which describes a processor which operates as a half duplex system, i.e. either in a transmitting or receiving mode. In the "transmit" mode, the input signal to the processor is a speech wave and the processor perform the vocoder function of speech wave analysis and the modem function of generating a modulating ~-r ~h f~ ~13~B~
signal which carrie~ the results Or the speech wave analysls.
Thls modulating signal becomes the output signal Or the processor ln this mode, which is used ln the modulator Or a conventlonal communications transmitting ~ystem, In the "receive" mode the input signal to the processor is a speech information bearing communlcations signal such as may be derived from a conventional communications receiver, The same t~pe Or modulation used ln the transmlt mode mu~t be used in this mode. The processor performs the modem runction Or demodulating o~ the input signal and the vocoder runction Or the synthesis Or the speech wave.
The output signal Or the processor in this mode is then used to drive conventional volce reproduction circuitry.
Acoustic si~nal processors employing F~ techniques are also disclosed in U,S. Patent 3,662,108, entitled "Apparatus ~or Reducing Multi-Path Distortion Or Signals Utilizing Cepstrum Technique", by J,L, Flannigan and U,S, Patent 3,634,668, entitled "Log Spectrum Decoding Utilizing Symmetry", by H,L, Sharfer, Accordingly, an object of the present invention is to provide a system including a digital processor which is particu-larly adapted for voice communications systems, Another obJect Or the present invention is to provlde a system including a digital processor which ls adapted to runc-tion as both the vocoder and modem in a voice security communl-cations system, It i9 a rurther ob~ect Or the present invention to provide a system includlng a digital signal processor which combines the function Or both a vocoder and a modem in a halr duplex mode Or operation, Summary 3o Disclosed is a vocoder modem system having a digital signal processor which implements a Fourier transrorm (FFT) ~` ~
¦algorlthm used rOr both vocoder and modem processing in both the transmit and receive modes. In the transmlt mode, a 256 point ¦(sample) FFT algorithm erfecting a complex DFT is used to simul-taneously perrorm both spectrum and cepstrum analysis for pitch detectlon and cepstral encodlng of input speech signals. The pltch and cepstral parameters provide the input to a rour phase dirferentlally coherent phase shirt keying (DPSK) table loolc-up modulator which generates 25 digltal modem tones which are converted to the audio band and transmitted over a wire line or high ~requency llnk. In the receive mode, a 64 point complex FFT algorithm is used to perrorm a 12~ point, real input DFT
on the modem input which is ~irst used to demodulate the tone stack Or the input. The same 64 point F~r algorithrn is then used again to transform the decoded cepstral parameters into a log spectrum and to an impulse response which is used with pitch inrormation to generate an output speech waverorm ln a syntheslzer.
Synchronization is provided by an iterative dirrerence equation rilter centered on an empty slot ln the tone stack, Brier Description Or the Drawings These and other ob~ects of the present invention will become more fully apparent from the followlng detailed descriptio taken in connection with the following drawings ln whlch:
Figure 1 is a slmplified block diagram of a hal~ duplex vocoder modem;
Figure 2 is a graph illustrative of the general appear-ance o~ a logged magnitude spectrum;
Figure 3 i~ a graph illustrative of the Fourier trans-form o~ the log spectrum magnitude shown in Figure 2;
Flgure 4 is a block diagram illustrative of the pre-3o rerred embodiment of the sub~ect invention;
~ ~ ~ il39W~
l Figure 5 lq a detailed block diagram Or the portlon ¦of the embodiment shown in Figure 4 utilized ln the transmlt ¦ mode;
¦ Figure 6 ls a detalled block diagram illustrative Or ¦the portion of the embodiment shown in Figure 4 utilized in the ¦ receive mode;
Figure 7 is a graph illustrative Or the samples Or the input speech ~Javeform red lnto the FFT computer portion of the l apparatus shown in Figure 5;
LO Figure 8 is a graph illustrative Or the logged spectrum magnitude Or the samples shown in Figure 7;
Figure 9 is a graph illustrative of the cepstrum of the samples shown in Figures 7 and 8;
Figure 10 is a graph illustrative o~ the data input tlming dlagram for the receive mode of operation;
Figure 11 is a graph illustrative of the synchroniza-tion alignment ror the system during the receive mode;
Figure 12 is a graph illustrative Or the cepstrum sample~ ror a modem signal received by the subJect invention;
O Figu~e 13 is a graph illustrative of the sampled received spectrum magnitude o~ the modem signal received; and Figure 14 is a graph il1ustrative Or the sampled syn-thesizer implllse response o~ the received modem signal, Description Or the Prererred Embod~ments A simplified block di.agram Or the present inventlon 1~ shown in Figure 1. In the "transmit" mode, the input speech i8 sampled and digitized in analog to digital converter means 10 and red into a computer (digital data processor) 12 which analyses the speech in the spectral and pltch parameters. These ,0. parameters are quantized into a multi-bit data stream which acts ~ i~9W~ ' as an input to a dif~erential phase shift lceying (DPSK) modulator 14 The DPSK modulator 14 constructs a multi-tone modem slgnal whlch i3 then converted into an analog signal by digital to ana-log means 16 and ~ed as a transmitted modem waverorm into a communlcations channel. In the "receive" mode, the same analog to digital converter means 10 accepts a modem input signal and outputs a digital data stream wh~ch is red to a DPSK demodulator 18 which separates the pitch and spectral coefricients which are then used by a synthesi~er 20 to reconstruct the input speech in digltal rorm. The output speech is then converted in the digital to analog converter means 16 which was used to produce the line signal when the system operated in the transmit mode.
The sub~ect invention is an outgrowth of thc ~ r-cepstrum vocoder technique which has been described, for example, in the above rererenced U S. Patent 3,681,530. This patent, moreover, discloses a complex 256-point fast Fourier translorm algorithm hereinafter referred to simply as FFT, which is used to simultaneously generate the spectrum magnitude and cepstrum Or an input speech waveform. In the FFT~cepstrum approach, a 256 sample segment Or input speech is transrormed to produce spectral magnitude information at 128 frequencies across the speech band This i~ surricient to resolve the pitch harmonics Or voice sounds so that the spectrum envelope may be extracted without the confounding Or the envelope and pitch inrormation that occurs, ror example in a channel vocoder. The resulting measured magnltude spectrum is the product Or the spectrum envelope with pitch harmonic line spectrum. This product is separated into a sum by taking the logarlthm o~ the high resolu-tion magnitude spectrum. The general appearance o~ such a high 3o resolution, logged magnitude spectrum is snown in Flgure 2.
il3981~
The cepstrum or FFT of this logged spectrurn is shown ln Figure 3.
The transform o~ the slowly varying part of the spectral envelope appears in the low "que~rency" or low delay part of the cepstrum below 3 milliseconds. This low delay portion is selected ror quantization and transmission the er~ect of which is to remove the pitch harmonic ripple in the high resolution, logged magnitude spectrum by low pass ril~ering rrhis method, both theoretically and experirnentally, yields the most accurate possible smooth spectrum envelope description with no confounding Or pitch and spectrum envelope inrormation.
The spectrum parameters obtained by the FFT-cepstrum vocoder are nearly uncorrolated with each other, i.e. rrom channel to channel, and thererore are in an efricient representation Or spectrum shape from a data reduction vicwpoint. They are also in an ideal form to enable the system to take advantage Or corro-lations rrom ~rame to rrame, i.e across time Therefore, an interlacing technique is employed in which the low delay cepstrum values which characterize the gross spectrum shape are thus trans-mltted every ~rame and the higher order coefficients are trans-mitted every other rrame is utilized by the quantizer Pltch deteotion ln the FFT-cepstrum approach i9 accomplished by detect-ing and tracking the cepstral peak which occurs in the high "querrency" or high delay part o~ the cepstrum during voiced sound9. The time delay to this peak is the pitch period. The cepstral peak can be thought of as the "frequency" of the rlpples in the log spectrum magnitude Voiclng is detected in the cepstrum pitch extractor by reserving both the ratio Or low band to high band spectral magnitude energy and by observlng the presence Or a signiricant cepstral peak The present invention as wlll be shown employs a l;
il3~8t~4 single N point, radix-4 FFT algorithm which is capable Or operat-ing either in the transmit or the receive mode~. In the transmit mode an N=256 point complex FFT operation is performed while ln the receive mode an N~64 point complex ~FT ls utllized to perrorm a 2N=128 point real input discrete Fourier transform which allows the use of the same coding in each instance Rererring now to Figure 4, there is disclosed a com-posite block diagram o~ the present invention. What is si~niri-cant is the utilization o~ the same input and output sections 22 and 24, respectively, with common FFT and even/odd separator computer means 26 in both the transmit and receive modes o~
operation. This is carried out in accordance with the synchronous operation Or a plurality of signal switching elements shown in Figure ~ as the transmit/receive (T/R) switch blocks 28, 30, 32, 34, 36 and 38. As shown, the analog speech input and the analog modem input are both coupled to the switch block 28 which alternately operates in the transmit mode and receive mode, respectively, to couple the respective analog waveforms to an analog lnput processor 40 whose output is coupled to an analog to digital (A/D) converter 42. The A/D converter 42 operates to converk the signal inputted thereto to a 12 bit digital data ~tream whioh i~ red into a data bufrer store 44 whlch is adapted to accommodate 256 12-bit words in storage at a tlme. In the receive mode, the 12-bit data stream is additionally fed to a sync detector block 46 which operates to control the data bufrer over two baud intervals in a manner to be discussed subsequently In the transmit mode, the T/R block 30 operates to couple the digitized speech input into a speech pre-processor 48 where it i~ then red to ~he FFT and even/odd separator computer 3o means 26 through the T/R block 32. The FFT and E/0 separator v 1~
computer means 26 is coupled to a transmit ~pectrum processor 50 and to a pitch extractor 52 through the actlon of the T/R
block 34. The output Or 3pectrum processor 50 whlch comprlses the logged spectrum Or the FFT samples is coupled back to the input o~ the FFT apparatus 26 in a manner as disclosed ln the arorementioned U,S. Patent 3,681,530 to perrorm the Fourier transrorm Or the log Or the speech spectrum (cepstrum) whlch signals are outputted to a quantizer 54 along with the output Or the pitch extractor 52. The quantized cepstrum and pitch signals are fed to a DPSK modulator 56 which forms a digital modem signal which is coupled to an output digital to analog converter 58 through the T/R block 36. The analog modem signals outputted rrom the converter 58 are fed to the transmlsslon channel via an analog output signal processor (low pass filter) 60 and the T/R block 38.
In the receive mode, the analog modem input is digitlzed into a data stream and ~ed via the T/R blocks 30 and 32 to the FFT and ~/0 computer means 26, wherein a ~irst pass o~ the data stream is made through the computer means 26 in a 64 point complex FFT routine in order to perform a 128 point real lnput DFT
transrorm, The results Or the rirst pass is outputted to a demodulator 62 where ceRstrum signals are developed. The~e signals are ~ed back into the computer mean9 26 through a signal summer 64 and the T/R device 32, Again a 128 point real input DFT trans~orm is formed by means Or a 64 point complex FFT
routine. A ~irst portion (lmaginary) o~ the second pas9 compu-tation is fed as an impulse response signal to a syntheslzer 66 whlle a second portion (real) is ~ed to a recelve spectrum processor 68 which provides a feedback path back into the FFT
3o ~or second pass operation Or the FFT Or the data. This operatlon will be considered further when Figure 6 is considered. The synthesizer 66 also receives pitch in~ormation ~rom the demodu-lator 62 along with the impulse response signals developed ~3~3884 during the second 64 point calnplex Fi~r operation to provl~e digltlzed speech waveforms which are then coupled to the digltal to analog converter 58 through the T/R device 36. The analog speech waverorms are then fed to the analog output processor 66 and to a transducer, not shown, through the T/R device 38.
In order to more rully understand the operation Or the subject invention, the separate portions of the system shown in Figure 4 are separated and expanded in the rcspective transmit and receive mode apparatus shown in Figures 5 and 6.
Referring now to Figure 5, shown is a detailed electrical block diagram illustrative Or the means ~or implementing the digital processing which is performed in a transmit mode Or operation, The heart Or the speech analysis portion Or the system is a 256-point, radix-4 FFT algorithm which is shown being performed by the block 26 which operates simultaneously to generate the spectrum and cepstrum Or the speech input. The modulation section 56 runctionally implements a table look-up routine, a typical example Or which is disclosed in the above-rererenced U.S. Patent 3,617,941, and is operable to generate four phase DPSK stacked tones, The FFT routine is adapted to output data in rrames at a system updating interval which is typically every 20,8 milllseconds ror a 2400 bit pcr second (b/~) mode and every 30 milliseconds ror a 1200 b/s mode, The lnput speech waveform coupled rrom an input channel, not æhown, is applied to the analog input processor 40 which is comprlsed o~ an audio ~ain control circuit 70 and a low pass llter 72 coupled to the output thereof, The analog input speech thus processed is fed to the analog to di~ltal converter 42 where it is converted into a 12 bit digital form and red serially lnto the bufrer store 44 ~hich is adzpted to accommodate _!- _ ~ ^ ~1398~
256 12-blt words Or storage durlng each rrame interval At the start Or each frame interval, l e. each 20 8 mllll~econds or 30 milliseconds depending upon the system updating rate, the 256 samples contained in the bufrer store 44 are fed ~nto the speech pre-processor 48 which i5 comprised Or a circult ror normali~lng the lnput data 74 and a circuit 76 for implementing the well known Hanning weightlng runction The 256 samples outputted from circuit 76 are fed into the 256- sample working storage area, not shown, Or the real part input Rk(1) Or the FFT sectlon 78 of the computer means 26. The even/odd (E/0) separator section o~ the computer means 26 is designated by rererence numeral 80.
