语声转换技术发展及展望

An Overview of Voice Conversion

语声转换通过改变语音信号的声学特征参数来调整语音的个性特征,从而使得转换后的源说话人语音听起来就像是目标说话人的声音一样。系统地介绍了当前语声转换技术的发展状况,在描述语声转换技术的应用场景和系统框架的基础上,着重阐述了系统的转换模块,即声道特性的转换和韵律转换,特别是重点介绍了声道特性的转换算法。简要地介绍了系统性能的测试方法,最后对全文进行了总结,并针对当前语声转换技术还存在的一些问题,对未来的发展进行了展望。

Voice conversion attempts to transform the personal characteristics of speech through adapting the acoustic parameters. The object is to make the speech uttered by a particular source speaker sound as if spoken by a designed target speaker. This paper introduces the development of voice conversion techniques in details. Firstly, the application of voice conversion and its system framework are described. Then, current conversion algorithms for the characteristics of vocal tract and prosody are presented, which is the core process of voice conversion. After that, the system performance evaluation methods, including subjective and objective measure, are introduced. Finally, the summary is given with a discussion of some existing problems in the current proposed algorithms.