人脸语音动画中基于PSOLA的情感语音合成系统

Emotional speech synthesis system based on PSOLA in facial speech animation

提出一种基于时域基音同步叠加TD-PSOLA算法的情感语音合成系统。根据情感语音库分析总结情感规则,在此基础上利用TD-PSOLA算法对中性语音的韵律参数进行改变,并提出一种能够对基频曲线尾部形状改变的方法,使句子表达出丰富的情感。实验表明,合成出的语音具有明显的情感色彩,证明了该系统能以简单明了的方式实现情感语音的合成,有助于提高人脸语音动画表达的丰富性和生动性。

This paper proposed a emotional speech synthesis system based on pitch synchronous overlap-add（PSOLA）.Prosodic parameters could be changed in this system freely.First,analyzing pre-recorded emotional speech samples it concluded some acoustic features associated closely with happiness,angry,surprise and sadness.Then it used TD（time domain）-PSOLA algorithm to change the speech prosodic parameters of neutral speeches.Especially,it proposed a approach to change the F0 contour.Experiments demonstrates that the system is effective,which helps to express the facial speech animation more vivi-dly.