The input da~a normalizer 74 o~cratca ~y slli~tln~ an entire set of 256- ~nput samples lert or right until the most significant bit of the largest sample is typically in the 9th bit position. The sample set is then scaled up by a maximum gain Or four, or scaled down by a maximum attenuation of four.
The scaling ~actor is stored for later use in the quantizer section 54 which removes the normalization in the cepstrum before transmission.
In a manner disclosed in the above rererenced U.S.
Patent 3,681,530 the FFT section 78 next computes the FFT
trans~orm Or 256 complex samples in each time rrame. The real part input Rk( ) corresponds to 256 speech samples to be analyzed and transmitted. The imaginary part input Ik(l) corresponds to 256 samples generated in the previous time frame and coupled to the spectrum processor section 50 which provides the even functions o~ the logarithm of the spectrum magnitude Or the previous 2$6- input sample set.
3o The FFT complex output sample set Rn(2) and In( ) is _ ~ . _ . ..
~ 113~B84 odd/even (E/0) separated to provide two separate transrorms, one of whlch corresponds to the spectrum Or the input speech slgnal and is denoted as the "n-th spectrum" It is comprised of a real part~ n and an imaginary part~ n ~he other transform consists Or the (n - l)th cepstrum and is comprised Or the imaginary part~ c and is derived rrom the previous input sample aek fed back to the Ik input. Althou~h this type of cornpu-tation is disclosed in detail in U.S Patent 3,681,530, a brief generalized statement of this technique will now be presented.
The FFT section takes an N sample, complex inpùt vector Sk(l) and computes an N sample complex output vector Sn(2) in accordance with the discrete complex Fourier transform relation:
S(2) = ;~ S(l) e j2,~ N (1) where k = 0, l, 2, 3, . N-l, and n = 0, l, 2, 3, .. N-l The complex input vector Sk(l) has a real part Rk(l) ¦and an imaginary part Ik(l) so that:
S(l) = R(l) + jI(l) (2) ~he Rk( ) lnputs are thus the samples Or the input speech waverorm and is comprised Or the sum of the odd and even part o~ the N samples Figure 7 exempliries such a waverorm Each sample Or Rk( ) is stored in one Or N storage locations in the FFT section 78 with the k-th sample of Rk(l) being in the k-th location, k = 0, l, 2, 3, ... N-l.
S~nce Rk(l) ls the sum Or both even and oàd functlons, it will have both a non-zero discrete cosine transrorm ~ n as wel~
as a non-zero discrete sine transrorm ~ n. Q n and ~ n are respectlvely samples of the real and imaginary parts Or the 3o di9crete Fourier transform Or the analyzed speech waverorm Rk(l) ~ ~ l~a~?H~
and they are two Or the outputs derived rrom the comblned operation Or the FFT section 78 and the E/0 sectlon 80 of the computer means 26, The lnput Ik~l) which is the imaginary lnput vector to the FFT section 78 is contained ln another set Or N
storage locations where agaln the k-th sample Or Ik belng.ln the k-th location, In the sub~ect system, the even part Or Ik( ) is made to be equal to the logarithm Or the magnitude Or the speech slgnal to be transmltted which ~as spectrum analyzed in th~
immediately past operation Or the FFT section 78, The odd part Or the I~(l) is unused.
The even part Or Ik(l) which is outputted from the even ~unctlon generator 82 corresponds to l/2(Ik(l) + I(l)N_k) which for a volced speech lnput typlcally appears as the sampled runction shown in Figure 8, The term l/2(Ik ~ IN_k) comprlses the logged spectrum magnitude Or the signal to be transmitted and is a purely even function Or k, centered around k = N/2.
It wlll have a non-zero discrete cosine transform Cn and an ldentically zero discrete.sine transform, Thus, the Fourier . transform Or l/2(Ik ~ IN_k) is Cn and comprises ;one of the outputs obtained rrom the ~/0 section 80, A typical Cn runction is shown in Figure 9, These points desi~nate samples o~ the cosine transrorm o~ the logarithm Or the magnitude spectrum o~
the input speech signal, The function Cn is "even" and is referred to as the "cepstrum" Or the input speech signal samples Rk(l). The samples Or Cn for n = 0, l, 2, 3, ... 18~
are used as the æpectrum envelope inrormation to be transmitted and as such ls coupled to the quantizer section 54 and to a pitch detection and the voicing logic section 86, Returnlng now to the discrete Fourier transrorm set 3o rorth ln equation (2), the rollowing is intended to demonstrate il;~9~!i84 exactly how each of the inputs and outputs discussed above is obtained from the DFT trans~orm, Fir~t, the identity ( ) ( ) is substituted into equation (1), obtaining:
S( ) = ~ ~R(l) cos ( 7TN ) ~ iRk sin ( N~) + jI(1) cos ( N~) + Ik ( N )~
At the end Or each data pass or operation of the F~l section 78, the output complex vector Sn(2~ appears in t~Jo sets Or N storage locations each. One set Or N memory locations contains the real part Or Sn(2) given by:
n k~O ~ R~ cos ( N ) + I( ) sin (2 N~X) ¦ (5) The other set of N memory locations contains the imaginary part Or Sn(2) glven by:
n k~O l k sin ( N ) ~ Ik ) cos (2rN~k) ¦ (6) In each case the two sets Or registers are numbered n = O, 1, 2, 3, ... N-l.
The ~/0 separator section 80 operates on Rn(2) to produce the even and odd parts Or Rn(2). The even part corres-ponds to:
ll~ N~ ) os (27Ink) = ~ (7) ~n comprises the real part Or the spectrum Or Rk(l), the analyzed ~peech waverorm. The odd part Or Rn(2) corresponds to:
Rn RN n = ~ ) sin ( ~ ) = (8) _ _ ~' 11~
whlch is the unused part of the transrorm. The E/0 separator section 80 similarly operates on In(2) to produce the even part of In(2) which corresponds to:
` k~0 k ( ~ ) n (9) where Cn is the cepstrum. The odd part Or In(2) corresponds to: 2 = ~0 Rk sin ( N ) gn (10) ~ n comprises the imaginary part Or the spectrum Or Rk(l), the analyzed speech waveform.
Thus there are basically four parts of the FFT output as given by equations 7, 8, 9 and 10. Equations 7 and 8 provide the real and imaginary part, respcctively, Or the ~`our1er spcctru of the speech input waveform Rk(l) which is to be analyzed and transmitted while quation 9 provides the cepstrum function.
As noted above, the Ik(l) input to the FF~ section 78 comprises the even function of the logged spectrum magnitude Mn at the n-th rrequency. Mn is defined as:
Mn = ~n ~ gn (11) where Qn and ~ n are derined in equations 7 and 10. In practicality, inasmuch as binary numbers are being dealt with, it is more desirable to use the approximation (MAX + 1/4 MIN) to avoid squaring the numbers ~ n and ~ n. This can be mathematically stated as rOllOws:
M = max ~ n~ n¦} + 4 min {¦~ nl~ }
Equation 12 indicates that either the magnitude of ~n or the magnitude -~ n, whichever is greater, ls taken to which is 3o added one rourth r~n¦~ n¦whichever is less. In general ~ _ _ .. . .
other approximations such as the 1/4 may be replaced by 1/2 may be utili2ed without changing the general idea; however, the 1/4 ractor is simply the ractor which gives the best approxlma-tion to ~ ~ n2 + ~ n2 Apparatus for performln~ thls runction is designated by reference numeral 88 in the spectrum processor section 50 In addition to being fed to the logging algorlthm computer 84, the spectrum magnitude of the n-th frequency is fed to the pitch extractor section 52 and respectlvely to three portions thereor, namely an energy sum noise stripper 90, a low band sum circuit 92, and a high band sum circuit 94. ~he noise stripper 90 is utilized in order to increase the accuracy Or the voicing under heavy acoustic bac~ground nolse conditions and provides a noise threshold which establishes a minimum level for the low band and the high band, for example, durlng 4 second intervals. Since it is highly unlikely that speech would be sustained for greater than 4 second intervals, the mlnlma can be considered measurements Or the magnitude Or noise in the two bands Throughout each 4-second estimation perlod, the noise estimates rrom the previous estimation period are subtracted from the high band and low band energy sums. The circuits 92 and 94 comprise an implementation Or the simple 9ubtractlon algorithm which is a crude approximation o~ the ~ollowlng pair O r expressions .
SL0 = ~MLo ~ Nlo (13.) SHI = ~ MHI - NHI (14) .
where S i3 the speech energy sumJ M is the measured energy sum, 3o and N is the esti~ated noise energy sum. These equatlons are derived from the relationship:
.
. ` 1139W~
M2 s2 N2 ( 15 Accordingly the noise stripper operates to improve the perfor-mance o~ the pitch extractor 52 The outputs Or the su~ners 92 and 94 are logged in apparatus denoted by reference numerals 96 and 98 which corres-pond to the logging algorithm cornputer 84, referred to above, and may be comprlsed of, ror example, a Sylvania Electrlc Products Inc., ACP-l computer wherein the log runction is approximated by ten inscribed straight lines This logged low and high band sum inrormation is then red into the pitch detection and voicing loglc circuit 86, which additionally receives an input o~ the cepstrum Cn outputted from the E/0 section 80.
Pitch detection is accomplished by rirst testing the voicing detection criterion and after voic~ng has been detected, searching the cepstrum runction in a predetermined search range ror a peak that occurs whenever the input speech is voiced.
The position Or the tracking peak ln the search range is then taken to be rundamental pitch period o~ the speaker's voice.
This concept is known and described in detail in the arore-mentioned Patent 3,681,530.
This pitch value which is extracted is fed to the modulator 56 along with quantized cepstrum values outputted ~rom the quantizer section 54 which receives a set Or cepstrum coef~icients Cn that characterize the spectral envelope Or the lnput speech data. Quantizer 54 acts to remove normalization gain of the cepstrum coefficients after which they are scaled, quantized and fed to the modulator 56.
The quantizer 54 is designed as a dual updating rate quantizer and operates on the theory that the hearing process is not sensitive to fine detail in the spectrum shape during 3o rapld transition, i.e. when the spectrum shape is changing 11~
¦rapidly. At these times lt is important to ralthrully transmlt ¦the rapld changes of the general spectrum shape by sampling those coefficients which determine the general ~hape and in particular the overall energy level as often as possible. Indeed at times such rapid change in spectral content the usual assumptions quazi-stationarity in the speech spectrum breakdown completely.
The theory rurther states that during relatlvely stable sounds such as near the middle Or vowels, the hearing mechanism wlll be quite sensitive to spectrum envelope detall At such tlmes, however, the spectrum envelope ls changing relatively slowly so that the higher order coerricients which convey the required spectral detail may be sampled at a lower rate.
The cepstrum vocoder concept set rorth above is ideally suited to take advantage Or these af~ects since the very low order coerricients, i.e. the first four or five cepstrum valuesJ
convey the general spectrum envelope shape with the zero delay value; Or course, carrying the overall gain or energy level.
The higher order coer~icients rrom the rourth or firth co-erricient out to about 3 milliseconus define the final spectrum envelope in detail.
Thus advantage can be taken Or the e~rects postulated in the above dlscussion in a simple way by samplin~ the ~irst rour or rive cepstrum values at the hlghest practical rate, i e every rrame, while the rest Or the spectrum coerricients are sampled on an interleaved basis every other rrame. This automatlcally provldes the most rrequent possible sampling Or the overall spectrum shape during intervals Or rapidly c~anging spectral content together with many higher order spectrum coerficlents sampled at a lower rate to provide faithrul repro-duction of rine spectrum envelope detail during the more nearly 8 teady 8 tate sounds.
~J ~ .
The modulator section 56 accordingly lncludes a rrame rormatter Or the quantized cepstrum and pitch parameters whlch are repetitively updated ror example 50 tlmes per second, into a multi-bit digital word containing, for example, 7 blts Or pltch and 41 bits Or cepstrum information ln the 2400 bit per second mode and 6 bits Or pitch and 30 bits Or cepstrum ln the 1200 blt per second mode. The data thus rormatted is red to a phase shirt control. unit 102 which is coupled to a table look-up modulator 104. The phàse control unlt 102 and the table look-up modulator operate to generate in digital form a ~irst and second plurality Or tones in a four phase DPSK operational mode as de~-cribed ln the above rererenced U.S. Patent 3~617J941. Table I
discloses a typical tone gen~ration output for a 25 tone look-up modulator which accepts input data at a rate of 2400 bits per second. These direct and heterodyned sampled waverorms generated in the table look-up modulator 104 are outputted to the dlgltal analog converter 58 which generates and holds the analog output value Or the tone rrequency until the next rrequency sample ls ed thereto The lowpass rilter 60 coupled to the D/A converter ~ 58 is used to eliminate unused tones generated by the table look-up modulator 104. Thus the transmltted slgnal comprises an analog modem waverorm generated rrom a 25 tone stack `
Rererring now to the receive mode and Figure 6, modem lnput signals rrom a communications channelJ not shown, are red lnto the aforementioned analog input processor 40 ln-cludlng the gain control circuit 70 and low pass rllter 72 and then to the analog to digital converter 42. The digitized modem input is red to the same 128 sample bur~er storage 44 as berore, however, it is now also red into the sync detector sectlon 46 ¦ which consists Or a dirrerence equatlon rilter 106 and a baud synchronization circuit 108. The input tlmlng provided by the ¦sync detector 46 is shown in Figure 10 ~or a 2 baud inter~al The 128 sample burrer data in the burrer store 44 is taken - - 20 _ '' : ~
l ~A~
¦Modem fS/2 GeneratiOn Heterodyne Flnal Output ¦ Tone ~ Frequency Frequency Frequency ¦ 1 10 562.5 Hz ~ 562,5 Hz
Vocoder systems function to transmit speech signals in a coded manner to reduce the transmission bandwidth which would other- I
wise be required if the speech was to be transmitted in an uncoded manner. Thus a vocoder system includes a transmit terminal to analyze the characteristics of the speech wave to be encoded and to encode the speech wave and a receive terminal which is used to synthesize, from the coded signal set to it, a reconstruction of the original speech wave. Dats modems on the other hand function to facilitate the transmission of data, for example data from a speech vocoder, over a transmission medium. Thus a modcm includes a trunsmil tcrminal to convert the encoded data into a modulating signal which is used to modulate a carrier frequency, and a receive terminal to demodulate the received signal and thereby recover the transmitted data. Both vocoder and modem equipment are therefore required for transmission of speech signal in an efficient manner.
