SOUND SOURCE SEPARATOR DEVICE, SOUND SOURCE SEPARATOR METHOD, AND PROGRAM

摘要

According to a conventional sound source separation device, in an environment where diffuse noises having arrival direction uncertain are present, a certain frequency domain is largely suppressed, and the diffuse noises are irregularly sorted into sound source separation results, resulting in musical noises. According to an aspect of the invention, a beamformer unit (3) of a sound source separation device (1) multiplies output signals by microphones (10 and 11) having undergone spectrum analysis by a weighting factor in a relationship of complex conjugate, thereby performing a beamformer process of attenuating respective sound source signals arrived from a region including a general direction of a target sound and a region opposite to that region with a plane intersecting with a line interconnecting the two microphones (10 and 11) being as a boundary. A weighting-factor calculation unit (50) calculates, based on a difference between power spectrum information calculated by power calculation units (40 and 41), the weighting factor.