VOICE DETECTION DEVICE, VOICE DETECTION METHOD, VOICE DETECTION PROGRAM, AND RECORDING MEDIUM

摘要

It is possible to improve a detection accuracy of a voice section in a conversation system allowing simultaneous utterance of more than one speakers via microphones.  A voice detection device (20) includes: a per-band power calculation unit (200) which calculates for each predetermined frequency width (sub-band), a sum (sub-band power) of powers of signals inputted from respective microphones; a per-band noise estimation unit (202) which estimates a noise power for each of the sub-bands; a per-band SNR calculation unit (203) which calculates a sub-band SNR for each of the sub-bands and outputs the largest sub-band SNR as the SNR of the microphone; and a voice/non-voice judgment unit (104) which judges whether a sound is a human voice or not by using the SNR.