A known system of particular interest is disclosed in U.S. Patent 3,681,530, entitled "Method and Apparatus for Signal Bandwidth Compression Utilizing The Fourier Transform Of The Logarithm Of The Frequency Spectrum Magnitude", issued to H.~. Manley, et al.
There a system is described wherein an input speech waveform i9 con-verted into an electrical signal which is then digitized by an analog to digital converter. Following this the digitized signal is directed through a device where magnitudes of the frequency spectrum of the input speech wave are obtained. These magnitudes are then directed to a logging circuit to obtain a logarithm of the frequency spectrum magnitudes of the input speech signal. The log magnitudes of the frequency spectrum MR/
~39884 are then directed to a computer where the discrete Fourier transform (DFT) of the log spectrum magnitudes are obtained to form what is commonly referred to as the "cepstrum" of the input speech signal. The system also includes an analysis section of the vocoder terminal which operates to decode received data and separate it into pitch data and vocal track impulse data. Also included is a computing device for computing the logarithm of the spectrum envelope of the vocal track impulse response function using FFT techniques. A convolution unit then convolves the pitch data with the impulse response data to yield the desired synthesized speech signal.
A digital modem adapted to be utilized in conjunction with the system described in U.S. Patent 3,681,530 is disclosed in U.S. Patent 3,617,941, entitled "Table Look Up Modulator", which issued to J. DeLellis, Jr., one of the subject inventors, and which discloses a differential phase shift keying (DPSK) system for generating in digital form a plurality of tones at a predetermined baud rate.
Additionally a digital processor which can use the same hardware to implement both the vocoder and modem functions is disclosed in U.S. Patent 3,706,929, entitled "Combined Modem and Vocoder Pipeline Processor", by J.L. Robinson, et al., which describes a processor which operates as a half duplex system, i.e. either in a transmitting or receiving mode. In the "transmit" mode, the input signal to the processor is a speech wave and the processor perform the vocoder function of speech wave analysis and the modem function of generating a modulating ~-r ~h f~ ~13~B~
signal which carrie~ the results Or the speech wave analysls.
Thls modulating signal becomes the output signal Or the processor ln this mode, which is used ln the modulator Or a conventlonal communications transmitting ~ystem, In the "receive" mode the input signal to the processor is a speech information bearing communlcations signal such as may be derived from a conventional communications receiver, The same t~pe Or modulation used ln the transmlt mode mu~t be used in this mode. The processor performs the modem runction Or demodulating o~ the input signal and the vocoder runction Or the synthesis Or the speech wave.
The output signal Or the processor in this mode is then used to drive conventional volce reproduction circuitry.
Acoustic si~nal processors employing F~ techniques are also disclosed in U,S. Patent 3,662,108, entitled "Apparatus ~or Reducing Multi-Path Distortion Or Signals Utilizing Cepstrum Technique", by J,L, Flannigan and U,S, Patent 3,634,668, entitled "Log Spectrum Decoding Utilizing Symmetry", by H,L, Sharfer, Accordingly, an object of the present invention is to provide a system including a digital processor which is particu-larly adapted for voice communications systems, Another obJect Or the present invention is to provlde a system including a digital processor which ls adapted to runc-tion as both the vocoder and modem in a voice security communl-cations system, It i9 a rurther ob~ect Or the present invention to provide a system includlng a digital signal processor which combines the function Or both a vocoder and a modem in a halr duplex mode Or operation, Summary 3o Disclosed is a vocoder modem system having a digital signal processor which implements a Fourier transrorm (FFT) ~` ~
¦algorlthm used rOr both vocoder and modem processing in both the transmit and receive modes. In the transmlt mode, a 256 point ¦(sample) FFT algorithm erfecting a complex DFT is used to simul-taneously perrorm both spectrum and cepstrum analysis for pitch detectlon and cepstral encodlng of input speech signals. The pltch and cepstral parameters provide the input to a rour phase dirferentlally coherent phase shirt keying (DPSK) table loolc-up modulator which generates 25 digltal modem tones which are converted to the audio band and transmitted over a wire line or high ~requency llnk. In the receive mode, a 64 point complex FFT algorithm is used to perrorm a 12~ point, real input DFT
on the modem input which is ~irst used to demodulate the tone stack Or the input. The same 64 point F~r algorithrn is then used again to transform the decoded cepstral parameters into a log spectrum and to an impulse response which is used with pitch inrormation to generate an output speech waverorm ln a syntheslzer.
Synchronization is provided by an iterative dirrerence equation rilter centered on an empty slot ln the tone stack, Brier Description Or the Drawings These and other ob~ects of the present invention will become more fully apparent from the followlng detailed descriptio taken in connection with the following drawings ln whlch:
Figure 1 is a slmplified block diagram of a hal~ duplex vocoder modem;
Figure 2 is a graph illustrative of the general appear-ance o~ a logged magnitude spectrum;
Figure 3 i~ a graph illustrative of the Fourier trans-form o~ the log spectrum magnitude shown in Figure 2;
Flgure 4 is a block diagram illustrative of the pre-3o rerred embodiment of the sub~ect invention;
~ ~ ~ il39W~
l Figure 5 lq a detailed block diagram Or the portlon ¦of the embodiment shown in Figure 4 utilized ln the transmlt ¦ mode;
¦ Figure 6 ls a detalled block diagram illustrative Or ¦the portion of the embodiment shown in Figure 4 utilized in the ¦ receive mode;
Figure 7 is a graph illustrative Or the samples Or the input speech ~Javeform red lnto the FFT computer portion of the l apparatus shown in Figure 5;
LO Figure 8 is a graph illustrative Or the logged spectrum magnitude Or the samples shown in Figure 7;
Figure 9 is a graph illustrative of the cepstrum of the samples shown in Figures 7 and 8;
Figure 10 is a graph illustrative o~ the data input tlming dlagram for the receive mode of operation;
Figure 11 is a graph illustrative of the synchroniza-tion alignment ror the system during the receive mode;
Figure 12 is a graph illustrative Or the cepstrum sample~ ror a modem signal received by the subJect invention;
O Figu~e 13 is a graph illustrative of the sampled received spectrum magnitude o~ the modem signal received; and Figure 14 is a graph il1ustrative Or the sampled syn-thesizer implllse response o~ the received modem signal, Description Or the Prererred Embod~ments A simplified block di.agram Or the present inventlon 1~ shown in Figure 1. In the "transmit" mode, the input speech i8 sampled and digitized in analog to digital converter means 10 and red into a computer (digital data processor) 12 which analyses the speech in the spectral and pltch parameters. These ,0. parameters are quantized into a multi-bit data stream which acts ~ i~9W~ ' as an input to a dif~erential phase shift lceying (DPSK) modulator 14 The DPSK modulator 14 constructs a multi-tone modem slgnal whlch i3 then converted into an analog signal by digital to ana-log means 16 and ~ed as a transmitted modem waverorm into a communlcations channel. In the "receive" mode, the same analog to digital converter means 10 accepts a modem input signal and outputs a digital data stream wh~ch is red to a DPSK demodulator 18 which separates the pitch and spectral coefricients which are then used by a synthesi~er 20 to reconstruct the input speech in digltal rorm. The output speech is then converted in the digital to analog converter means 16 which was used to produce the line signal when the system operated in the transmit mode.
The sub~ect invention is an outgrowth of thc ~ r-cepstrum vocoder technique which has been described, for example, in the above rererenced U S. Patent 3,681,530. This patent, moreover, discloses a complex 256-point fast Fourier translorm algorithm hereinafter referred to simply as FFT, which is used to simultaneously generate the spectrum magnitude and cepstrum Or an input speech waveform. In the FFT~cepstrum approach, a 256 sample segment Or input speech is transrormed to produce spectral magnitude information at 128 frequencies across the speech band This i~ surricient to resolve the pitch harmonics Or voice sounds so that the spectrum envelope may be extracted without the confounding Or the envelope and pitch inrormation that occurs, ror example in a channel vocoder. The resulting measured magnltude spectrum is the product Or the spectrum envelope with pitch harmonic line spectrum. This product is separated into a sum by taking the logarlthm o~ the high resolu-tion magnitude spectrum. The general appearance o~ such a high 3o resolution, logged magnitude spectrum is snown in Flgure 2.
il3981~
The cepstrum or FFT of this logged spectrurn is shown ln Figure 3.
The transform o~ the slowly varying part of the spectral envelope appears in the low "que~rency" or low delay part of the cepstrum below 3 milliseconds. This low delay portion is selected ror quantization and transmission the er~ect of which is to remove the pitch harmonic ripple in the high resolution, logged magnitude spectrum by low pass ril~ering rrhis method, both theoretically and experirnentally, yields the most accurate possible smooth spectrum envelope description with no confounding Or pitch and spectrum envelope inrormation.
The spectrum parameters obtained by the FFT-cepstrum vocoder are nearly uncorrolated with each other, i.e. rrom channel to channel, and thererore are in an efricient representation Or spectrum shape from a data reduction vicwpoint. They are also in an ideal form to enable the system to take advantage Or corro-lations rrom ~rame to rrame, i.e across time Therefore, an interlacing technique is employed in which the low delay cepstrum values which characterize the gross spectrum shape are thus trans-mltted every ~rame and the higher order coefficients are trans-mitted every other rrame is utilized by the quantizer Pltch deteotion ln the FFT-cepstrum approach i9 accomplished by detect-ing and tracking the cepstral peak which occurs in the high "querrency" or high delay part o~ the cepstrum during voiced sound9. The time delay to this peak is the pitch period. The cepstral peak can be thought of as the "frequency" of the rlpples in the log spectrum magnitude Voiclng is detected in the cepstrum pitch extractor by reserving both the ratio Or low band to high band spectral magnitude energy and by observlng the presence Or a signiricant cepstral peak The present invention as wlll be shown employs a l;
il3~8t~4 single N point, radix-4 FFT algorithm which is capable Or operat-ing either in the transmit or the receive mode~. In the transmit mode an N=256 point complex FFT operation is performed while ln the receive mode an N~64 point complex ~FT ls utllized to perrorm a 2N=128 point real input discrete Fourier transform which allows the use of the same coding in each instance Rererring now to Figure 4, there is disclosed a com-posite block diagram o~ the present invention. What is si~niri-cant is the utilization o~ the same input and output sections 22 and 24, respectively, with common FFT and even/odd separator computer means 26 in both the transmit and receive modes o~
operation. This is carried out in accordance with the synchronous operation Or a plurality of signal switching elements shown in Figure ~ as the transmit/receive (T/R) switch blocks 28, 30, 32, 34, 36 and 38. As shown, the analog speech input and the analog modem input are both coupled to the switch block 28 which alternately operates in the transmit mode and receive mode, respectively, to couple the respective analog waveforms to an analog lnput processor 40 whose output is coupled to an analog to digital (A/D) converter 42. The A/D converter 42 operates to converk the signal inputted thereto to a 12 bit digital data ~tream whioh i~ red into a data bufrer store 44 whlch is adapted to accommodate 256 12-bit words in storage at a tlme. In the receive mode, the 12-bit data stream is additionally fed to a sync detector block 46 which operates to control the data bufrer over two baud intervals in a manner to be discussed subsequently In the transmit mode, the T/R block 30 operates to couple the digitized speech input into a speech pre-processor 48 where it i~ then red to ~he FFT and even/odd separator computer 3o means 26 through the T/R block 32. The FFT and E/0 separator v 1~
computer means 26 is coupled to a transmit ~pectrum processor 50 and to a pitch extractor 52 through the actlon of the T/R
block 34. The output Or 3pectrum processor 50 whlch comprlses the logged spectrum Or the FFT samples is coupled back to the input o~ the FFT apparatus 26 in a manner as disclosed ln the arorementioned U,S. Patent 3,681,530 to perrorm the Fourier transrorm Or the log Or the speech spectrum (cepstrum) whlch signals are outputted to a quantizer 54 along with the output Or the pitch extractor 52. The quantized cepstrum and pitch signals are fed to a DPSK modulator 56 which forms a digital modem signal which is coupled to an output digital to analog converter 58 through the T/R block 36. The analog modem signals outputted rrom the converter 58 are fed to the transmlsslon channel via an analog output signal processor (low pass filter) 60 and the T/R block 38.
In the receive mode, the analog modem input is digitlzed into a data stream and ~ed via the T/R blocks 30 and 32 to the FFT and ~/0 computer means 26, wherein a ~irst pass o~ the data stream is made through the computer means 26 in a 64 point complex FFT routine in order to perform a 128 point real lnput DFT
transrorm, The results Or the rirst pass is outputted to a demodulator 62 where ceRstrum signals are developed. The~e signals are ~ed back into the computer mean9 26 through a signal summer 64 and the T/R device 32, Again a 128 point real input DFT trans~orm is formed by means Or a 64 point complex FFT
routine. A ~irst portion (lmaginary) o~ the second pas9 compu-tation is fed as an impulse response signal to a syntheslzer 66 whlle a second portion (real) is ~ed to a recelve spectrum processor 68 which provides a feedback path back into the FFT
3o ~or second pass operation Or the FFT Or the data. This operatlon will be considered further when Figure 6 is considered. The synthesizer 66 also receives pitch in~ormation ~rom the demodu-lator 62 along with the impulse response signals developed ~3~3884 during the second 64 point calnplex Fi~r operation to provl~e digltlzed speech waveforms which are then coupled to the digltal to analog converter 58 through the T/R device 36. The analog speech waverorms are then fed to the analog output processor 66 and to a transducer, not shown, through the T/R device 38.
