APPARATUS AND METHOD FOR SIGNAL
SEPARATION AND RECORDING MEDIUM
FOR THE SAME
5
BACKGROUND OF THE INVENTION
The present invention relates generally to signal processing techniques. More particularly, the invention is concerned with a method and an apparatus for performing signal separation processings and a medium recording the signal 10 separation method in the form of a program executable with a computer.
In recent years, a human interface has been spotlighted according to the progress of computerization of consumer products. Especially, a hands-free operation is preferred in 15 the case of the car navigation system for safety and convenience, so that the expectation is increasing for a speech recognition system within a surrounding noise. As well known, a surrounding noise degrades the performance of speech recognizer dramatically. To overcome this 2o problem, the noise cancellers based on an adaptive algorithm such as LMS are used. Although they are effective when the system between noise source and observation is stable and noise is separately measurable, their performance degrades if measurement of noise is not precise or a transfer system is unstable.
The blind signal separation or blind noise canceller that does not require any reference noise signal, is preferred for these applications. There are several approaches to build blind signal separation systems. Because those alogrithms based on the gradient algorithm for convergence, there is a 30 similar problem on local minimums on cost function. Also, these algorithms use high order statistics, so that the computational load is not small.
In this paper, a new simple signal separation method is proposed, for example. This method separates signals using 35 the information on relative relationship between source signals.
In transmission of signals originating in different signal sources or systems, there may arise such situation that these signals undergo mutual interference or superposition with 40 given amplification factors in the course of transmission to such extent that they can not be discriminated by the receiver, as exemplified by crosstalk phenomenon. For coping with this problem, there has heretofore been known a technique for performing signal separation processing on the 45 received signals with a view to restoring the original signals from the mutually superposed state. With the conventional signal separation technique, the original signals of two discrete signal sources or systems sent through transmission line(s) or channels and received in mutually indiscernible 5Q state can certainly be restored approximately to the original sgnals.
For better understanding of the concept underlying the present invention, description will first be made in some detail of the conventional signal separation technique by 5J reference to FIG. 5 of the accompanying drawings which shows in a functional block diagram a typical one of the signal separation apparatuses known heretofore.
The signal separation apparatus shown in FIG. 5 includes a signal separation means or unit and a transmission channel characteristics estimation means or unit. In the figure, ref- 60 erence numeral 1 denotes a first filter element or module of a variable tap coefficient type for performing filtering operation or processing on an input signal received from the transmission path or channel and originating in a first signal source or system (hereinafter referred to as the first input 65 signal) with a given tap coefficient value, numeral 2 denotes a second filter element or module of a variable tap coefficient
type for performing filtering operation or processing on an input signal received from the transmission channel and originating in a second signal source or system (hereinafter referred to as the second input signal) with a given tap coefficient value, numeral 3 denotes a difference calculation module for arithmetically determining a difference between the second input signal and the output signal of the first filter module 1, numeral 4 denotes a difference calculation module for arithmetically determining a difference between the first input signal and the output signal of the second filter module 2, numeral 5 denotes a third filter element or module of a variable tap coefficient type for performing filtering operation or processing on the output signal of the difference calculation module 3 with a given tap coefficient value, numeral 6 denotes a fourth filter element or module of a variable tap coefficient type for performing filtering processing on the output signal of the difference calculation module 4 with a given tap coefficient value, numeral 7 denotes a first cross-correlation calculation module for arithmetically determining cross-correlation between the second input signal and the output signal of the difference calculation module 3, numeral 8 denotes a second cross-correlation calculation module for arithmetically determining crosscorrelation between the first input signal and the output signal of the difference calculation module 3, numeral 9 denotes a third cross-correlation calculation module for arithmetically determining cross-correlation between the second input signal and the output signal of the difference calculation module 4, numeral 10 denotes a fourth crosscorrelation calculation module for arithmetically determining cross-correlation between the first input signal and an output signal of the difference calculation module 4, numeral 11 denotes a first inverse function calculation module for arithmetically determining an inverse function of the output signal of the first cross-correlation calculation module 7, numeral 12 denotes a second inverse function calculation module for arithmetically determining an inverse function of the output signal of the third cross-correlation calculation module 9, numeral 13 denotes a first multiplication module for determining a product of output signals of the first inverse function calculation module 11 and the second cross-correlation calculation module 8, and numeral 14 denotes a second multiplication module for determining a product of the output signals of the second inverse function calculation module 12 and the fourth cross-correlation calculation module 10. As can be seen in FIG. 5, the signal separation unit is comprised of the first and second filter elements or modules 1 and 2, the first and second difference calculation modules 3 and 4 and the third and fourth filter modules 5 and 6, while the transmission channel characteristics estimation unit is constituted by the first to fourth cross-correlation calculation modules 7 to 10, the first and second inverse function calculation modules 11 and 12, and the first and second multiplication modules 13 and 14.
Next, referring to FIG. 6, description will be directed to operation of the conventional signal separation apparatus of the structure shown in FIG. 5.
For convenience of the description, the original signals of two different signal sources or systems are represented in terms of the time-based notation as follows.
