基于基音同步的时频域插值的汉语语音合成

Mandarin speech synthesis based on pitch synchronous time-frequency interpolation

针对TD-PSOLA韵律调整能力的不足,将基于基音同步的时频域插值(TFI)方法应用于汉语语音合成中,该方法能够保证基频调整和时长的调整不会相互影响.为了提高计算精度,在频谱的插值计算中还引入了差商型插值方法.实验结果表明,采用差商型插值的TFI方法能取得比较好的合成效果.

The ability of wide range of prosody adjustment will be helpful to improve the naturalness and expressiveness of synthesized speech. The method of pitch synchronous time-frequency interpolation （TFI） is applied to Mandarin speech synthesis. Compared with TD-PSOLA, TFI can adjust the pitch without affecting the duration. The difference quotient interpolation is also used to improve the accuracy of the interpolation cal-culation, which shows satisfied quality of synthesized speech.