In order to more rully understand the operation Or the subject invention, the separate portions of the system shown in Figure 4 are separated and expanded in the rcspective transmit and receive mode apparatus shown in Figures 5 and 6.
Referring now to Figure 5, shown is a detailed electrical block diagram illustrative Or the means ~or implementing the digital processing which is performed in a transmit mode Or operation, The heart Or the speech analysis portion Or the system is a 256-point, radix-4 FFT algorithm which is shown being performed by the block 26 which operates simultaneously to generate the spectrum and cepstrum Or the speech input. The modulation section 56 runctionally implements a table look-up routine, a typical example Or which is disclosed in the above-rererenced U.S. Patent 3,617,941, and is operable to generate four phase DPSK stacked tones, The FFT routine is adapted to output data in rrames at a system updating interval which is typically every 20,8 milllseconds ror a 2400 bit pcr second (b/~) mode and every 30 milliseconds ror a 1200 b/s mode, The lnput speech waveform coupled rrom an input channel, not æhown, is applied to the analog input processor 40 which is comprlsed o~ an audio ~ain control circuit 70 and a low pass llter 72 coupled to the output thereof, The analog input speech thus processed is fed to the analog to di~ltal converter 42 where it is converted into a 12 bit digital form and red serially lnto the bufrer store 44 ~hich is adzpted to accommodate _!- _ ~ ^ ~1398~
256 12-blt words Or storage durlng each rrame interval At the start Or each frame interval, l e. each 20 8 mllll~econds or 30 milliseconds depending upon the system updating rate, the 256 samples contained in the bufrer store 44 are fed ~nto the speech pre-processor 48 which i5 comprised Or a circult ror normali~lng the lnput data 74 and a circuit 76 for implementing the well known Hanning weightlng runction The 256 samples outputted from circuit 76 are fed into the 256- sample working storage area, not shown, Or the real part input Rk(1) Or the FFT sectlon 78 of the computer means 26. The even/odd (E/0) separator section o~ the computer means 26 is designated by rererence numeral 80.
The input da~a normalizer 74 o~cratca ~y slli~tln~ an entire set of 256- ~nput samples lert or right until the most significant bit of the largest sample is typically in the 9th bit position. The sample set is then scaled up by a maximum gain Or four, or scaled down by a maximum attenuation of four.
The scaling ~actor is stored for later use in the quantizer section 54 which removes the normalization in the cepstrum before transmission.
In a manner disclosed in the above rererenced U.S.
Patent 3,681,530 the FFT section 78 next computes the FFT
trans~orm Or 256 complex samples in each time rrame. The real part input Rk( ) corresponds to 256 speech samples to be analyzed and transmitted. The imaginary part input Ik(l) corresponds to 256 samples generated in the previous time frame and coupled to the spectrum processor section 50 which provides the even functions o~ the logarithm of the spectrum magnitude Or the previous 2$6- input sample set.
3o The FFT complex output sample set Rn(2) and In( ) is _ ~ . _ . ..
~ 113~B84 odd/even (E/0) separated to provide two separate transrorms, one of whlch corresponds to the spectrum Or the input speech slgnal and is denoted as the "n-th spectrum" It is comprised of a real part~ n and an imaginary part~ n ~he other transform consists Or the (n - l)th cepstrum and is comprised Or the imaginary part~ c and is derived rrom the previous input sample aek fed back to the Ik input. Althou~h this type of cornpu-tation is disclosed in detail in U.S Patent 3,681,530, a brief generalized statement of this technique will now be presented.
The FFT section takes an N sample, complex inpùt vector Sk(l) and computes an N sample complex output vector Sn(2) in accordance with the discrete complex Fourier transform relation:
S(2) = ;~ S(l) e j2,~ N (1) where k = 0, l, 2, 3, . N-l, and n = 0, l, 2, 3, .. N-l The complex input vector Sk(l) has a real part Rk(l) ¦and an imaginary part Ik(l) so that:
S(l) = R(l) + jI(l) (2) ~he Rk( ) lnputs are thus the samples Or the input speech waverorm and is comprised Or the sum of the odd and even part o~ the N samples Figure 7 exempliries such a waverorm Each sample Or Rk( ) is stored in one Or N storage locations in the FFT section 78 with the k-th sample of Rk(l) being in the k-th location, k = 0, l, 2, 3, ... N-l.
S~nce Rk(l) ls the sum Or both even and oàd functlons, it will have both a non-zero discrete cosine transrorm ~ n as wel~
as a non-zero discrete sine transrorm ~ n. Q n and ~ n are respectlvely samples of the real and imaginary parts Or the 3o di9crete Fourier transform Or the analyzed speech waverorm Rk(l) ~ ~ l~a~?H~
and they are two Or the outputs derived rrom the comblned operation Or the FFT section 78 and the E/0 sectlon 80 of the computer means 26, The lnput Ik~l) which is the imaginary lnput vector to the FFT section 78 is contained ln another set Or N
storage locations where agaln the k-th sample Or Ik belng.ln the k-th location, In the sub~ect system, the even part Or Ik( ) is made to be equal to the logarithm Or the magnitude Or the speech slgnal to be transmltted which ~as spectrum analyzed in th~
immediately past operation Or the FFT section 78, The odd part Or the I~(l) is unused.
The even part Or Ik(l) which is outputted from the even ~unctlon generator 82 corresponds to l/2(Ik(l) + I(l)N_k) which for a volced speech lnput typlcally appears as the sampled runction shown in Figure 8, The term l/2(Ik ~ IN_k) comprlses the logged spectrum magnitude Or the signal to be transmitted and is a purely even function Or k, centered around k = N/2.
It wlll have a non-zero discrete cosine transform Cn and an ldentically zero discrete.sine transform, Thus, the Fourier . transform Or l/2(Ik ~ IN_k) is Cn and comprises ;one of the outputs obtained rrom the ~/0 section 80, A typical Cn runction is shown in Figure 9, These points desi~nate samples o~ the cosine transrorm o~ the logarithm Or the magnitude spectrum o~
the input speech signal, The function Cn is "even" and is referred to as the "cepstrum" Or the input speech signal samples Rk(l). The samples Or Cn for n = 0, l, 2, 3, ... 18~
are used as the æpectrum envelope inrormation to be transmitted and as such ls coupled to the quantizer section 54 and to a pitch detection and the voicing logic section 86, Returnlng now to the discrete Fourier transrorm set 3o rorth ln equation (2), the rollowing is intended to demonstrate il;~9~!i84 exactly how each of the inputs and outputs discussed above is obtained from the DFT trans~orm, Fir~t, the identity ( ) ( ) is substituted into equation (1), obtaining:
S( ) = ~ ~R(l) cos ( 7TN ) ~ iRk sin ( N~) + jI(1) cos ( N~) + Ik ( N )~
At the end Or each data pass or operation of the F~l section 78, the output complex vector Sn(2~ appears in t~Jo sets Or N storage locations each. One set Or N memory locations contains the real part Or Sn(2) given by:
n k~O ~ R~ cos ( N ) + I( ) sin (2 N~X) ¦ (5) The other set of N memory locations contains the imaginary part Or Sn(2) glven by:
n k~O l k sin ( N ) ~ Ik ) cos (2rN~k) ¦ (6) In each case the two sets Or registers are numbered n = O, 1, 2, 3, ... N-l.
The ~/0 separator section 80 operates on Rn(2) to produce the even and odd parts Or Rn(2). The even part corres-ponds to:
ll~ N~ ) os (27Ink) = ~ (7) ~n comprises the real part Or the spectrum Or Rk(l), the analyzed ~peech waverorm. The odd part Or Rn(2) corresponds to:
Rn RN n = ~ ) sin ( ~ ) = (8) _ _ ~' 11~
whlch is the unused part of the transrorm. The E/0 separator section 80 similarly operates on In(2) to produce the even part of In(2) which corresponds to:
` k~0 k ( ~ ) n (9) where Cn is the cepstrum. The odd part Or In(2) corresponds to: 2 = ~0 Rk sin ( N ) gn (10) ~ n comprises the imaginary part Or the spectrum Or Rk(l), the analyzed speech waveform.
Thus there are basically four parts of the FFT output as given by equations 7, 8, 9 and 10. Equations 7 and 8 provide the real and imaginary part, respcctively, Or the ~`our1er spcctru of the speech input waveform Rk(l) which is to be analyzed and transmitted while quation 9 provides the cepstrum function.
As noted above, the Ik(l) input to the FF~ section 78 comprises the even function of the logged spectrum magnitude Mn at the n-th rrequency. Mn is defined as:
Mn = ~n ~ gn (11) where Qn and ~ n are derined in equations 7 and 10. In practicality, inasmuch as binary numbers are being dealt with, it is more desirable to use the approximation (MAX + 1/4 MIN) to avoid squaring the numbers ~ n and ~ n. This can be mathematically stated as rOllOws:
M = max ~ n~ n¦} + 4 min {¦~ nl~ }
Equation 12 indicates that either the magnitude of ~n or the magnitude -~ n, whichever is greater, ls taken to which is 3o added one rourth r~n¦~ n¦whichever is less. In general ~ _ _ .. . .
other approximations such as the 1/4 may be replaced by 1/2 may be utili2ed without changing the general idea; however, the 1/4 ractor is simply the ractor which gives the best approxlma-tion to ~ ~ n2 + ~ n2 Apparatus for performln~ thls runction is designated by reference numeral 88 in the spectrum processor section 50 In addition to being fed to the logging algorlthm computer 84, the spectrum magnitude of the n-th frequency is fed to the pitch extractor section 52 and respectlvely to three portions thereor, namely an energy sum noise stripper 90, a low band sum circuit 92, and a high band sum circuit 94. ~he noise stripper 90 is utilized in order to increase the accuracy Or the voicing under heavy acoustic bac~ground nolse conditions and provides a noise threshold which establishes a minimum level for the low band and the high band, for example, durlng 4 second intervals. Since it is highly unlikely that speech would be sustained for greater than 4 second intervals, the mlnlma can be considered measurements Or the magnitude Or noise in the two bands Throughout each 4-second estimation perlod, the noise estimates rrom the previous estimation period are subtracted from the high band and low band energy sums. The circuits 92 and 94 comprise an implementation Or the simple 9ubtractlon algorithm which is a crude approximation o~ the ~ollowlng pair O r expressions .
SL0 = ~MLo ~ Nlo (13.) SHI = ~ MHI - NHI (14) .
where S i3 the speech energy sumJ M is the measured energy sum, 3o and N is the esti~ated noise energy sum. These equatlons are derived from the relationship:
.
. ` 1139W~
M2 s2 N2 ( 15 Accordingly the noise stripper operates to improve the perfor-mance o~ the pitch extractor 52 The outputs Or the su~ners 92 and 94 are logged in apparatus denoted by reference numerals 96 and 98 which corres-pond to the logging algorithm cornputer 84, referred to above, and may be comprlsed of, ror example, a Sylvania Electrlc Products Inc., ACP-l computer wherein the log runction is approximated by ten inscribed straight lines This logged low and high band sum inrormation is then red into the pitch detection and voicing loglc circuit 86, which additionally receives an input o~ the cepstrum Cn outputted from the E/0 section 80.
Pitch detection is accomplished by rirst testing the voicing detection criterion and after voic~ng has been detected, searching the cepstrum runction in a predetermined search range ror a peak that occurs whenever the input speech is voiced.
The position Or the tracking peak ln the search range is then taken to be rundamental pitch period o~ the speaker's voice.
This concept is known and described in detail in the arore-mentioned Patent 3,681,530.
This pitch value which is extracted is fed to the modulator 56 along with quantized cepstrum values outputted ~rom the quantizer section 54 which receives a set Or cepstrum coef~icients Cn that characterize the spectral envelope Or the lnput speech data. Quantizer 54 acts to remove normalization gain of the cepstrum coefficients after which they are scaled, quantized and fed to the modulator 56.
The quantizer 54 is designed as a dual updating rate quantizer and operates on the theory that the hearing process is not sensitive to fine detail in the spectrum shape during 3o rapld transition, i.e. when the spectrum shape is changing 11~
¦rapidly. At these times lt is important to ralthrully transmlt ¦the rapld changes of the general spectrum shape by sampling those coefficients which determine the general ~hape and in particular the overall energy level as often as possible. Indeed at times such rapid change in spectral content the usual assumptions quazi-stationarity in the speech spectrum breakdown completely.
The theory rurther states that during relatlvely stable sounds such as near the middle Or vowels, the hearing mechanism wlll be quite sensitive to spectrum envelope detall At such tlmes, however, the spectrum envelope ls changing relatively slowly so that the higher order coerricients which convey the required spectral detail may be sampled at a lower rate.
The cepstrum vocoder concept set rorth above is ideally suited to take advantage Or these af~ects since the very low order coerricients, i.e. the first four or five cepstrum valuesJ
convey the general spectrum envelope shape with the zero delay value; Or course, carrying the overall gain or energy level.
The higher order coer~icients rrom the rourth or firth co-erricient out to about 3 milliseconus define the final spectrum envelope in detail.
Thus advantage can be taken Or the e~rects postulated in the above dlscussion in a simple way by samplin~ the ~irst rour or rive cepstrum values at the hlghest practical rate, i e every rrame, while the rest Or the spectrum coerricients are sampled on an interleaved basis every other rrame. This automatlcally provldes the most rrequent possible sampling Or the overall spectrum shape during intervals Or rapidly c~anging spectral content together with many higher order spectrum coerficlents sampled at a lower rate to provide faithrul repro-duction of rine spectrum envelope detail during the more nearly 8 teady 8 tate sounds.
~J ~ .
