Method and system for speech analysis with input via a microphone assembly

摘要

PROBLEM TO BE SOLVED: To make possible precisely estimating a voice source position or direction with less operation amounts. ;SOLUTION: The device and method are constituted so that an input signal from a microphone array inputted by a voice input part 1 is frequency divided by a band-pass filter in a frequency analysis part 2, and a band-pass waveform at every frequency band is obtained by microphone channel classifications, and a band-pass power distribution at every voice source position or direction is obtained at every frequency band from the band-pass waveform by a voice source position inquiry part 3, and the band-pass power distribution is integrated ranging to all frequency bands, and the voice source position or direction is estimated from the integrated band-pass power distribution. A voice parameter extraction part 4 extracts the relevant band-pass power from the band-pass power distribution at every frequency band as a voice parameter based on the voice source position or direction, and a voice recognition part 5 uses it for voice recognition.;COPYRIGHT: (C)1997,JPO