摘要 |
PROBLEM TO BE SOLVED: To provide a speaking speed converting device capable of converting a speaking speed by a simple method while keeping original sound quality. SOLUTION: This is a speaking speed converting device which converts a speaking speed of a recorded speech by extending/compressing only a stationary section of a speech waveform by inserting/deleting a fundamental section waveform (pitch period waveform) of the speech waveform into/from the speech waveform. In this case, this device comprises a threshold value input part 3 and a speaking speed conversion rate input part 4 which give a waveform similarity lower limit β and a speaking speed conversion rate α as parameters, a frame slicing part 5 for sequentially slicing an input speech waveform into short frame sections from the head, an auto-correlation function calculating part 6 for calculating an auto-correlation function of the waveform in a sliced frame section, and a waveform inserting/deleting part 8 which regards the waveform in the frame section as stationary when a maximum value y(T) in an auto-correlation function is larger than the waveform similarity lower limit β, and inserts/deletes a pitch section waveform having a period T giving the maximum value to the frame section waveform as a fundamental period. |