The modulator section 56 accordingly lncludes a rrame rormatter Or the quantized cepstrum and pitch parameters whlch are repetitively updated ror example 50 tlmes per second, into a multi-bit digital word containing, for example, 7 blts Or pltch and 41 bits Or cepstrum information ln the 2400 bit per second mode and 6 bits Or pitch and 30 bits Or cepstrum ln the 1200 blt per second mode. The data thus rormatted is red to a phase shirt control. unit 102 which is coupled to a table look-up modulator 104. The phàse control unlt 102 and the table look-up modulator operate to generate in digital form a ~irst and second plurality Or tones in a four phase DPSK operational mode as de~-cribed ln the above rererenced U.S. Patent 3~617J941. Table I
discloses a typical tone gen~ration output for a 25 tone look-up modulator which accepts input data at a rate of 2400 bits per second. These direct and heterodyned sampled waverorms generated in the table look-up modulator 104 are outputted to the dlgltal analog converter 58 which generates and holds the analog output value Or the tone rrequency until the next rrequency sample ls ed thereto The lowpass rilter 60 coupled to the D/A converter ~ 58 is used to eliminate unused tones generated by the table look-up modulator 104. Thus the transmltted slgnal comprises an analog modem waverorm generated rrom a 25 tone stack `
Rererring now to the receive mode and Figure 6, modem lnput signals rrom a communications channelJ not shown, are red lnto the aforementioned analog input processor 40 ln-cludlng the gain control circuit 70 and low pass rllter 72 and then to the analog to digital converter 42. The digitized modem input is red to the same 128 sample bur~er storage 44 as berore, however, it is now also red into the sync detector sectlon 46 ¦ which consists Or a dirrerence equatlon rilter 106 and a baud synchronization circuit 108. The input tlmlng provided by the ¦sync detector 46 is shown in Figure 10 ~or a 2 baud inter~al The 128 sample burrer data in the burrer store 44 is taken - - 20 _ '' : ~
l ~A~
¦Modem fS/2 GeneratiOn Heterodyne Flnal Output ¦ Tone ~ Frequency Frequency Frequency ¦ 1 10 562.5 Hz ~ 562,5 Hz
2 12 675,0 ____ 675.0
3 13 731.25 ____ 731,25
4 14 787,5 _ _ _ _ 787.5 843.75 ____ 843,75 6 16 900.0 ---- 900.0 7 17 956,25 ____ 956,25 8 18 1012,5 - - -- 1012,5 9 19 106~,75 --~- 1068,75 1125,0 ---- 1125,0 11 21 1181.25 ---- 1181.25 12 22 1237,5 _ _ _ _ 1237.5 13 23 1293.75 ---- 1293.75 14 25 1406.25 __ _ 1406.25 26 1462.5 ____ 1462,5 16 27 1518.75 ---- 1518.75 17 10 562.5 2137.5 Hz 1575.0 18 9 506.25 2137.5 1631.25 19 8 450.0 2137.5 1687.5 7 393.75 2137.5 1743.75 21 6 337.5 2137,5 1800,0 22 5 281,25 2137,5 1856,25 23 4 225,0 2137,5 1912,5 24 3 168,75 2137,5 1968.75 Z 1~2.5 2137,5 2025,0 ~`
~ `:
i~39~8~
~ . . .
from the center Or the baud so that the input data is lsolated as much as posslble rrom the baud transition. The data ln this 128 sample orthogonal window is trans orlned once per baud to provide the Fourier analysis Or the input tones which ls necessary ~or demodulation.
The interval designated D~ input salnplc~ ln l~ ure 10 illustrates the gating Or input sarnples to the dirrerence equa-tion rilter 106 In each 2 baud interval, the last 128 samples o~ the first baud and the rirst 128 samples of the second baud are fed into an integrate and dump dirference equation imple-mented thereby. The algorithm used to process the synchroniza-tion input data is represented by the rollowing dif~erence equation:
Yk+1 ~ 2(cos ~pT) Yk+Yk-1 gk (16) where ~p is the radian center rrequency of the rilter, T ls the sampling period, and gk is the input. This dirference equation has a solution: -Yk sin ~ T ~1g(n)sin(k-n) ~pT + Clcos(kwpT)+ C2sin(kwpT) (17) where the last two terms represent the transient part Or the solution and may be dropped by starting the dir~erence equation with zero inltial conditions. Thus when the rilter 46 15 excited at its center rrequency, the envelope amplitude is directly proportional to the input amplitude and the number Or input samples processed and the phase Or the output is the same as the input phase. In the synchronization process, 127 samples Or the input are processed while the 128th sample is set to zero to allow the dirferenceequation to rlng once.
3o Baud synchronization ls accomplished by adding a correctlon + A to the number o~ samples processed every other 11398~4 ¦baud (Figure 10). The difrerence equation rilter is adapted to ¦detect rour alignments Or the actual received baud with respect ¦to the receiver timing These conditions are diagral~med in ¦Figure 11. In the instances (a) and (b) Or Figure 11J energy ¦is measured in only one Or the two intervals, the synchroniza-¦tion correction - 1 is calculated rrom the following expresslon:
l ~= C(m~ - me) (1~) ¦where ml ls the magnitude of the late filter, me is the magnitude ¦o~ the early filter, and C is a synchronization constant. The ¦constant C is chosen so that the maximum value o~ ~ during the ¦lock-in phase synchronization is three of rour samples. Thus ~hen the received baud is in sync, neither rilter will measure ¦energy and the synchronization correction goes to zero. This ¦condition i9 illustrated as portion (c) Or Figure 11. Since the number o~ samples monitored by the early and late rilters exceeds the number of samples in a baud, it is possible that a baud transitlon will lie in both the early and late rilters, as I shown in Figure ll(d). This condition occurs only durlng ¦initial synchronization lock-up and could prevent the synchronl-zation acquisition. To prevent suoh an occurrencel the values the early and late magnltudes are checked be~ore being utlllzed l in equation 18. If both magnitudes exceed an experimenta}ly ¦ determined threshold, the value ~ is forced to +1. In this manner the program will tend to walk out of the indeterminate posltlon. When the input to the receive mode program is set ¦ to lts normal level, synchronization acquisition is achieved ¦ in a very short time span. When the input level ls reduced to zero, as might be experienced in a deep signal rade, the 3o correction goes to zero. Thus synchronization will be maintained ~' ~
¦in a deep fade as lon~ as the accuracy Or the transm1tter and ¦receiver clocks are surricient ¦ Considering again the FFT computer 26, lt now operates ¦to implement a 64 point complex FFT while accepting samples at ¦the real only input and thus is capable Or performing a 128 polnt sample transrorm which is used twice, rirst to analyze the incomlng baud for demodulation, and second to generate the impulse response from the decoded cepstrum. The rollowing dls-cussion illustrates the relationships necessary to obtain the discrete Fourier transrorm Or 128 sample (2N) real only input ~rom the N= 64 complex sample FFT algorithm. The input is packed into the complex input vector Or the FFT and the output of the FFT i9 even/odd separated and weighted to rorm the N
sample, non-symmetric DFT of the real value input.
The DFT of a 2N sample real inputJ x(tn), n = 0, 1, ...
2N-l is expressed in the rollowing equation:
1 2N-1 -i ~N n (19) Slnce the input i9 real, the real part Or X(fk) is even symmetric about N/2 and the imaginary part of x(rk) is odd symmetric about N/2. Because of these symmetries, the output x(rk) is des-cribed ~or k~ N-l.
The summation o~ equatlon 19 may be separated lnto two summations, the rirst over the even n samples Or the input and the second over the odd n samples Or the input according to the ~ollowlng expression:
N-~ k2n (20) 1 1~ ( N+l) ~' ` il3~W~ `
Substituting the Identlty:
e N = e e N (21) Equation 20 may be rewritten as:
k 2 N [~ 0 e x(t2 ~¦+ 2 e n ~ i2~n The rirst bracketed term ls the DFT Or the even sample part ¦of the lnput and the second bracketed term ls the DF~r of the odd ¦sample par~ Or the lnput. Thus equation 22 can be rewritten as ¦ X(fk) = 2 F ¦X(t2n)] ~ 2 e N F ¦X(t2n~1)] (23) Equation 23 thus provides the basis for the real input transform algorithm.
Accordingly, using this algorithm~ e~ricient use can be made o~ the complex FFT when the lnput is a single real valued function. At the cost of even/odd separating the output and applylng a weighting runction, an N complex sample FFT can be used ror a 2N sarnple real input FFTj resulting in an approximate 50~ reduction in running time and working storage.
Referring now back to Figure 6, 128 samples o~ the modem waveform are trans~erred rrom the burrer 44 to the real input Rkf ) where the samples are placed in the working storage, not shown, Or the FFT sectlon 78, ~lhereupon the transrorm pro-duces 64 non-symmetric complex output vectors Or which 25 are us~d to provide modem tone analysis Since the input ls real, the output vector will consist Or real and ima~inary parts ~ n and ~ n which is fed into a rour phase DPSK decoder 110 which is operable to provide packed cepstral coefficients which are red to a cepstrum decoder 112 The decoder is operable to 3o form cepstral coef~icients by applying inverse scaling and r ~ ~ 1139Wll ¦quantizing operations such as described with respect to the ¦quantizer 54 shown ln Figure 5. The received cepstrum is de-¦coded into even function coerricients Ck and red to the summer 164 along with odd runction coefficients Mk representative Or ¦the spectral envelope, which results from the real part ~ n ¦or a previous path Or data through the FFT computer 26 while ¦ perrorming a 128 point real DFT. The spectral envelope samples Mk 19 shown in Figure 12 and results ~rom the operation Or the spectrum processor sectlon 68 on the logged spectral envelope Mn~
l e the coslne transrorm o~ the.received cepstrum Ck. The spectrum processor section 68 includes means 114 and 116 ror perrorming the inverse process o~ the logging algoritllm and ror multiplying the spectral envelope by a series of random positive and negative unit impulses when the spectral envelope represents an unvoiced section ol speech.
The imaginary part Or the second 128 poin~ real DFT
.which comprises the part~ n comprises the impulse response Hn which ls the sine trans~orm of the spectral magnitude Mk and i8 illustrated, rOr example in Figure 13. Since the impulse response is derived rrom the Fourier trans~orm o~ the spectral magnitude, the impulse response samples are aharaoterlzed by the dlagram o~ Figure 13 For voiced synthesis, the impulse response data samples have applied thereto Hanning weighting in a circult 118, while ror unvoiced synthesis the impulse response is ~ed unwelghted into the synthesizer section 120. An impulse quietlng response circuit is prererably lnserted between the Hannlng weighting circuit 118 and the synthesizer 122 to remove any low level noise which may arise ~rom time to time, particularly when an unvoiced speech lnput i3 present. In addition to the .
impulse response, the synthesizer 120, a typical example Or which ~ ~ ' ~13g88~ ' ¦is disclosed in U S. Patent 3,681,530, also recelves a pitch ¦error correction ~ignal from the block 122 which is coupled to ¦the four phase DPSK decoder 110 which operate~ to ~inimlze the ¦effect of error rate on the re~enerated pitch parameter ror ¦value~ Or pitch which lie outside the range of the pltch detector and are thus trapped berore the pitch is transmitted to the synthesizer. The circuit 122 operates in such a rashlon that ¦when activated the previous value Or pitch received is trans-l mitted to the synthesizer 120. The synthesizer 120 operates ln ¦ a well known manner to provide digitally coded signals which when red to the digital to analog converter 58 provides syn-thesized output speech ~rom the modem input originally.
Thus what has been shown and described is an all digital processor which combines the functions o~ a vocoder and a modem in a half duplex mode Or operation while making maximum practlcal use of a single FFT algorithm which ls used for both vocoder and modem processing in both the transmit and receive modes o~
operation.
While there has been shown and described what is at present considered to be the preferred embodiment of the subject invention, modi~ications and alterations wlll readily occur to tho8e ~killed in the art, It is not de~irable, thererore, that the lnvention be limited to the speci~ic arran~ements shown and described, but it is intended to cover all such modlricatlons as ~all within the spirit and scope o~ the invention as derined ln e appended o1aim~.
~ `:
i~39~8~
~ . . .
from the center Or the baud so that the input data is lsolated as much as posslble rrom the baud transition. The data ln this 128 sample orthogonal window is trans orlned once per baud to provide the Fourier analysis Or the input tones which ls necessary ~or demodulation.
The interval designated D~ input salnplc~ ln l~ ure 10 illustrates the gating Or input sarnples to the dirrerence equa-tion rilter 106 In each 2 baud interval, the last 128 samples o~ the first baud and the rirst 128 samples of the second baud are fed into an integrate and dump dirference equation imple-mented thereby. The algorithm used to process the synchroniza-tion input data is represented by the rollowing dif~erence equation:
Yk+1 ~ 2(cos ~pT) Yk+Yk-1 gk (16) where ~p is the radian center rrequency of the rilter, T ls the sampling period, and gk is the input. This dirference equation has a solution: -Yk sin ~ T ~1g(n)sin(k-n) ~pT + Clcos(kwpT)+ C2sin(kwpT) (17) where the last two terms represent the transient part Or the solution and may be dropped by starting the dir~erence equation with zero inltial conditions. Thus when the rilter 46 15 excited at its center rrequency, the envelope amplitude is directly proportional to the input amplitude and the number Or input samples processed and the phase Or the output is the same as the input phase. In the synchronization process, 127 samples Or the input are processed while the 128th sample is set to zero to allow the dirferenceequation to rlng once.
