| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a voice processing method which estimates multiple sound model parameters together with multiple excitation model parameters to maximize the likelihood of a speech waveform. <P>SOLUTION: When a text is inputted, a voice processing device generates a voice output in accordance with the input text using a probabilistic theory model including a sound model which contains multiple model parameters for describing multiple probability distributions to associate a word or the part of a word with a feature, and a excitation model which contains excitation model parameters used for modeling a voice cord and a lung to generate the voice output using the feature. The sound parameters and the excitation parameters are estimated together and the voice output is generated. <P>COPYRIGHT: (C)2012,JPO&INPIT |
