VOICE SIGNAL PROCESSING APPARATUS, VOICE SIGNAL PROCESSING METHOD, AND PROGRAM

摘要

PROBLEM TO BE SOLVED: To provide an apparatus for separating with high accuracy a command of a specified sound source from a voice signal in which a plurality of sounds are mixed.

SOLUTION: Learning data consisting of base frequencies B corresponding to sound sources respectively is generated based on a learning voice signal containing sounds from a plurality of the sound sources, and all base frequencies Ball combining the base frequencies B which correspond to the sound sources respectively. Moreover, time frequency analysis is applied to an input voice signal to produce a time frequency analysis result. Base decomposition applying the all base frequencies Ball is applied to the time frequency analysis result for the input voice signal to produce a base activity H for the input voice signal. Identification processing of the base activity H produced last is executed to perform command identification. Highly accurate command identification is achieved by sound source separation processing based on the learning data.

COPYRIGHT: (C)2012,JPO&INPIT