Method, medium, and apparatus for extracting target sound from mixed sound
A method, medium, and apparatus for extracting a target sound from a mixed sound. The method includes obtaining the mixed signal from a microphone array, generating a first signal which is emphasized and directed toward a target sound source, and a second signal which is suppressed and directed toward the target sound source, calculating a non-linear filter which is adaptive to at least one of an amplitude ratio of the first signal to the second signal in a time-frequency domain, frequencies of the first and second signals, and a ratio of an interference signal to the mixed signal, and filtering the first signal by the non-linear filter.
1. A method of adaptively extracting a target sound signal from a mixed signal, the method comprising:
obtaining the mixed signal from a microphone array;
generating a first signal which is emphasized and directed toward a target sound source, and a second signal which is suppressed and directed toward the target sound source;
calculating a non-linear filter which is adaptive to at least one of an amplitude ratio of the first signal to the second signal in a time-frequency domain, frequencies of the first and second signals, and a ratio of an interference signal to the mixed signal; and
filtering the first signal by the non-linear filter.
2. The method of claim 1 , wherein the calculating of the non-linear filter comprises determining a coefficient of the non-linear filter in such a manner that the greater the amplitude ratio, the greater an output value of the non-linear filter.
3. The method of claim 1 , wherein the calculating of the non-linear filter comprises determining a coefficient of the non-linear filter in such a manner that the greater the frequencies of the first and second signals, the smaller an output value of the non-linear filter.
4. The method of claim 1 , wherein the calculating of the non-linear filter comprises determining a coefficient of the non-linear filter in such a manner that the greater the ratio of the interference signal to the mixed signal, the smaller an output value of the non-linear filter.
5. The method of claim 1 , wherein the calculating of the non-linear filter comprises calculating the non-linear filter by using a sigmoid function which is adaptive to at least one of the amplitude ratio, the frequencies of the first and second signals, and the ratio of the interference signal to the mixed signal.
6. The method of claim 1 , further comprising detecting a direction of the target sound source from the mixed signal by using a predetermined sound source search algorithm.
7. The method of claim 6 , wherein the predetermined sound source search algorithm is used to determine a direction relative to the microphone array of a sound source generating a sound signal having a relatively high SNR (signal-to-noise ratio) compared to SNRs of sound signals generated by a plurality of sound sources around the microphone array, the determined direction directing towards the target sound source.
8. A computer readable recording medium on which a program causing a computer to execute the method of claim 1 is recorded.
9. An apparatus for adaptively extracting a target sound signal from a mixed signal, the apparatus comprising:
a microphone array receiving the mixed signal;
a beam former generating a first signal which is emphasized and directed toward a target sound source, and a second signal which is suppressed and directed toward the target sound source;
a non-linear filter calculation unit calculating a non-linear filter which is adaptive to at least one of an amplitude ratio of the first signal to the second signal in a time-frequency domain, frequencies of the first and second signals, and a ratio of an interference signal to the mixed signal; and
a signal extractor extracting the target sound signal from the first signal by filtering the first signal by the non-linear filter.
10. The apparatus of claim 9 , wherein the non-linear filter calculation unit comprises an amplitude ratio calculation unit calculating the amplitude ratio of the first signal to the second signal in the time-frequency domain, as an input variable of the non-linear filter.
11. The apparatus of claim 9 , wherein the non-linear filter calculation unit comprises a frequency-adaptive coefficient calculation unit calculating an inclination coefficient of the non-linear filter by using the frequencies of the first and second signals.
12. The apparatus of claim 9 , wherein the non-linear filter calculation unit comprises an interference signal ratio-adaptive coefficient calculation unit calculating the ratio of the interference signal to the mixed signal as a bias coefficient of the non-linear filter.
13. The apparatus of claim 9 , further comprising a sound source search unit detecting a direction of the target sound source from the mixed signal by using a predetermined sound source search algorithm.
14. The apparatus of claim 13 , wherein the predetermined sound source search algorithm is used to determine a direction relative to the microphone array of a sound source generating a sound signal having a relatively high SNR (signal-to-noise ratio) compared to SNRs of sound signals generated by a plurality of sound sources around the microphone array, the determined direction directing towards the target sound source.