韵律参数和频谱包络修改相结合的情感语音合成技术研究  

Emotional speech synthesis based on the modification of prosody parameters and spectral envelope

作　　者：邵艳秋[1] 韩纪庆[1] 王卓然[1] 刘挺[1] 

机构地区：[1]哈尔滨工业大学计算机学院,哈尔滨150000

出　　处：《信号处理》2007年第4期526-530,共5页Journal of Signal Processing

摘　　要：情感语音合成可以增强合成语音的表现力、人情味,是近年来的新兴课题。除了韵律特征之外,音质类和发声器官类参数对情感语音的表达也有着至关重要的影响,而通常的研究大多都是基于规则或者预先为某种情感设计的滤波器来进行这两类参数的修改。本文提出了通过频谱包络综合地调整音质类和发声器官类参数来合成情感语音的方法,并通过实验验证了这一方法的有效性。另外,实验结果也显示了当韵律参数和频谱包络同时得到修改时,相对于单独修改某类参数可以获得更好的情感合成效果。Emotional speech synthesis, a recently developed research subject, is expected to make the synthesized speech more expressive and human-like. Besides prosody features, voice quality and articulatory parameters are also the important factors that should be considered in emotional speech synthetic systems. Generally, rules and filters are designed to process these two kinds of parameters respectively. This paper presents that by modifying spectral envelope, the voice quality and articulatory could be adjusted as a whole. The experiments results also show that when the prosody features and spectral envelope are all modified, the best synthetic emotional speech could be got.

关 键 词：情感语音合成 频谱包络 韵律修改 

分 类 号：TN912.33[电子电信—通信与信息系统]