CN104217725A - Audio watermarking method based on multi-echo core - Google Patents
Audio watermarking method based on multi-echo core Download PDFInfo
- Publication number
- CN104217725A CN104217725A CN201410507604.5A CN201410507604A CN104217725A CN 104217725 A CN104217725 A CN 104217725A CN 201410507604 A CN201410507604 A CN 201410507604A CN 104217725 A CN104217725 A CN 104217725A
- Authority
- CN
- China
- Prior art keywords
- watermark
- audio
- echo
- signal
- section
- 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.)
- Pending
Links
Abstract
The invention discloses an audio watermarking method based on a multi-echo core and belongs to the field of audio signal processing. The method includes first processing binary original watermarking signals, determining the echo core according to the attenuation coefficient and the echo delay point position, embedding the multiple ways of echo into the original audio signals, adopting an energy reversed spectral analysis method to decode the signals during watermarking extraction and extracting the corresponding watermarking information. The echo cache means is adopted in the embedding process, the synchronous performance is good, the problem of watermarking damage caused by the fact that the audio signals are shorn randomly is solved, the problem of echo interference in the multiple echo is solved, the data capacity is improved, the odd-even check information is added, the system hostile attack resistance is improved, and the method has high robustness.
Description
Technical field
The present invention relates to a kind of multimedia messages hidden method, particularly relate to a kind of audio-frequency water mark method based on many echoes core, can be used as audio-frequency information label mark, or for Information hiding, belong to Audio Signal Processing field and information security field.
Background technology
The copyright protection of audio multimedia and content retrieval are all concerned problems all the time; digital watermark technology is the approach that this problem provides solution; using label as watermark embedment audio multimedia; label can mark the source of audio frequency, copyright, synopsis or other attributes; simultaneously; the embedding of watermark label should be pursued and improve robustness and watermark capacity under the prerequisite not reducing audio quality, and can realize the embedding of watermark quick, convenient, steadily and extract detection.
For example, when employing FM audio broadcasting is as carrier, carry the satellite information (as copyright, source, Checking label etc.) relevant with broadcasted content, these information as watermark embedment in audio frequency, and the embedding of watermark can not destroy the audio quality of broadcast itself, together transmit with audio-frequency information during transmission.Audio watermarking technique is exactly under the prerequisite not affecting the quality of audio frequency own, will hide Info and be embedded in host signal, and extracts a kind of technology of watermark by corresponding decoding algorithm.Originally, as long as its function is limited in the aspects such as copyright protection, secret communication, tampering detection, now, along with the development of technology, algorithms of digital audio watermarking can be divided into following four classes: phase encoding, band spectrum modulation, joining method, echo hiding.Echo hiding technology, based on the masking effect of people's ear, artificial adds echo signal to audio frequency, under people's perception is less than the prerequisite of echo by information insertion in voice signal, in guarantee voice quality, there is natural advantage.In addition, the advantages such as it also has low complex degree, do not have noise to superpose, busyly detect, the good performance of net synchronization capability is good, so range of application is more extensive.At present, there are some related invention, they discuss by be applied to based on the audio frequency watermark of echo hiding be similar to FM broadcast media program in.
Such as, the title of Bender is the U.S. Patent No. 5,893 of " Method and Apparatus for Echo Data Hiding in Audio Signals ", and one or more echo is embedded in host audio signal by 067.The title of Yoiti Suzuki is that US Patent No. 2003/0172277 A1 of " Digital Watermark System " discloses and a kind ofly the echo signal of generation is inserted into data waterprint embedded method in original audio signal by expanding echo signal on a timeline.The title of Chen Ning is that " multi-watermarking based on echo hiding embeds and extraction algorithm " proposes and be embedded in same host signal by multi-watermarking signal.
But the echo hiding method capacity of Bender is very low and decode procedure is dangerous.The methods combining of Yoiti echo hiding and spread-spectrum, to improve capacity and security.But consider that the down-sampling in media interaction scenarios is attacked, the length of the embedding echo array allowed is limited, thus it can not provide length to be enough to ensure the PN sequence of good statistical attribute.On the other hand, this method is fragile for echo jitter attack.Chen Ning improves the embedded mobile GIS of echo hiding, but does not carry out rational theoretical validation to selected parameter, once the quantity of watermark is much, system performance can become very poor, and it does not consider the impact of FM modulation /demodulation on system yet.The behaviors such as the random editing especially often occurred in FM broadcast, require that system has good anti-lock-out attacking ability.
Therefore, prior art fails to provide a kind of effective ways by obtaining media program relevant information based on audio frequency watermark.
For the audio frequency watermark system being applied to FM broadcast, meet the following conditions:
(1) related content FM can broadcasted as watermark embedment in original broadcast, and can not affect the voice quality of broadcast itself.
