| 摘要 |
PROBLEM TO BE SOLVED: To provide a voice emphasizing device which has high precision when a noise mixed with a voice picked up by a microphone is suppressed. SOLUTION: The voice emphasizing device includes a voice emphasizing unit 113 which applies a voice emphasizing filter to a voice signal with which the noise is mixed to generate a voice-emphasized signal, a noise estimating filter generation unit 120 which stores a sound-received signal and performs adaptive learning of a filter by independent component analysis algorithm to suppress the noise by integration with microphone array technique, a noise estimating unit 114 which generates a noise-emphasized signal by using the generated filter, and a subtraction processing unit 115 which outputs a voice signal obtained by subtracting the noise-estimated signal generated by the noise estimating unit 114 from the voice-emphasized signal generated by the voice emphasizing unit 113, wherein precision of the filter generated by the noise estimating filter generating unit 120 is estimated on the basis of distribution kurtoses of the estimated voice signal and filter update is performed only when the filter precision is satisfactory. COPYRIGHT: (C)2010,JPO&INPIT |
