一种自然语流的声源分析方法  

Novel Glottal Analyzing Algorithm for Natural Utterance

作　　者：周斌[1] 戴礼荣[1] 凌震华[1] 王仁华[1] 

机构地区：[1]中国科学技术大学电子工程与信息科学系,合肥230027

出　　处：《数据采集与处理》2005年第3期297-301,共5页Journal of Data Acquisition and Processing

摘　　要：在研究传统的逆滤波和声源建模相结合的分析方法所存在的问题的基础上,提出了一种改进的声源分析算法和利用该算法自动分析自然语流的方法。该方法对于传统的方法所存在着的鲁棒性低、分析不精确、不能自动分析大规模的自然语流等问题具有较好的改进;文中给出的对合成元音以及自然语流的实验结果验证了所提方法的有效性。Based on the analysis of the existed problems of traditional vocal tract modeling and glottal measurement extraction methods, an improved glottal source analyzing method is presented and an algorithm based on the method is used to analyze continuous natural utterance automatically. The method improves the accuracy and robustness of glottal source analysis, and also exhibits good performance in automatically analyzing large scale of continuous natural utterance compared with the traditional method. Experimental results of synthesized vowels and continuous natural utterances prove the effectivity of the method.

关 键 词：声源 源-滤波器模型 逆滤波 自然语流 

分 类 号：TN911[电子电信—通信与信息系统]