(2) just can obtain by simple method, step the relevant information hidden in broadcast, to synchronous not requirement.
(3) compression, resampling, re-quantization, the attack maliciously such as wave filter, neighbourhood noise, synchronization attack can be resisted.
(4) after FM modulation, demodulation, watermark information is not destroyed, the information of our needs of decoding out normally.
(5) because FM broadcasted content enriches, require that this power system capacity is high.
Summary of the invention
The object of the invention is overcome the deficiency of above system and meet the above condition of system requirements, a kind of audio-frequency water mark method based on many echoes core is provided, the watermark label that can embed in quick, convenient, sane acquisition audio content, and do not affect the quality of broadcast own, the attack of multiple malice can be resisted.
A kind of audio-frequency water mark method based on many echoes core of the present invention, comprises the steps:
The telescopiny of watermark:
Step one: be scale-of-two watermark sequence { a by information coding to be embedded
0a
1a
2a
(J*U)-1, wherein a
n0 or 1,0≤n≤(J*U)-1; Described scale-of-two watermark sequence length is M=U*J bit, and wherein J, U are the positive integer being more than or equal to 1; J is the way of default echo, and U is the binary sequence bit number that every road echo embeds;
Such as, if No. ID of to be embedded is each audio frequency, distribute unique No. ID (one-to-one relationship) so first to each audio multimedia data, this No. ID is converted to scale-of-two watermark sequence (such as by mapping table or be directly converted to scale-of-two);
Step 2: scale-of-two watermark sequence step one obtained is embedded in corresponding sound signal, and concrete grammar is as follows:
1, first original audio signal X (n) is carried out segmentation, split time is
second, wherein F
sfor sampling rate, the i-th section audio signal writing X after segmentation
i(n);
2, M-bit scale-of-two watermark information { a step one obtained
0a
1a
2a
(J*U)-1change into the matrix that size is U × J:
Then will often arrange the binary number that writing one is new from top to bottom, and be translated into decimal number, obtain J decimal value { b
1b
2b
jand preserve;
3, the delay of J road echo signal is determined according to this J decimal value: d
1, d
2..., d
j;
d
1＝A+F×b
1
d
2＝d
1+F×b
2+K
.
.
.
d
J＝d
J-1+F×b
J+K
Wherein parameter A represents initial delay point value; K is intersegmental boundary belt length, namely minimum between different delay interval, and preferred value is 6; The F section of being intrinsic resolution, namely same road postpones the interval between different value, the optimal value of preferred value to be the preferred value of 3, K and F be experimentally result verification;
Then, to each section audio X
in (), by X
in () postpones d
1individual delay point, is then multiplied by attenuation coefficient α, namely obtains the 1st road echo signal α X
i(n-d
1); In like manner, X is obtained successively
iall the other echo signals α X of (n)
i(n-d
2) ..., α X
i(n-d
j), namely obtain J road echo signal; As preferably, α=0.4;
4, the parity information of 1 bit is determined according to the number of " 1 " in M-bit scale-of-two watermark sequence in step one; Scale-of-two watermark sequence is embedded into as follows each section audio X
iin (n), until whole audio frequency all embeds complete:
If parity information is ' 0 ', the number representing " 1 " in M-bit scale-of-two watermark sequence is odd number, then i-th section containing watermark signal:
Y
i(n)＝X
i(n)+αX
i(n-d
1)+αX
i(n-d
2)+…+αX
i(n-d
J).
That is now:
h(n)＝δ(n)+αδ(n-d
1)+…+αδ(n-d
J)
Y
i(n)＝X
i(n)*h(n)
If parity information is ' 0 ', the number representing " 1 " in M-bit scale-of-two watermark sequence is even number, then i-th section containing watermark signal:
Y
i(n)＝X
i(n)-αX
i(n-d
1)-αX
i(n-d
2)-…-αX
i(n-d
J)
That is now:
h(n)＝δ(n)-αδ(n-d
1)-…-αδ(n-d
J)；
Y
i(n)＝X
i(n)*h(n)
5, by every section audio signal Y of embed watermark
in () head and the tail couple together and obtain containing watermark audio signal Y (n);
The leaching process of watermark:
Step one: carry out segmentation to sound signal Y ' (n) to be detected, section length is
second, the i-th section audio signal after the section of scoring is Y '
i(n);
Step 2: the energy cepstrum calculating every section audio signal to be detected, the i-th section audio signal Y '
ithe energy cepstrum P of (n)
y ' i(n) be:
P
y′i(n)＝{IFFT(log(|FFT(Y
i′(n))|
2))}
2
Defined can be obtained by energy cepstrum
P
yi(n)＝{IFFT(log(|FFT(Y
i(n))|
2))}
2＝P
xi(n)+P
h(n)，
The end be arbitrary value can, choose as single 10 or e be the end; Thus obtain:
P
y(n)≈P
x(n)+n
2α
2δ(n)
+α
2{[δ(n-d
1)+δ(n+d
1)]+…+[δ(n-d
J)+δ(n+d
J)]}
+α
4{[δ(n-(d
2-d
1))+δ(n+(d
2-d
1))]+[δ(n-(d
3-d
1))+δ(n+(d
3-d
1))]+…+[δ(n-(d
J-d
1))+δ(n+(d
J-d
1))]}
+α
4{[δ(n-(d
3-d
2))+δ(n+(d
3-d
2))]+[δ(n-(d
4-d
2))+δ(n+(d
4-d
2))]+…+[δ(n-(d
J-d
2))+δ(n+(d
J-d
2))]}
+
.
