| 摘要 |
PROBLEM TO BE SOLVED: To synthesize a voice waveform giving a more natural impression to a listener. SOLUTION: In a voice synthesizer 100, a parameter determination part 20 performs phoneme analysis or the like on the basis of document data and supplies pitch information P and gain information G to be used for voice waveform synthesis to a sound source waveform generator 30. In this case, an information representative of the degree of voice hoarseness brought about by subtle fluctuation of the pitch or the gain, which is included in a vocal waveform, is outputted also from the parameter determination part 20. A pulse generation part 33 of the sound source waveform generator 30 generates a pulse train according with the pitch information P and the gain information G and corrects each pulse of the pulse train on the basis of information representative of the degree of hoarseness. In the case of this correction, the pulse generation part 33 refers to a correction information database 34 where correction information for reproducing subtle fluctuation of the pitch and the gain is preliminarily stored.
|
