Sound source separation device, voice recognition device, mobile phone, sound source separation method, and program

摘要

Make it possible without being affected by variations in the sensitivity of the microphone element from the mixed sound source signals emitted from a plurality of sound sources are mixed, to separate the source signals from the target source. By multiplying a weighting factor in the complex conjugate relationship with respect to the output signal from the microphone 10 and 11 after the spectrum analysis, the beamformer section 3 of the sound source separation device 1, the straight line connecting the microphones 10, 11 of the two The beamformer performs processing for attenuating each of the sound source signals arriving from directions symmetrical with respect to vertical. Power calculation unit 40 and 41 calculates the power spectrum information, target sound spectrum extraction section 50, 51 to extract the spectral information of a target sound source on the basis of the difference between the power spectrum information.