.
.
+α
4{[δ(n-(d
J-1-d
J-2))+δ(n+(d
J-1-d
J-2))]+[δ(n-(d
J-d
J-2))+δ(n+(d
J-d
J-2))]+…+[δ(n-(d
J-d
2))+δ(n+(d
J-d
2))]}
+α
4[δ(n-(d
J-d
J-1))+δ(n+(d
J-d
J-1))]
Can find out according to formula, the sound signal containing watermark will at delay point d
1, d
2..., d
jthere is peak value in place;
Step 3: extract watermark information; Concrete grammar is as follows:
(1) to each segmentation sound signal Y ' to be detected
in (), detects its energy cepstrum P
y ' ieach peak point position of (n), and the peak point position detected is recorded as d
1', d
2' ..., d
j'
(2) { b is calculated according to the inverse operation of formula in the 3rd small step embedded in step 2
1' b
2' ... b
j', that is:
b
1′＝(d
1′-A)/F
b
2′＝(d
2′-d
1′-K)/F
.
.
.
b
J′＝(d
J′-d
J-1′-K)/F
(3) by { b
1' b
2' ... b
j' in each element round up be taken as integer after be converted to U position scale-of-two respectively, and the new binary number of the end to end composition of binary number one after these being changed, think that this binary number is extracted watermark;
Further, also comprise:
(4) by extracted watermark and the scale-of-two watermark sequence { a embedded in step
0a
1a
2α
(J*U)-1contrast, to judge that whether this sound signal is containing watermark.
Below be namely the audio-frequency water mark method based on many echoes core proposed by the invention.
Contrast prior art, beneficial effect of the present invention is:
1) the present invention is of many uses, such as, people can be made from media program, to obtain relevant information easily and efficiently when broadcast listening.
2) present invention employs many echoes core, and derived the upper limit of echo quantity in theory, improve the capacity of data, only need the broadcast of recording 1s, can 2 be identified
16individual FM broadcast program.
3) the present invention with the addition of parity information, enhances the ability of the anti-malicious attack of system, has very high robustness.
4) the present invention well solves in multiple echo, the problem that echo is disturbed mutually, and this postpones dot matrix to determine the delay point of several roads echo owing to adopting.
5) the present invention analyzes the performance of the band audio frequency watermark signal after FM modulation /demodulation, makes it to be more suitable for being applied to FM live broadcast.
6) echo telescopiny of the present invention takes echo buffer memory means, makes net synchronization capability good, solves the problem that FM broadcasts the destruction watermark that random shearing brings thus.
Accompanying drawing explanation
Fig. 1 is audio frequency watermark corresponding relation schematic diagram of the present invention;
Fig. 2 is the inventive method implementing procedure figure;
Fig. 3 is that embodiment adopts the inventive method to embed the process flow diagram of audio frequency watermark to sound signal;
Fig. 4 is that embodiment adopts the inventive method from sound signal, extract the process flow diagram of watermark;
Fig. 5 is the analogous diagram of the present invention's detected energy cepstrum peak point when detecting watermark;
Fig. 6 is multichannel echo delay signal schematic representation of the present invention.
Embodiment
Below in conjunction with drawings and Examples, the present invention is described in detail, also describe technical matters and the beneficial effect of technical solution of the present invention solution simultaneously, it is pointed out that described embodiment is only intended to be convenient to the understanding of the present invention, and any restriction effect is not play to it.
A kind of method to sound signal embed digital watermark of the present invention, see accompanying drawing 2, comprising:
Binary original watermark signal is processed, obtains the watermark signal after processing.According to the value of watermark information determination parity check bit, determine attenuation coefficient, determine the delay point position of echo.By the echo core that the delay point position of attenuation coefficient and echo determines, multichannel echo is embedded in original sound signal.Comprise when extracting watermark: adopt the method for energy cepstrum analysis to decode to signal, extract watermark information.
The present invention has also derived the upper limit of echo quantity in multiple echo core theoretically.
The following describes the theoretical foundation by the theoretical foundation of four tunnel inhibit signals and the value of parameter A.
