基于FD—PSOLA算法的语音合成分析方法

介绍了一种基于FD－PSOLA算法来实现汉语韵律特征的修改。在短时信号频域修改的过程中，通过同态滤波处理分离了频谱包络和激励源频谱，并通过修改频率轴坐标来实现激励源频谱的压缩或拉伸。实验结果表明，FD－PSOLA算法比TD－PSOLA算法更适合于较高频率调整范围的语音合成分析。

Analyzing Speech Synthesis Based on FD-PSOLA

This paper introduces FD-PSOLA (Frequency Domain Pitch-Synchronous Overlap-add),which is used to modify prosodic features of the Chinese speech. In the process of frequency-domain modification of short-time signals,the spectral envelope and activating source spectrum get separated by homomorphic filtering,which is compressed or stretched by adjusting the frequency axis. Experiments show that FD-PSOLA is more suitable than TD-PSOLA for speech synthesis analysis in higher frequency adjustment.