3o Baud synchronization ls accomplished by adding a correctlon + A to the number o~ samples processed every other 11398~4 ¦baud (Figure 10). The difrerence equation rilter is adapted to ¦detect rour alignments Or the actual received baud with respect ¦to the receiver timing These conditions are diagral~med in ¦Figure 11. In the instances (a) and (b) Or Figure 11J energy ¦is measured in only one Or the two intervals, the synchroniza-¦tion correction - 1 is calculated rrom the following expresslon:
l ~= C(m~ - me) (1~) ¦where ml ls the magnitude of the late filter, me is the magnitude ¦o~ the early filter, and C is a synchronization constant. The ¦constant C is chosen so that the maximum value o~ ~ during the ¦lock-in phase synchronization is three of rour samples. Thus ~hen the received baud is in sync, neither rilter will measure ¦energy and the synchronization correction goes to zero. This ¦condition i9 illustrated as portion (c) Or Figure 11. Since the number o~ samples monitored by the early and late rilters exceeds the number of samples in a baud, it is possible that a baud transitlon will lie in both the early and late rilters, as I shown in Figure ll(d). This condition occurs only durlng ¦initial synchronization lock-up and could prevent the synchronl-zation acquisition. To prevent suoh an occurrencel the values the early and late magnltudes are checked be~ore being utlllzed l in equation 18. If both magnitudes exceed an experimenta}ly ¦ determined threshold, the value ~ is forced to +1. In this manner the program will tend to walk out of the indeterminate posltlon. When the input to the receive mode program is set ¦ to lts normal level, synchronization acquisition is achieved ¦ in a very short time span. When the input level ls reduced to zero, as might be experienced in a deep signal rade, the 3o correction goes to zero. Thus synchronization will be maintained ~' ~
¦in a deep fade as lon~ as the accuracy Or the transm1tter and ¦receiver clocks are surricient ¦ Considering again the FFT computer 26, lt now operates ¦to implement a 64 point complex FFT while accepting samples at ¦the real only input and thus is capable Or performing a 128 polnt sample transrorm which is used twice, rirst to analyze the incomlng baud for demodulation, and second to generate the impulse response from the decoded cepstrum. The rollowing dls-cussion illustrates the relationships necessary to obtain the discrete Fourier transrorm Or 128 sample (2N) real only input ~rom the N= 64 complex sample FFT algorithm. The input is packed into the complex input vector Or the FFT and the output of the FFT i9 even/odd separated and weighted to rorm the N
sample, non-symmetric DFT of the real value input.
The DFT of a 2N sample real inputJ x(tn), n = 0, 1, ...
2N-l is expressed in the rollowing equation:
1 2N-1 -i ~N n (19) Slnce the input i9 real, the real part Or X(fk) is even symmetric about N/2 and the imaginary part of x(rk) is odd symmetric about N/2. Because of these symmetries, the output x(rk) is des-cribed ~or k~ N-l.
The summation o~ equatlon 19 may be separated lnto two summations, the rirst over the even n samples Or the input and the second over the odd n samples Or the input according to the ~ollowlng expression:
N-~ k2n (20) 1 1~ ( N+l) ~' ` il3~W~ `
Substituting the Identlty:
e N = e e N (21) Equation 20 may be rewritten as:
k 2 N [~ 0 e x(t2 ~¦+ 2 e n ~ i2~n The rirst bracketed term ls the DFT Or the even sample part ¦of the lnput and the second bracketed term ls the DF~r of the odd ¦sample par~ Or the lnput. Thus equation 22 can be rewritten as ¦ X(fk) = 2 F ¦X(t2n)] ~ 2 e N F ¦X(t2n~1)] (23) Equation 23 thus provides the basis for the real input transform algorithm.
Accordingly, using this algorithm~ e~ricient use can be made o~ the complex FFT when the lnput is a single real valued function. At the cost of even/odd separating the output and applylng a weighting runction, an N complex sample FFT can be used ror a 2N sarnple real input FFTj resulting in an approximate 50~ reduction in running time and working storage.
Referring now back to Figure 6, 128 samples o~ the modem waveform are trans~erred rrom the burrer 44 to the real input Rkf ) where the samples are placed in the working storage, not shown, Or the FFT sectlon 78, ~lhereupon the transrorm pro-duces 64 non-symmetric complex output vectors Or which 25 are us~d to provide modem tone analysis Since the input ls real, the output vector will consist Or real and ima~inary parts ~ n and ~ n which is fed into a rour phase DPSK decoder 110 which is operable to provide packed cepstral coefficients which are red to a cepstrum decoder 112 The decoder is operable to 3o form cepstral coef~icients by applying inverse scaling and r ~ ~ 1139Wll ¦quantizing operations such as described with respect to the ¦quantizer 54 shown ln Figure 5. The received cepstrum is de-¦coded into even function coerricients Ck and red to the summer 164 along with odd runction coefficients Mk representative Or ¦the spectral envelope, which results from the real part ~ n ¦or a previous path Or data through the FFT computer 26 while ¦ perrorming a 128 point real DFT. The spectral envelope samples Mk 19 shown in Figure 12 and results ~rom the operation Or the spectrum processor sectlon 68 on the logged spectral envelope Mn~
l e the coslne transrorm o~ the.received cepstrum Ck. The spectrum processor section 68 includes means 114 and 116 ror perrorming the inverse process o~ the logging algoritllm and ror multiplying the spectral envelope by a series of random positive and negative unit impulses when the spectral envelope represents an unvoiced section ol speech.
The imaginary part Or the second 128 poin~ real DFT
.which comprises the part~ n comprises the impulse response Hn which ls the sine trans~orm of the spectral magnitude Mk and i8 illustrated, rOr example in Figure 13. Since the impulse response is derived rrom the Fourier trans~orm o~ the spectral magnitude, the impulse response samples are aharaoterlzed by the dlagram o~ Figure 13 For voiced synthesis, the impulse response data samples have applied thereto Hanning weighting in a circult 118, while ror unvoiced synthesis the impulse response is ~ed unwelghted into the synthesizer section 120. An impulse quietlng response circuit is prererably lnserted between the Hannlng weighting circuit 118 and the synthesizer 122 to remove any low level noise which may arise ~rom time to time, particularly when an unvoiced speech lnput i3 present. In addition to the .
impulse response, the synthesizer 120, a typical example Or which ~ ~ ' ~13g88~ ' ¦is disclosed in U S. Patent 3,681,530, also recelves a pitch ¦error correction ~ignal from the block 122 which is coupled to ¦the four phase DPSK decoder 110 which operate~ to ~inimlze the ¦effect of error rate on the re~enerated pitch parameter ror ¦value~ Or pitch which lie outside the range of the pltch detector and are thus trapped berore the pitch is transmitted to the synthesizer. The circuit 122 operates in such a rashlon that ¦when activated the previous value Or pitch received is trans-l mitted to the synthesizer 120. The synthesizer 120 operates ln ¦ a well known manner to provide digitally coded signals which when red to the digital to analog converter 58 provides syn-thesized output speech ~rom the modem input originally.
Thus what has been shown and described is an all digital processor which combines the functions o~ a vocoder and a modem in a half duplex mode Or operation while making maximum practlcal use of a single FFT algorithm which ls used for both vocoder and modem processing in both the transmit and receive modes o~
operation.
While there has been shown and described what is at present considered to be the preferred embodiment of the subject invention, modi~ications and alterations wlll readily occur to tho8e ~killed in the art, It is not de~irable, thererore, that the lnvention be limited to the speci~ic arran~ements shown and described, but it is intended to cover all such modlricatlons as ~all within the spirit and scope o~ the invention as derined ln e appended o1aim~.
Claims (10)
PROPERTY OR PRIVILEGE IS CLAIMED ARE DEFINED AS FOLLOWS:
1. In signal processing apparatus including means implementing both vocoder and modem functions in a half duplex mode of operation in response to a voice signal in a transmit mode and a modem signal in a receive mode, the improvement comprising:
input means selectively coupled to the voice signal and the modem signal to provide respective digital data sequences therefrom;
common computer means coupled to said input means implementing a single fast Fourier transform algorithm adapted to perform a first multi-point complex fast Fourier transform in the transmit mode and to perform a second multi-point complex fast Fourier transform in the receive mode, said first transform being operable to effect spectrum and cepstrum analysis of the digital data sequence of the voice signal to provide pitch detection and cepstral encoding signals therefrom, and said second transform being operable to effect demodulation of the digital data sequence of the modem signal to provide pitch and impulse response signals therefrom;
first circuit means coupled to said common computer means in the transmit mode and being responsive to the pitch detection and cepstral encoding signals to generate a plurality of information bearing tone waveforms to provide a modem signal in digitized form;
first feedback circuit means connected to the output of said common computer means for feeding back a processed spectrum signal in the transmit mode to the input of said common computer means;
second circuit means coupled to said common computer means in the receive mode and being responsive to the pitch and impulse response signals to generate an output speech waveform in digitized form;
second feedback means connected to the output of said common computer means for feeding back a processed spectrum signal in the receive mode to the input of said common means; and output means selectively coupled to said first and second circuit means in synchronism with said input means to provide an analog signal of either modem signal in the transmit mode or a voice signal in the receive mode.
input means selectively coupled to the voice signal and the modem signal to provide respective digital data sequences therefrom;
common computer means coupled to said input means implementing a single fast Fourier transform algorithm adapted to perform a first multi-point complex fast Fourier transform in the transmit mode and to perform a second multi-point complex fast Fourier transform in the receive mode, said first transform being operable to effect spectrum and cepstrum analysis of the digital data sequence of the voice signal to provide pitch detection and cepstral encoding signals therefrom, and said second transform being operable to effect demodulation of the digital data sequence of the modem signal to provide pitch and impulse response signals therefrom;
first circuit means coupled to said common computer means in the transmit mode and being responsive to the pitch detection and cepstral encoding signals to generate a plurality of information bearing tone waveforms to provide a modem signal in digitized form;
first feedback circuit means connected to the output of said common computer means for feeding back a processed spectrum signal in the transmit mode to the input of said common computer means;
second circuit means coupled to said common computer means in the receive mode and being responsive to the pitch and impulse response signals to generate an output speech waveform in digitized form;
second feedback means connected to the output of said common computer means for feeding back a processed spectrum signal in the receive mode to the input of said common means; and output means selectively coupled to said first and second circuit means in synchronism with said input means to provide an analog signal of either modem signal in the transmit mode or a voice signal in the receive mode.
2. The signal processing apparatus as defined by claim 1 wherein said second multi-point complex fast Fourier transform providing pitch and impulse response signals comprises two routines of a real input discrete Fourier transform.
3. The signal processing apparatus as defined by claim 1 wherein said first multi-point complex fast Fourier transform comprises an N-point FFT and wherein said second multi-point complex fast Fourier transform comprises an N/4 FFT used to perform an N/2 point, real input discrete Fourier transform, the latter being used in a first routine for demodulating the digitized modem input signal and in a second routine to transform decoded cepstral parameters resulting from said demodulation into a log spectrum and to transform said spectrum into said impulse res-ponse signals.
4. The signal processing apparatus as defined by claim 1 wherein said first multi-point complex transform com-prises a 256-point fast Fourier transform and wherein said second multi-point complex transform comprises a 64-point fast Fourier transform used to effect a 128-point real input discrete Fourier transform.
5. The signal processing apparatus as defined by claim 1 wherein said input means includes analog to digital converter means being operable to generate an input data sequence from either said voice signal or said modem signal inputted thereto and digital data storage means coupled to said analog to digital converter means and being operable to couple the digital data sequence to said computer means in successive sets of input samples; and wherein said output means comprises digital to analog converter means for converting the digitized waveforms from said first and second circuit means to respective analog representa-tions thereof.
6. The signal processing apparatus as defined by claim 1 wherein said first circuit means comprises modulator means incorporating a table look-up differentially coherent phase shift keying modulation routine which is operable to generate said plurality of information bearing tones,
7. The signal processing apparatus as defined by claim 1 wherein said second circuit means includes synthesizer means responsive to said pitch and impulse response signals which is operable to generate said output speech waveform in digitized form.
8. The signal processing apparatus as defined by claim 1 and additionally including speech pre-processor means coupled between said input means and said computer means in said transmit mode, said speech pre-processor means including means for normalizing said digital data sequence and providing Hanning weighting to said digital data prior to being fed to said computer means,
9. A signal processing apparatus as defined by claim 1 and additionally including means coupled between said computer means and said second circuit means in said receive mode for providing Hanning weighting to said impulse response signals.
10. The signal processing apparatus as defined by claim 1 and additionally including synchronization means between said input means and said computer means in said receive mode for regulating the digital data sequence inputted to said computer means.