1, by the theoretical foundation of multipath delay signal
Be provided with J road to postpone, that is
H (n)=δ (n)+α δ (n-d
1)+... + α δ (n-d
j), calculating energy cepstrum is:
P
y(n)≈P
x(n)+n
2α
2δ(n)
+α
2{[δ(n-d
1)+δ(n+d
1)]+…+[δ(n-d
J)+δ(n+d
J)]}
+α
4{[δ(n-(d
2-d
1))+δ(n+(d
2-d
1))]+[δ(n-(d
3-d
1))+δ(n+(d
3-d
1))]+…+[δ(n-(d
J-d
1))+δ(n+(d
J-d
1))]}
+α
4{[δ(n-(d
3-d
2))+δ(n+(d
3-d
2))]+[δ(n-(d
4-d
2))+δ(n+(d
4-d
2))]+…+[δ(n-(d
J-d
2))+δ(n+(d
J-d
2))]}
+
.
.
.
+α
4{[δ(n-(d
J-1-d
J-2))+δ(n+(d
J-1-d
J-2))]+[δ(n-(d
J-d
J-2))+δ(n+(d
J-d
J-2))]+…+[δ(n-(d
J-d
2))+δ(n+(d
J-d
2))]}
+α
4[δ(n-(d
J-d
J-1))+δ(n+(d
J-d
J-1))]
Analyze above formula can obtain: containing in the energy cepstrum of watermark signal, have two quasi-peak values.The first kind is the peak value produced in the position postponing point, and amplitude is α
2, we are referred to as to postpone some peak value, and we can utilize this peak value to determine to postpone the value of point, thus extract watermark information.Equations of The Second Kind peak value is at d
i-d
jthe peak value that (1 <=j < i <=J) place produces, amplitude is α
4, total (J-1) J/2 is individual.They can produce interference to delay point peak value, and then cause occurring decoding error, and we are referred to as to disturb peak value.Supposing a kind of limiting case, if all interference peak values all appear at same position, if amplitude now reaches with postponing some peak amplitude suitable, so just will inevitably occur decoding error, that is:
and α=0.4, we can calculate J=4.07.Also the theoretical upper limit of many echoes core is just said.Based on this theory deduction, we get J=4.
2, the theoretical foundation of the value of parameter A
Have above-mentioned analysis to obtain, during Dang You tetra-road peak value, the position of interference peak value is
(d
2-d
1)，(d
3-d
1)，…，(d
J-d
1)；
(d
3-d
2)，(d
4-d
2)，…，(d
J-d
2)；
…….
(d
J-1-d
J-2)，(d
J-d
J-2)
(d
J-d
J-1)
Known accordingly:
d
2-d
1＝3×(b
1+2)＝interp1
d
3-d
1＝3×(b
2+b
3+4)＝interp2
d
4-d
1＝3×(b
2+b
3+b
4+6)＝interp3
.
.
.
d
J-d
J-1＝3×(b
J-1+2)＝interp[(J-1)J/2]
Thus the total size calculating interference peak value is [min (interp1, interp2, interp3, interp4, interp [(J-1) J/2]), max (interp1, interp2, interp3, interp4 ..., interp [(J-1) J/2])].In order to make interference peak value can not affect postpone some a peak value, we get A=max (interp1, interp2, interp3, interp4 ..., interp [(J-1) J/2]) and+1 be optimal value.
Below in conjunction with embodiment, the specific embodiment of the present invention is made an explanation.Take watermark as the ID of audio data be example:
First multimedia database is set up, as Fig. 1.
Specific practice: compile unique No. ID (one-to-one relationship) to each multimedia document
Watermark embed step (accompanying drawing 3):
Step one: be embedded into No. ID in sound signal, if No. ID is the binary message of 16bits;
Specific practice is as follows:
1, first original complete audio frequency is carried out segmentation, split time for (44100/fs) s wherein fs=441.00 be sampling rate.If originally audio frequency is X (n), then after segmentation i-th section be X
i(n).
2,16bits binary message is divided into 4 parts, every part is 4bits information, then every part of 4bits binary number is converted into corresponding decimal number, obtains four decimal value [b
1b
2b
3b
4], wherein b
i∈ [0,15]
3, the delay d of four road echo signals is determined according to four decimal values
1, d
2, d
3, d
4.(F＝3，K＝6)
D
1=A+F × b
1, by X
in () postpones d
1individual delay point, is then multiplied by attenuation coefficient α=0.4, obtains first via echo signal α X
i(n-d
1). in like manner, we calculate other delay: d
2=d
1+ F × b
2+ K; d
3=d
2+ F × b
3+ K; d
4=d
3+ F × b
4+ K; Thus obtain remaining 3 road echo signal α X
i(n-d
2), α X
i(n-d
3), α X
i(n-d
4).
