摘要 |
PURPOSE:To prevent the naturalness of a low-speed reproduced sound by clearly setting the border point between a voiced selection and a voiceless section and controlling connecting operation for a voice waveform by using information on the border point. CONSTITUTION:The voiced section and voiceless section of a voice signal are clearly discriminated under constant conditions to extract the point of time corresponding to the border point and merge it with a time point extracted from the voice waveform, and address information on a memory 8 corresponding to the generation point of time of the obtained time point sequence is stored in a memory 27 in sequence. In this case, the address data is read out of the storage device 27 at the time of demodulation, a coincidence with an address indicated by a counter 9 driving the storage device 8 is detected, and the address of the storage device 8 is varied with the detection signal; and then data read out of the storage device 8 is updated and this process is repeated to eliminate the mixture of an unnecessary signal in the voiceless section in the low-speed reproduction of a voice signal having an intermittent part. Consequently, the naturalness of the reproduced voice signal is not spoiled. |