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US06/114,777 US4310721A (en) | 1980-01-23 | 1980-01-23 | Half duplex integral vocoder modem system |
| US114,777 | 1980-01-23 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA1139884A true CA1139884A (en) | 1983-01-18 |
Family
ID=22357363
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA000354955A Expired CA1139884A (en) | 1980-01-23 | 1980-06-19 | Half duplex integral vocoder modem system |
Country Status (2)
| Country | Link |
|---|---|
| US (1) | US4310721A (en) |
| CA (1) | CA1139884A (en) |
Families Citing this family (176)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US4454610A (en) * | 1978-05-19 | 1984-06-12 | Transaction Sciences Corporation | Methods and apparatus for the automatic classification of patterns |
| WO1983003917A1 (en) * | 1982-04-29 | 1983-11-10 | Massachusetts Institute Of Technology | Voice encoder and synthesizer |
| US4554658A (en) * | 1983-07-07 | 1985-11-19 | Motorola, Inc. | System for producing encoded and decoded voice, internal audio tones, and data communication within a time division multiplex network |
| US4620294A (en) * | 1983-09-09 | 1986-10-28 | Cts Corporation | Digital signal processor modem |
| US4680797A (en) * | 1984-06-26 | 1987-07-14 | The United States Of America As Represented By The Secretary Of The Air Force | Secure digital speech communication |
| US4790016A (en) * | 1985-11-14 | 1988-12-06 | Gte Laboratories Incorporated | Adaptive method and apparatus for coding speech |
| US5345473A (en) * | 1987-01-22 | 1994-09-06 | Outokumpu Oy | Apparatus for providing two-way communication in underground facilities |
| US4884247A (en) * | 1987-03-09 | 1989-11-28 | Mobil Oil Company | Method of processing geophysical data to compensate for earth filter attenuation |
| US4780859A (en) * | 1987-03-09 | 1988-10-25 | Mobil Oil Corporation | Method of interpreting seismic data |
| US4809334A (en) * | 1987-07-09 | 1989-02-28 | Communications Satellite Corporation | Method for detection and correction of errors in speech pitch period estimates |
| GB2209648B (en) * | 1987-09-10 | 1991-10-23 | Ncr Co | Modem communication system having main and secondary channels |
| US5034962A (en) * | 1988-07-01 | 1991-07-23 | Oki Electric Industry Co., Ltd. | Voice-band signal processor |
| US4983906A (en) * | 1989-08-17 | 1991-01-08 | Hewlett-Packard Company | Frequency estimation system |
| US5297186A (en) * | 1991-07-29 | 1994-03-22 | Codex Corporation | Device and method for on-line adaptive selection of baud rate and carrier frequency |
| US5884253A (en) * | 1992-04-09 | 1999-03-16 | Lucent Technologies, Inc. | Prototype waveform speech coding with interpolation of pitch, pitch-period waveforms, and synthesis filter |
| FR2696297B1 (en) * | 1992-09-25 | 1994-11-04 | Alcatel Radiotelephone | Method for determining the transmittance of a filtering circuit provided for transforming the impulse response of a filter into a minimum phase response and filter using this method. |
| US5692104A (en) * | 1992-12-31 | 1997-11-25 | Apple Computer, Inc. | Method and apparatus for detecting end points of speech activity |
| US5596680A (en) * | 1992-12-31 | 1997-01-21 | Apple Computer, Inc. | Method and apparatus for detecting speech activity using cepstrum vectors |
| US7082106B2 (en) * | 1993-01-08 | 2006-07-25 | Multi-Tech Systems, Inc. | Computer-based multi-media communications system and method |
| DE69614799T2 (en) * | 1995-05-10 | 2002-06-13 | Koninkl Philips Electronics Nv | TRANSMISSION SYSTEM AND METHOD FOR VOICE ENCODING WITH IMPROVED BASIC FREQUENCY DETECTION |
| SE9702316L (en) | 1997-06-18 | 1998-12-19 | Telia Ab | VDSL modems |
| US6529866B1 (en) * | 1999-11-24 | 2003-03-04 | The United States Of America As Represented By The Secretary Of The Navy | Speech recognition system and associated methods |
| US8645137B2 (en) | 2000-03-16 | 2014-02-04 | Apple Inc. | Fast, language-independent method for user authentication by voice |
| KR100383668B1 (en) * | 2000-09-19 | 2003-05-14 | 한국전자통신연구원 | The Speech Coding System Using Time-Seperated Algorithm |
| ITFI20010199A1 (en) | 2001-10-22 | 2003-04-22 | Riccardo Vieri | SYSTEM AND METHOD TO TRANSFORM TEXTUAL COMMUNICATIONS INTO VOICE AND SEND THEM WITH AN INTERNET CONNECTION TO ANY TELEPHONE SYSTEM |
| FR2848715B1 (en) * | 2002-12-11 | 2005-02-18 | France Telecom | METHOD AND SYSTEM FOR MULTI-REFERENCE CORRECTION OF SPECTRAL VOICE DEFORMATIONS INTRODUCED BY A COMMUNICATION NETWORK |
| US7231190B2 (en) * | 2003-07-28 | 2007-06-12 | Motorola, Inc. | Method and apparatus for terminating reception in a wireless communication system |
| US8677377B2 (en) | 2005-09-08 | 2014-03-18 | Apple Inc. | Method and apparatus for building an intelligent automated assistant |
| US7633076B2 (en) | 2005-09-30 | 2009-12-15 | Apple Inc. | Automated response to and sensing of user activity in portable devices |
| JP2007114417A (en) * | 2005-10-19 | 2007-05-10 | Fujitsu Ltd | Voice data processing method and device |
| US9318108B2 (en) | 2010-01-18 | 2016-04-19 | Apple Inc. | Intelligent automated assistant |
| US20080129520A1 (en) * | 2006-12-01 | 2008-06-05 | Apple Computer, Inc. | Electronic device with enhanced audio feedback |
| US8977255B2 (en) | 2007-04-03 | 2015-03-10 | Apple Inc. | Method and system for operating a multi-function portable electronic device using voice-activation |
| US9053089B2 (en) * | 2007-10-02 | 2015-06-09 | Apple Inc. | Part-of-speech tagging using latent analogy |
| US8620662B2 (en) | 2007-11-20 | 2013-12-31 | Apple Inc. | Context-aware unit selection |
| US10002189B2 (en) * | 2007-12-20 | 2018-06-19 | Apple Inc. | Method and apparatus for searching using an active ontology |
| US9330720B2 (en) | 2008-01-03 | 2016-05-03 | Apple Inc. | Methods and apparatus for altering audio output signals |
| US8065143B2 (en) | 2008-02-22 | 2011-11-22 | Apple Inc. | Providing text input using speech data and non-speech data |
| US8996376B2 (en) | 2008-04-05 | 2015-03-31 | Apple Inc. | Intelligent text-to-speech conversion |
| US10496753B2 (en) | 2010-01-18 | 2019-12-03 | Apple Inc. | Automatically adapting user interfaces for hands-free interaction |
| US8464150B2 (en) | 2008-06-07 | 2013-06-11 | Apple Inc. | Automatic language identification for dynamic text processing |
| US20100030549A1 (en) | 2008-07-31 | 2010-02-04 | Lee Michael M | Mobile device having human language translation capability with positional feedback |
| US8768702B2 (en) * | 2008-09-05 | 2014-07-01 | Apple Inc. | Multi-tiered voice feedback in an electronic device |
| US8898568B2 (en) * | 2008-09-09 | 2014-11-25 | Apple Inc. | Audio user interface |
| US8583418B2 (en) | 2008-09-29 | 2013-11-12 | Apple Inc. | Systems and methods of detecting language and natural language strings for text to speech synthesis |
| US8712776B2 (en) * | 2008-09-29 | 2014-04-29 | Apple Inc. | Systems and methods for selective text to speech synthesis |
| US8676904B2 (en) | 2008-10-02 | 2014-03-18 | Apple Inc. | Electronic devices with voice command and contextual data processing capabilities |
| WO2010067118A1 (en) | 2008-12-11 | 2010-06-17 | Novauris Technologies Limited | Speech recognition involving a mobile device |
| US8862252B2 (en) | 2009-01-30 | 2014-10-14 | Apple Inc. | Audio user interface for displayless electronic device |
| US8380507B2 (en) | 2009-03-09 | 2013-02-19 | Apple Inc. | Systems and methods for determining the language to use for speech generated by a text to speech engine |
| US10241644B2 (en) | 2011-06-03 | 2019-03-26 | Apple Inc. | Actionable reminder entries |
| US20120311585A1 (en) | 2011-06-03 | 2012-12-06 | Apple Inc. | Organizing task items that represent tasks to perform |
| US10540976B2 (en) * | 2009-06-05 | 2020-01-21 | Apple Inc. | Contextual voice commands |
| US10241752B2 (en) | 2011-09-30 | 2019-03-26 | Apple Inc. | Interface for a virtual digital assistant |
| US9858925B2 (en) | 2009-06-05 | 2018-01-02 | Apple Inc. | Using context information to facilitate processing of commands in a virtual assistant |
| US9431006B2 (en) * | 2009-07-02 | 2016-08-30 | Apple Inc. | Methods and apparatuses for automatic speech recognition |
| US8682649B2 (en) * | 2009-11-12 | 2014-03-25 | Apple Inc. | Sentiment prediction from textual data |
| US8600743B2 (en) * | 2010-01-06 | 2013-12-03 | Apple Inc. | Noise profile determination for voice-related feature |
| US8381107B2 (en) | 2010-01-13 | 2013-02-19 | Apple Inc. | Adaptive audio feedback system and method |
| US8311838B2 (en) | 2010-01-13 | 2012-11-13 | Apple Inc. | Devices and methods for identifying a prompt corresponding to a voice input in a sequence of prompts |
| US10705794B2 (en) | 2010-01-18 | 2020-07-07 | Apple Inc. | Automatically adapting user interfaces for hands-free interaction |
| US10276170B2 (en) | 2010-01-18 | 2019-04-30 | Apple Inc. | Intelligent automated assistant |
| US10553209B2 (en) | 2010-01-18 | 2020-02-04 | Apple Inc. | Systems and methods for hands-free notification summaries |
| US10679605B2 (en) | 2010-01-18 | 2020-06-09 | Apple Inc. | Hands-free list-reading by intelligent automated assistant |
| WO2011089450A2 (en) | 2010-01-25 | 2011-07-28 | Andrew Peter Nelson Jerram | Apparatuses, methods and systems for a digital conversation management platform |
| US8682667B2 (en) | 2010-02-25 | 2014-03-25 | Apple Inc. | User profiling for selecting user specific voice input processing information |
| US8713021B2 (en) | 2010-07-07 | 2014-04-29 | Apple Inc. | Unsupervised document clustering using latent semantic density analysis |
| US8719006B2 (en) | 2010-08-27 | 2014-05-06 | Apple Inc. | Combined statistical and rule-based part-of-speech tagging for text-to-speech synthesis |
| US8719014B2 (en) | 2010-09-27 | 2014-05-06 | Apple Inc. | Electronic device with text error correction based on voice recognition data |
| US10762293B2 (en) | 2010-12-22 | 2020-09-01 | Apple Inc. | Using parts-of-speech tagging and named entity recognition for spelling correction |
| US10515147B2 (en) | 2010-12-22 | 2019-12-24 | Apple Inc. | Using statistical language models for contextual lookup |
| US8781836B2 (en) | 2011-02-22 | 2014-07-15 | Apple Inc. | Hearing assistance system for providing consistent human speech |
| US9262612B2 (en) | 2011-03-21 | 2016-02-16 | Apple Inc. | Device access using voice authentication |
| US20120310642A1 (en) | 2011-06-03 | 2012-12-06 | Apple Inc. | Automatically creating a mapping between text data and audio data |
| US10057736B2 (en) | 2011-06-03 | 2018-08-21 | Apple Inc. | Active transport based notifications |
| US8812294B2 (en) | 2011-06-21 | 2014-08-19 | Apple Inc. | Translating phrases from one language into another using an order-based set of declarative rules |
| US8706472B2 (en) | 2011-08-11 | 2014-04-22 | Apple Inc. | Method for disambiguating multiple readings in language conversion |
| US8994660B2 (en) | 2011-08-29 | 2015-03-31 | Apple Inc. | Text correction processing |
| US8762156B2 (en) | 2011-09-28 | 2014-06-24 | Apple Inc. | Speech recognition repair using contextual information |
| US10134385B2 (en) | 2012-03-02 | 2018-11-20 | Apple Inc. | Systems and methods for name pronunciation |
| US9483461B2 (en) | 2012-03-06 | 2016-11-01 | Apple Inc. | Handling speech synthesis of content for multiple languages |
| US9280610B2 (en) | 2012-05-14 | 2016-03-08 | Apple Inc. | Crowd sourcing information to fulfill user requests |
| US10417037B2 (en) | 2012-05-15 | 2019-09-17 | Apple Inc. | Systems and methods for integrating third party services with a digital assistant |
| US8775442B2 (en) | 2012-05-15 | 2014-07-08 | Apple Inc. | Semantic search using a single-source semantic model |
| WO2013185109A2 (en) | 2012-06-08 | 2013-12-12 | Apple Inc. | Systems and methods for recognizing textual identifiers within a plurality of words |
| US9721563B2 (en) | 2012-06-08 | 2017-08-01 | Apple Inc. | Name recognition system |
| US9495129B2 (en) | 2012-06-29 | 2016-11-15 | Apple Inc. | Device, method, and user interface for voice-activated navigation and browsing of a document |
| US9576574B2 (en) | 2012-09-10 | 2017-02-21 | Apple Inc. | Context-sensitive handling of interruptions by intelligent digital assistant |
| US9547647B2 (en) | 2012-09-19 | 2017-01-17 | Apple Inc. | Voice-based media searching |
| US8935167B2 (en) | 2012-09-25 | 2015-01-13 | Apple Inc. | Exemplar-based latent perceptual modeling for automatic speech recognition |
| KR102516577B1 (en) | 2013-02-07 | 2023-04-03 | 애플 인크. | Voice trigger for a digital assistant |
| US10572476B2 (en) | 2013-03-14 | 2020-02-25 | Apple Inc. | Refining a search based on schedule items |
| US9977779B2 (en) | 2013-03-14 | 2018-05-22 | Apple Inc. | Automatic supplementation of word correction dictionaries |
| US9368114B2 (en) | 2013-03-14 | 2016-06-14 | Apple Inc. | Context-sensitive handling of interruptions |
| US9733821B2 (en) | 2013-03-14 | 2017-08-15 | Apple Inc. | Voice control to diagnose inadvertent activation of accessibility features |
| US10652394B2 (en) | 2013-03-14 | 2020-05-12 | Apple Inc. | System and method for processing voicemail |