4, the parity information of 1bits is determined according to 16bits binary message, if be ' 0 ', the band watermark signal after embedding:
Y
i(n)＝X
i(n)+αX
i(n-d
1)+αX
i(n-d
2)+αX
i(n-d
3)+αX
i(n-d
4).
That is:
H (n)=δ (n)+α δ (n-d
1)+... + α δ (n-d
4), Y
i(n)=X
i(n) * h (n); Otherwise:
Y
i(n)＝X
i(n)-αX
i(n-d
1)-αX
i(n-d
2)-αX
i(n-d
3)-αX
i(n-d
4)。
That is h (n)=δ (n)-α δ (n-d
1)-...-α δ (n-d
4), Y
i(n)=X
i(n) * h (n); See accompanying drawing 6;
5, by id information loop embedding to next section audio X
i+1n (), until whole audio frequency all embeds complete.Every section audio signal of embed watermark is coupled together.
Introduce the extraction step (accompanying drawing 4) of watermark below: from audio signal decoding out watermark information (No. ID)
Calculate the energy cepstrum of band watermark signal
1, carry out segmentation to sound signal Y ' (n) to be detected, split time is (44100/fs) s;
2, defined can be obtained by energy cepstrum:
P
y(n)＝{IFFT(log(|FFT(y(n))|
2))}
2＝P
x(n)+P
h(n)
, thus obtain:
P
y(n)≈P
x(n)+n
2α
2δ(n)+{α
2[δ(n-d
1)+δ(n+d
1)]+…+α
2[δ(n-
d
4)+δ(n+d
4)]}+α
4{[δ(n-(d
2-d
1))+δ(n+(d
2-d
1))]+…+[δ(n-
(d
J-d
1))+δ(n+(d
J-d
1))]}+…+α
4[δ(n-(d
4-d
3))+δ(n+(d
4-
d
3))]
3, according to above-mentioned formula, we can find out, and band watermark signal is at delay point d
1, d
2, d
3, d
4there is peak value in place, thus the position of detection peak point can obtain d
1, d
2, d
3, d
4value (accompanying drawing 5).
Calculate by the value postponing some the watermark information (No. ID) embedded, according to No. ID, in data bank, search corresponding multimedia document, multimedia document is fed back to user
Illustrate below in conjunction with embodiment: the theoretical foundation of the value of parameter A:
Have above-mentioned analysis to obtain, during Dang You tetra-road peak value, the position of interference peak value is (d
2-d
1), (d
3-d
1), (d
4-d
1), (d
3-d
2), (d
4-d
2), (d
4-d
3):
d
2-d
1＝3×(b
1+2)∈[6，51]
d
3-d
1＝3×(b
2+b
3+4)∈[12，102]
d
4-d
1＝3×(b
2+b
3+b
4+6)∈[18，153]
d
3-d
2＝3×(b
3+2)∈[6，51]
d
4-d
2＝3×(b
3+b
4+4)∈[12，102]
d
4-d
3＝3×(b
4+2)∈[6，51]
The total size that therefrom we can calculate interference peak value is [6,153].Can not affect to make interference peak value and postpone some peak value, we get A=154 is optimal value.
Further, based on this method, the present invention can be used for obtaining FM broadcast related information, comprising:
Embed step: in the sound signal that this step broadcast to FM, embedding comprises the sequence number watermark information of the relevant information of media program;
FM step of transmitting: the broadcast singal comprising watermark information is launched by this step after FM modulation;
Transmission step: this step F M broadcast singal transmits in the channel;
FM demodulation step: in this step, FM receiver receives FM broadcast singal and signal receiving out;
Recording step: the part in the broadcasted content of this step user record FM demodulator solution band watermark out;
Extraction step: this step extracts corresponding watermark information from the part sound signal of record, namely obtains the sequence number of this section of broadcast associated content;
Relevant information forwarding step: this step according to the sequence number of the related content extracted, by the content of sequence number sending to user.
During transmission of audio watermark signal, comprise FM transmitter device, this equipment has been used and has been launched after FM modulation by the voice signal that embedded in audio frequency watermark.And FM demodulated equipment, this equipment is used for receiving FM broadcast singal, and by signal receiving out, shows etc.
Below by the performance of experimental evaluation this method.First by changing the length of segmentation, changing transfer rate, thus contrasting the watermarking algorithm based on many echoes core of the present invention and existing monokaryon watermarking algorithm performance when different transmission rates.Some are attacked, such as wave filter, resampling, re-quantization, add white noise etc. and be used for the robustness of testing algorithm; Subjective test results shows, and the audio frequency watermark system based on many echoes core has well not sentience.Finally, we whole method is applied to FM broadcast in and test.Here is the experimental result of embodiment:
This experiment tests as carrier audio frequency with 20 dissimilar audio clips, and these 20 audio clips adopt the sampling of 44.1Khz, and 16-bit quantizes, and can be specifically a symbol with reference to table 1 watermark content, wherein comprise the binary message of 16bits.
