| 摘要 |
According to the method, the speech signal is reconstituted by selecting, in a dictionary, excitation wave forms for long term (2) and short term (3) predictor filters. In order to keep the same "pitch" and the same spectral characteristics as the messages read at normal speed for the synthetic speech messages read at slow or fast speed, the excitation duration of the long term (2) predictor filter and the calculating ranges of the two predictor filters (2, 3) are modified as a function of the synthesis speed without the pitch values (M) applied to the long-term predictor (2) being modified. Application: dictaphones, telephone answering machines, voice messaging systems. <IMAGE> |