| US10642574B2 (en) | 2013-03-14 | 2020-05-05 | Apple Inc. | Device, method, and graphical user interface for outputting captions |
| US11151899B2 (en) | 2013-03-15 | 2021-10-19 | Apple Inc. | User training by intelligent digital assistant |
| WO2014144949A2 (en) | 2013-03-15 | 2014-09-18 | Apple Inc. | Training an at least partial voice command system |
| CN112230878A (en) | 2013-03-15 | 2021-01-15 | 苹果公司 | Context-sensitive handling of interrupts |
| US10748529B1 (en) | 2013-03-15 | 2020-08-18 | Apple Inc. | Voice activated device for use with a voice-based digital assistant |
| WO2014144579A1 (en) | 2013-03-15 | 2014-09-18 | Apple Inc. | System and method for updating an adaptive speech recognition model |
| WO2014197334A2 (en) | 2013-06-07 | 2014-12-11 | Apple Inc. | System and method for user-specified pronunciation of words for speech synthesis and recognition |
| WO2014197336A1 (en) | 2013-06-07 | 2014-12-11 | Apple Inc. | System and method for detecting errors in interactions with a voice-based digital assistant |
| US9582608B2 (en) | 2013-06-07 | 2017-02-28 | Apple Inc. | Unified ranking with entropy-weighted information for phrase-based semantic auto-completion |
| WO2014197335A1 (en) | 2013-06-08 | 2014-12-11 | Apple Inc. | Interpreting and acting upon commands that involve sharing information with remote devices |
| US10176167B2 (en) | 2013-06-09 | 2019-01-08 | Apple Inc. | System and method for inferring user intent from speech inputs |
| EP3008641A1 (en) | 2013-06-09 | 2016-04-20 | Apple Inc. | Device, method, and graphical user interface for enabling conversation persistence across two or more instances of a digital assistant |
| CN105265005B (en) | 2013-06-13 | 2019-09-17 | 苹果公司 | System and method for the urgent call initiated by voice command |
| WO2015020942A1 (en) | 2013-08-06 | 2015-02-12 | Apple Inc. | Auto-activating smart responses based on activities from remote devices |
| US10296160B2 (en) | 2013-12-06 | 2019-05-21 | Apple Inc. | Method for extracting salient dialog usage from live data |
| US9620105B2 (en) | 2014-05-15 | 2017-04-11 | Apple Inc. | Analyzing audio input for efficient speech and music recognition |
| US10592095B2 (en) | 2014-05-23 | 2020-03-17 | Apple Inc. | Instantaneous speaking of content on touch devices |
| US9502031B2 (en) | 2014-05-27 | 2016-11-22 | Apple Inc. | Method for supporting dynamic grammars in WFST-based ASR |
| US10078631B2 (en) | 2014-05-30 | 2018-09-18 | Apple Inc. | Entropy-guided text prediction using combined word and character n-gram language models |
| US9842101B2 (en) | 2014-05-30 | 2017-12-12 | Apple Inc. | Predictive conversion of language input |
| US10170123B2 (en) | 2014-05-30 | 2019-01-01 | Apple Inc. | Intelligent assistant for home automation |
| EP3149728B1 (en) | 2014-05-30 | 2019-01-16 | Apple Inc. | Multi-command single utterance input method |
| US10289433B2 (en) | 2014-05-30 | 2019-05-14 | Apple Inc. | Domain specific language for encoding assistant dialog |
| US9760559B2 (en) | 2014-05-30 | 2017-09-12 | Apple Inc. | Predictive text input |
| US9785630B2 (en) | 2014-05-30 | 2017-10-10 | Apple Inc. | Text prediction using combined word N-gram and unigram language models |
| US9715875B2 (en) | 2014-05-30 | 2017-07-25 | Apple Inc. | Reducing the need for manual start/end-pointing and trigger phrases |
| US9633004B2 (en) | 2014-05-30 | 2017-04-25 | Apple Inc. | Better resolution when referencing to concepts |
| US9430463B2 (en) | 2014-05-30 | 2016-08-30 | Apple Inc. | Exemplar-based natural language processing |
| US9734193B2 (en) | 2014-05-30 | 2017-08-15 | Apple Inc. | Determining domain salience ranking from ambiguous words in natural speech |
| US9338493B2 (en) | 2014-06-30 | 2016-05-10 | Apple Inc. | Intelligent automated assistant for TV user interactions |
| US10659851B2 (en) | 2014-06-30 | 2020-05-19 | Apple Inc. | Real-time digital assistant knowledge updates |
| US10446141B2 (en) | 2014-08-28 | 2019-10-15 | Apple Inc. | Automatic speech recognition based on user feedback |
| US9818400B2 (en) | 2014-09-11 | 2017-11-14 | Apple Inc. | Method and apparatus for discovering trending terms in speech requests |
| US10789041B2 (en) | 2014-09-12 | 2020-09-29 | Apple Inc. | Dynamic thresholds for always listening speech trigger |
| US9886432B2 (en) | 2014-09-30 | 2018-02-06 | Apple Inc. | Parsimonious handling of word inflection via categorical stem + suffix N-gram language models |
| US9646609B2 (en) | 2014-09-30 | 2017-05-09 | Apple Inc. | Caching apparatus for serving phonetic pronunciations |
| US9668121B2 (en) | 2014-09-30 | 2017-05-30 | Apple Inc. | Social reminders |
| US10074360B2 (en) | 2014-09-30 | 2018-09-11 | Apple Inc. | Providing an indication of the suitability of speech recognition |
| US10127911B2 (en) | 2014-09-30 | 2018-11-13 | Apple Inc. | Speaker identification and unsupervised speaker adaptation techniques |
| US10552013B2 (en) | 2014-12-02 | 2020-02-04 | Apple Inc. | Data detection |
| US9711141B2 (en) | 2014-12-09 | 2017-07-18 | Apple Inc. | Disambiguating heteronyms in speech synthesis |
| US9865280B2 (en) | 2015-03-06 | 2018-01-09 | Apple Inc. | Structured dictation using intelligent automated assistants |
| US9886953B2 (en) | 2015-03-08 | 2018-02-06 | Apple Inc. | Virtual assistant activation |
| US9721566B2 (en) | 2015-03-08 | 2017-08-01 | Apple Inc. | Competing devices responding to voice triggers |
| US10567477B2 (en) | 2015-03-08 | 2020-02-18 | Apple Inc. | Virtual assistant continuity |
| US9899019B2 (en) | 2015-03-18 | 2018-02-20 | Apple Inc. | Systems and methods for structured stem and suffix language models |
| US9842105B2 (en) | 2015-04-16 | 2017-12-12 | Apple Inc. | Parsimonious continuous-space phrase representations for natural language processing |
| US10083688B2 (en) | 2015-05-27 | 2018-09-25 | Apple Inc. | Device voice control for selecting a displayed affordance |
| US10127220B2 (en) | 2015-06-04 | 2018-11-13 | Apple Inc. | Language identification from short strings |
| US10101822B2 (en) | 2015-06-05 | 2018-10-16 | Apple Inc. | Language input correction |
| US10186254B2 (en) | 2015-06-07 | 2019-01-22 | Apple Inc. | Context-based endpoint detection |
| US10255907B2 (en) | 2015-06-07 | 2019-04-09 | Apple Inc. | Automatic accent detection using acoustic models |
| US11025565B2 (en) | 2015-06-07 | 2021-06-01 | Apple Inc. | Personalized prediction of responses for instant messaging |
| US10671428B2 (en) | 2015-09-08 | 2020-06-02 | Apple Inc. | Distributed personal assistant |
| US10747498B2 (en) | 2015-09-08 | 2020-08-18 | Apple Inc. | Zero latency digital assistant |
| US9697820B2 (en) | 2015-09-24 | 2017-07-04 | Apple Inc. | Unit-selection text-to-speech synthesis using concatenation-sensitive neural networks |
| US11010550B2 (en) | 2015-09-29 | 2021-05-18 | Apple Inc. | Unified language modeling framework for word prediction, auto-completion and auto-correction |
| US10366158B2 (en) | 2015-09-29 | 2019-07-30 | Apple Inc. | Efficient word encoding for recurrent neural network language models |
| US11587559B2 (en) | 2015-09-30 | 2023-02-21 | Apple Inc. | Intelligent device identification |
| US10691473B2 (en) | 2015-11-06 | 2020-06-23 | Apple Inc. | Intelligent automated assistant in a messaging environment |
| US10049668B2 (en) | 2015-12-02 | 2018-08-14 | Apple Inc. | Applying neural network language models to weighted finite state transducers for automatic speech recognition |
| US10223066B2 (en) | 2015-12-23 | 2019-03-05 | Apple Inc. | Proactive assistance based on dialog communication between devices |
| US10446143B2 (en) | 2016-03-14 | 2019-10-15 | Apple Inc. | Identification of voice inputs providing credentials |
| US9934775B2 (en) | 2016-05-26 | 2018-04-03 | Apple Inc. | Unit-selection text-to-speech synthesis based on predicted concatenation parameters |
| US9972304B2 (en) | 2016-06-03 | 2018-05-15 | Apple Inc. | Privacy preserving distributed evaluation framework for embedded personalized systems |
| US10249300B2 (en) | 2016-06-06 | 2019-04-02 | Apple Inc. | Intelligent list reading |
| US10049663B2 (en) | 2016-06-08 | 2018-08-14 | Apple, Inc. | Intelligent automated assistant for media exploration |
| DK179309B1 (en) | 2016-06-09 | 2018-04-23 | Apple Inc | Intelligent automated assistant in a home environment |
| US10192552B2 (en) | 2016-06-10 | 2019-01-29 | Apple Inc. | Digital assistant providing whispered speech |
| US10490187B2 (en) | 2016-06-10 | 2019-11-26 | Apple Inc. | Digital assistant providing automated status report |
| US10067938B2 (en) | 2016-06-10 | 2018-09-04 | Apple Inc. | Multilingual word prediction |
| US10509862B2 (en) | 2016-06-10 | 2019-12-17 | Apple Inc. | Dynamic phrase expansion of language input |
| US10586535B2 (en) | 2016-06-10 | 2020-03-10 | Apple Inc. | Intelligent digital assistant in a multi-tasking environment |
| DK179049B1 (en) | 2016-06-11 | 2017-09-18 | Apple Inc | Data driven natural language event detection and classification |
| DK179415B1 (en) | 2016-06-11 | 2018-06-14 | Apple Inc | Intelligent device arbitration and control |
| DK201670540A1 (en) | 2016-06-11 | 2018-01-08 | Apple Inc | Application integration with a digital assistant |
| DK179343B1 (en) | 2016-06-11 | 2018-05-14 | Apple Inc | Intelligent task discovery |
| US10593346B2 (en) | 2016-12-22 | 2020-03-17 | Apple Inc. | Rank-reduced token representation for automatic speech recognition |
| DK179745B1 (en) | 2017-05-12 | 2019-05-01 | Apple Inc. | SYNCHRONIZATION AND TASK DELEGATION OF A DIGITAL ASSISTANT |
| DK201770431A1 (en) | 2017-05-15 | 2018-12-20 | Apple Inc. | Optimizing dialogue policy decisions for digital assistants using implicit feedback |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US3634668A (en) * | 1970-06-01 | 1972-01-11 | Sylvania Electric Prod | Log spectrum decoding utilizing symmetry |
| US3662108A (en) * | 1970-06-08 | 1972-05-09 | Bell Telephone Labor Inc | Apparatus for reducing multipath distortion of signals utilizing cepstrum technique |
| US3681530A (en) * | 1970-06-15 | 1972-08-01 | Gte Sylvania Inc | Method and apparatus for signal bandwidth compression utilizing the fourier transform of the logarithm of the frequency spectrum magnitude |
| US3617941A (en) * | 1970-08-03 | 1971-11-02 | Sylvania Electric Prod | Table look-up modulator |
| US3706929A (en) * | 1971-01-04 | 1972-12-19 | Philco Ford Corp | Combined modem and vocoder pipeline processor |
| DE2422504B2 (en) * | 1974-05-09 | 1977-09-29 | TE KA DE Feiten & Guilleaume Fernmeldeanlagen GmbH, 8500 Nürnberg | MESSAGE TRANSMISSION SYSTEM |
| US4085449A (en) * | 1976-11-26 | 1978-04-18 | Paradyne Corporation | Digital modem |
-
1980
- 1980-01-23 US US06/114,777 patent/US4310721A/en not_active Expired - Lifetime
- 1980-06-19 CA CA000354955A patent/CA1139884A/en not_active Expired
Also Published As
| Publication number | Publication date |
|---|---|
| US4310721A (en) | 1982-01-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA1139884A (en) | Half duplex integral vocoder modem system | |
| EP0363640B1 (en) | High-frequency anti-jam communication system terminal | |
| Warner et al. | OFDM/FM frame synchronization for mobile radio data communication | |
| CA2158440C (en) | Method and apparatus for signal transmission and reception | |
| US7173966B2 (en) | Compensation for non-linear distortion in a modem receiver | |
| US4140876A (en) | Compressed speech system and predictor | |
| US4881059A (en) | Manchester code receiver | |
| KR20040041590A (en) | Compensation for non-linear distortion in a modem receiver | |
| JPH02309736A (en) | Modem with line inspecting function | |
| JPH07508393A (en) | Quantized coherent RAKE receiver | |
| EP0099397A1 (en) | Adaptive differential pcm coding. | |
| US3706929A (en) | Combined modem and vocoder pipeline processor | |
| US4126761A (en) | Method of and means for processing an audio frequency signal to conceal intelligility | |
| US4719649A (en) | Autoregressive peek-through comjammer and method | |
| Nowsheen et al. | Design of a high frequency FPGA acoustic modem for underwater communication | |
| EP0718821A2 (en) | Linear prediction filter coefficient quantizer and filter set | |
| Dubnowski et al. | Variable rate coding of speech | |
| US4034160A (en) | System for the transmission of speech signals | |
| EP1965501A2 (en) | Methods and apparatus for baseband digital spectrum translation (BDST) | |
| US5742679A (en) | Optimized simultaneous audio and data transmission using QADM with phase randomization | |
| US3843942A (en) | Equalizer for phase modulation communication systems using the instantaneous signal amplitude weighted by signal envelope amplitude distortion as an adjustment control signal | |
| Schmidt | The development of an underwater telephone for digital communication purposes | |
| KR100335696B1 (en) | Device for communication system and method of using communication system | |
| EP0167364A1 (en) | Speech-silence detection with subband coding | |
| Zhu et al. | Combined iterative cross-correlation demodulation scheme for mixing space borne automatic identification system signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| MKEX | Expiry |