The carrier audio frequency of table 1. for testing
The accuracy of symbol is defined as:
One of them symbol contains the binary message of 16bits.
1, many echoes core audio frequency watermark system performance
(1) transfer rate
Adopting conventional single-core water mark method as a comparison in the present embodiment, by changing the length of segmentation, thus testing the performance of two kinds of methods at different rates.By result, we can find out, the water mark method based on many echoes core that we propose has obvious advantage, and especially when we improve transfer rate time, this advantage becomes more obvious.
Table 2.Result of transmission rate test
(2) robustness test
For FM broadcast system, evitable scarcely exist the behaviors such as resampling, quantification, compression, filtering, editing, so our system must have resistivity to the attack of these malice.In this experiment, what we thought adds attack, tests under difference is attacked, the performance of this system.Concrete attack can with reference to table 3 in this experiment, and section length is 1 second.As can be seen from the result of table 4 we, in contrast to traditional audio frequency watermark system, our system has higher accuracy, except resampling, the accuracy AR under other attack
sall exceed 90%.
Table 3.Detail of attack
Table 4.Result of attack test
(3) subjective testing
Take AB method for subjective testing to survey the quality of the audio frequency containing watermark.Wherein have 10 people to participate in test, everyone listens to 20 audio clips.In test process, they need to elect one from two options (raw tone and band watermark voice) and think that audio quality is better themselves.By this experiment, we can test the degree of injury increased to audio quality after watermark.Concrete outcome can with reference to table 5.
As can be seen from almost reciprocity result, we, add watermarking process and produce damage to audio frequency, is namely with watermarked audio to have good not sentience.
Table 5.Result of subjective test
| prefer | Host?audio | Watermarked?audio |
| number | 97 | 103 |
| rate | 0.485 | 0.515 |
2, performance test
In this experiment, the audio frequency watermark system of design is applied in FM broadcast by we, and split time is 1 second, and watermark is 1 symbol (16bits scale-of-two), adopt the mode record FM receiver decoding sound signal out of recording, test result can with reference to table 6.We know, along with the increase of recording distance, signal to noise ratio (S/N ratio) (SNR) can decline, accuracy AR
snature can decline, and result shows, and our system has showed performance well, and recording distance can reach 5 meters.
Table6.Test?result?of?the?whole?system
The above; be only the specific embodiment of the present invention; but protection scope of the present invention is not limited thereto; any people being familiar with this technology is in the technical scope disclosed by the present invention; the conversion and replacement expected can be understood; all should be encompassed in and of the present inventionly comprise within scope, therefore, protection scope of the present invention should be as the criterion with the protection domain of claims.
Claims (8)
1. based on an audio-frequency water mark method for many echoes core, it is characterized in that, comprise the steps:
The telescopiny of watermark:
Step one: be scale-of-two watermark sequence { a by information coding to be embedded
0a
1a
2a
(J*U)-1, wherein a
n0 or 1,0≤n≤(J*U)-1; Described scale-of-two watermark sequence length is M=U*J bit, and wherein J, U are the positive integer being more than or equal to 1; J is the way of default echo, and U is the binary sequence bit number that every road echo embeds;
Step 2: scale-of-two watermark sequence step one obtained is embedded in corresponding sound signal, and concrete grammar is as follows:
(1) first original audio signal X (n) is carried out segmentation, split time is
second, wherein F
sfor sampling rate, the i-th section audio signal writing X after segmentation
i(n);
(2) M-bit scale-of-two watermark information { a step one obtained
0a
1a
2a
(J*U)-1change into the matrix that size is U × J:
Then will often arrange the binary number that writing one is new from top to bottom, and be translated into decimal number, obtain J decimal value { b
1b
2b
jand preserve;
(3) delay of J road echo signal is determined according to this J decimal value: d
1, d
2..., d
j;
d
1＝A+F×b
1
d
2＝d
1+F×b
2+K
·
·
·
d
J＝d
J-1+F×b
J+K
Wherein parameter A represents initial delay point value; K is intersegmental boundary belt length, namely minimum between different delay interval, the F section of being intrinsic resolution, and namely same road postpones the interval between different value;
Then, to each section audio X
in (), by X
in () postpones d
1individual delay point, is then multiplied by attenuation coefficient α, namely obtains the 1st road echo signal α X
i(n-d
1); In like manner, X is obtained successively
iall the other echo signals α X of (n)
i(n-d
2) ..., α X
i(n-d
j), namely obtain J road echo signal;
(4) parity information of 1 bit is determined according to the number of " 1 " in M-bit scale-of-two watermark sequence in step one; Scale-of-two watermark sequence is embedded into as follows each section audio X
iin (n), until whole audio frequency all embeds complete:
If parity information is ' 0 ', the number representing " 1 " in M-bit scale-of-two watermark sequence is odd number, then i-th section containing watermark signal:
Y
i(n)＝X
i(n)+αX
i(n-d
1)+αX
i(n-d
2)+…+αX
i(n-d
J).
