| 摘要 |
PURPOSE: To shorten calculation time and to make it possible to extract a pitch without generating any error even in a section with small sound power by executing pitch extraction in a section judged as a voiced section by voiced/ voiceless sound judgement and retrieving a zero crossing position appearing at first at the time of going back a time axis from a pitch pulse. CONSTITUTION: The voiced section of a voice inputted from a voice input terminal 11 is detected by a voiced/voiceless sound judging part 12. A pitch pulse extraction part 15 extracts pulses corresponding to the periodical vibration of vocal chords from a pulsed signal. Namely a pulse with maximum amplitude is searched in a section with time length corresponding to the inverse of pitch frequency supposed as the maximum frequency in the input voice and extracts a pitch pulse by deleting the other pulses in the section. A zero crossing position detecting part 16 detects a zero crossing position by going back the time axis from the pitch pulse obtained by the extraction part 15. Consequently pitch extraction having high processing speed and high accuracy even in a section having comparatively small input voice power can be attained. |
