| 摘要 |
PROBLEM TO BE SOLVED: To prevent deterioration in sound quality by keeping senses of articulation of speech signals from a plurality of sound sources well-balanced in a speech signal having been compressed and expanded even when the speech signals of the plurality of sound sources are mixed with an input speech signal when a pitch cycle is detected from the input speech signal and the time base of the input speech signal is compressed or expanded on the basis of the pitch cycle. SOLUTION: A speech signal processor is equipped with a pitch cycle detection section 2 which detects a plurality of candidates for pitch cycles of the input speech signal M on the basis of a signal generated by subjecting the input speech signal to filter processing when one pitch cycle is detected from the input speech signal M, a pitch cycle selection section 3 which selects one pitch cycle out of the plurality of candidates for pitch cycles on the basis of the input speech signal M or the signal generated by subjecting the input speech signal to other filter processing different from the filter processing, and a signal compression/expansion section 4 which performs compression processing or expansion processing for the time base of the input speech signal M on the basis of the selected one pitch cycle. COPYRIGHT: (C)2006,JPO&NCIPI
|