If parity information is ' 0 ', the number representing " 1 " in M-bit scale-of-two watermark sequence is even number, then i-th section containing watermark signal:
Y
i(n)＝X
i(n)-αX
i(n-d
1)-αX
i(n-d
2)-…-αX
i(n-d
J)
(5) by every section audio signal Y of embed watermark
in () head and the tail couple together and obtain containing watermark audio signal Y (n);
The leaching process of watermark:
Step one: carry out segmentation to sound signal Y ' (n) to be detected, section length is
second, the i-th section audio signal after the section of scoring is Y '
i(n);
Step 2: the energy cepstrum calculating every section audio signal to be detected, the i-th section audio signal Y '
ithe energy cepstrum P of (n)
y ' i(n) be:
P
y′i(n)＝{IFFT(log(|FFT(Y
i′(n))|
2))}
2
Step 3: extract watermark information; Method is as follows:
(1) to each segmentation sound signal Y ' to be detected
in (), detects its energy cepstrum P
y ' ieach peak point position of (n), and the peak point position detected is recorded as d
1', d
2' ..., d
j'
(2) { b is calculated
1' b
2' ... b
j', that is:
(3) by { b
1' b
2' ... b
j' in each element round up be taken as integer after be converted to U position scale-of-two respectively, and the new binary number of the end to end composition of binary number one after these being changed, think that this binary number is extracted watermark.
2. according to a kind of audio-frequency water mark method based on many echoes core described in claim 1, it is characterized in that, the step 3 of described extraction step, after extracting watermark, also comprises step: (4) are by extracted watermark and the scale-of-two watermark sequence { a embedded in step
0a
1a
2a
(J*U)-1contrast, to judge that whether this sound signal is containing watermark.
3., according to a kind of audio-frequency water mark method based on many echoes core described in claim 1, it is characterized in that, α=0.4.
4., according to a kind of audio-frequency water mark method based on many echoes core described in claim 3, it is characterized in that, M=4.
5., according to a kind of audio-frequency water mark method based on many echoes core described in claim 3, it is characterized in that, U=4.
6., according to a kind of audio-frequency water mark method based on many echoes core described in claim 1, it is characterized in that, K=6.
7., according to a kind of audio-frequency water mark method based on many echoes core described in claim 1, its feature exists. in, F=3.
8., according to a kind of audio-frequency water mark method based on many echoes core described in claim 4, it is characterized in that, A=154.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410507604.5A CN104217725A (en) | 2014-09-29 | 2014-09-29 | Audio watermarking method based on multi-echo core |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410507604.5A CN104217725A (en) | 2014-09-29 | 2014-09-29 | Audio watermarking method based on multi-echo core |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104217725A true CN104217725A (en) | 2014-12-17 |
Family
ID=52099133
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410507604.5A Pending CN104217725A (en) | 2014-09-29 | 2014-09-29 | Audio watermarking method based on multi-echo core |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104217725A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105374360A (en) * | 2015-11-25 | 2016-03-02 | 武汉大学 | Interleaved additive spread spectrum audio watermark embedding method and detection method and system |
| CN109522445A (en) * | 2018-11-15 | 2019-03-26 | 辽宁工程技术大学 | A kind of audio classification search method merging CNNs and phase algorithm |
| CN110266889A (en) * | 2019-06-26 | 2019-09-20 | 四川神琥科技有限公司 | A method of record and protection recorded audio are distorted by editing |
| WO2021155697A1 (en) * | 2020-02-04 | 2021-08-12 | 北京达佳互联信息技术有限公司 | Watermark information addition method and extraction method, and device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6674861B1 (en) * | 1998-12-29 | 2004-01-06 | Kent Ridge Digital Labs | Digital audio watermarking using content-adaptive, multiple echo hopping |
| CN101266794A (en) * | 2008-03-27 | 2008-09-17 | 上海交通大学 | Multiple watermark inlay and exaction method based on echo hiding |
| CN102148034A (en) * | 2011-04-07 | 2011-08-10 | 南京邮电大学 | Echo hiding based watermark embedding and extracting method |
| CN103413552A (en) * | 2013-08-29 | 2013-11-27 | 四川大学 | Audio watermark embedding and extracting method and device |
-
2014
- 2014-09-29 CN CN201410507604.5A patent/CN104217725A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6674861B1 (en) * | 1998-12-29 | 2004-01-06 | Kent Ridge Digital Labs | Digital audio watermarking using content-adaptive, multiple echo hopping |
| CN101266794A (en) * | 2008-03-27 | 2008-09-17 | 上海交通大学 | Multiple watermark inlay and exaction method based on echo hiding |
