VOICE INPUT DEVICE

PURPOSE:To detect a voice section extremely accurately by finding an envelope signal as power information which is irrelevant to phase characteristics from a waveform signal and processing it. CONSTITUTION:A microphone unit M1 is for voice and placed nearby a voice generation source and a microphone unit M2 is for noises and picks up noises from a noise source. The electric outputs of the microphone units are amplified by amplifiers 13 and 14 respectively, and energy extraction parts 18 and 18' extract their energy components, which are processed by subtraction 12. The output of this subtracter 12 is used by a section detection part 17 to detect a voice section when the output is certain energy level higher than a properly set threshold value or a soundless section when lower according to an envelope level. The output of the amplifier 13, on the other hand, is A/D-converted 15, analyzed 16 for feature extraction, and sent to the section detection part 17, where the voice section is segmented from the signal segmented with the envelope, thereby performing recognition and decision making 21 by using a word dictionary 19.