汉语耳语音转换为正常音的研究  

Reconstruction of Normal Speech from Chinese Whispers

作　　者：樊星[1] 卢晶[1] 徐柏龄[1] 

机构地区：[1]南京大学声学所,江苏南京210093

出　　处：《电声技术》2005年第12期44-47,共4页Audio Engineering

基　　金：国家自然科学基金资助项目(60272037)

摘　　要：耳语音转换为正常音,可应用于公共场所的通信、失语者的语音恢复和司法工作的某些特殊需要等方 面。由于耳语发音和正常音在激励源和声道传递函数上存在差异,所以一般的语音合成方法不再适用。分析了耳语 音的发音及声学特点,初步建立了一个基于同态信号处理方法并结合相对熵声韵分割的耳语音重建系统,人耳对 恢复出的正常音可以正确分辨。Reconstruction of normal speech from Chinese whispers can be applied in several fields, such as the private speech communication in public, reconstruction of normal speech for the aphonic individuals, the special need for the forensic work and etc. However, because of the differences between normal and whispered speech in source and vocal tract transfer function, it is difficult to reconstruct whispers to normal speech with the ordinary synthesis method. In this paper, a reconstruction system based on the homomorphie signal processing and the symmetric relative entropy to detect the initial and unit has been founded. The Chinese whispers can be successfully transformed into normal speech using this method, and the reconstructed speech can be distinguished by human ear exactly.

关 键 词：耳语音 同态信号处理 语音合成 

分 类 号：TN912.34[电子电信—通信与信息系统]