| CN102148034A (en) * | 2011-04-07 | 2011-08-10 | 南京邮电大学 | Echo hiding based watermark embedding and extracting method |
| CN103413552A (en) * | 2013-08-29 | 2013-11-27 | 四川大学 | Audio watermark embedding and extracting method and device |
Non-Patent Citations (2)
| Title |
|---|
| XUEJUN ZHANG ET AL.: "《Audio Watermarking Based on Multiple Echoes Hiding for FM Radio》", 《INTERSPEECH 2014》 * |
| XUEJUN ZHANG ET AL.: "《Data Hiding System Based on New On New Multiple Echo Hiding Method》", 《THE 21ST INTERNATIONAL CONGRESS ON SOUND AND VIBRATION》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105374360A (en) * | 2015-11-25 | 2016-03-02 | 武汉大学 | Interleaved additive spread spectrum audio watermark embedding method and detection method and system |
| CN105374360B (en) * | 2015-11-25 | 2018-12-14 | 武汉大学 | Intersect additivity spread spectrum audio frequency watermark embedding grammar, detection method and system |
| CN109522445A (en) * | 2018-11-15 | 2019-03-26 | 辽宁工程技术大学 | A kind of audio classification search method merging CNNs and phase algorithm |
| CN110266889A (en) * | 2019-06-26 | 2019-09-20 | 四川神琥科技有限公司 | A method of record and protection recorded audio are distorted by editing |
| CN110266889B (en) * | 2019-06-26 | 2021-03-19 | 四川神琥科技有限公司 | Method for recording and preventing recorded audio from being edited and tampered |
| WO2021155697A1 (en) * | 2020-02-04 | 2021-08-12 | 北京达佳互联信息技术有限公司 | Watermark information addition method and extraction method, and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102859585B (en) | Watermark signal provider and method for providing a watermark signal | |
| CN102959622B (en) | Watermark signal provision and watermark embedding | |
| CN1282152C (en) | Decoding of information in audio signals | |
| CN102884571B (en) | Watermark generator, watermark decoder, method for providing a watermark signal, method for providing binary message data in dependence on a watermarked signal | |
| CN110163787B (en) | Audio digital robust blind watermark embedding method based on dual-tree complex wavelet transform | |
| CN102859587B (en) | Watermark generator, watermark decoder, method for providing a watermark signal in dependence on binary message data, method for providing binary message data in dependence on a watermarked signal | |
| CN102265344A (en) | Methods and apparatus to perform audio watermarking imbedding and watermark detection and extraction | |
| CN104217725A (en) | Audio watermarking method based on multi-echo core | |
| CN102859586B (en) | Watermark generator, watermark decoder, method for providing a watermark signal in dependence on binary message data, method for providing binary message data in dependence on a watermarked signal | |
| CN101441870A (en) | Robust digital audio watermark method based on discrete fraction transformation | |
| CN104658542A (en) | Additive spread spectrum audio watermarking embedding method, additive spread spectrum audio watermarking detection method and additive spread spectrum audio watermarking embedding system based on orthogonality | |
| Dhar et al. | Digital watermarking scheme based on fast Fourier transformation for audio copyright protection | |
| US9978382B2 (en) | Method and apparatus for embedding and extracting watermark data in an audio signal | |
| RU2586845C2 (en) | Watermark decoder and method of generating binary message data | |
| CN100524461C (en) | Didigal audio waterprint method based on low density parity check code | |
| Huang et al. | A Fast and Low-Distortion Capacity Adaptive Synchronized Acoustic-to-Acoustic Steganography Scheme | |
| CN102523535A (en) | Video stream and global positioning system (GPS) real-time registration and synchronous transmission method based on digital watermarking technology | |
| CN102222504A (en) | Digital audio multilayer watermark implanting and extracting method | |
| Kojima et al. | An audio data hiding based on complete complementary codes and its application to an evacuation guiding system | |
| CN105374360A (en) | Interleaved additive spread spectrum audio watermark embedding method and detection method and system | |
| Fallahpour et al. | High capacity method for real-time audio data hiding using the FFT transform | |
| CN104538038B (en) | Audio frequency watermark insertion and extracting method and device with robustness | |
| CN103745725B (en) | A kind of audio frequency watermark embedding grammar based on constant Q transform | |
| Dhavale et al. | Walsh Hadamard transform based robust blind watermarking for digital audio copyright protection | |
| Erkucuk et al. | Robust audio watermarking using a chirp based technique |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20141217 |
|
| WD01 | Invention patent application deemed withdrawn after publication |