To prevent a non-speech section of ambient noise, or sneeze or cough of a human from being misrecognized as a speech signal of the input speech of a user, and further, to eliminate errors such that the speech of a handicapped person, having frequency components exceeding the pitch range of everyday voicing by a nonhandicapped person, will not be decided as a speech.
This invention comprises a device which inputs a speech or sound signal and measures pitches for each analysis section (frame) for signal processing for sections decided as the start and end of an input signal so as to decide whether the input signal is a speech signal or a non-speech signal; a device which verifies the measured pitches and decides reliabilities of frames whose pitches are measured; a pitch frequency range deciding device which decides whether pitches of frames whose pitches are decided as reliable pitches are within the voice frequency range; and a device which decides whether a signal decided as a signal within the voice signal range is a speech signal or non-speech signal, based on the generation rate of pitches calculated by a pitch generation rate calculating device.
COPYRIGHT: (C)2007,JPO&INPIT
JP4346501 | Receiver |
JPS61273596 | VOICE SECTION DETECTION SYSTEM |
JPS6187199 | VOICE ANALYZER/SYNTHESIZER |
Lee Tochia
Hiroaki Kojima
JP8292787A | ||||
JP6083391A | ||||
JP2004061567A | ||||
JP9179587A |