基于小波变换的音素分段算法  

Phoneme Segmentation Algorithm Based on the Wavelet Transformation

作　　者：黄蓉[1] 汤一彬[1] 孙大飞[1,2] 吴镇扬[1] 

机构地区：[1]东南大学信息科学与工程学院,江苏南京210096 [2]南京工业大学理学院,江苏南京210009

出　　处：《信息与电子工程》2009年第1期44-47,共4页information and electronic engineering

摘　　要：在语音信号处理中,作为预处理前端的语音分段技术对于语音增强、编码和识别都有极其重要的作用。本文分析比较了几种传统的语音分段算法的性能,重点研究基于小波变换的语音音素分段算法,针对平滑渐变的低频能量不能对音素进行有效分割的缺点,提出了基于小波变换累积能量包络的语音分段算法。但由于小波的敏感性,会出现一定数目的伪点,本文提出2种去除伪点的方法,从而提高了音素分段的精确度。In speech signal processing, the techniques of speech segmentation as front end of preprocessing are of great importance to speech enhancement, coding and recognition. The performance of several typical algorithms of speech segmentation is analyzed by comparing with each other. It put emphasis on the study of the algorithm based on the wavelet transformation. In order to solve the problem that the smooth and gradual changing low frequency component can not segment the speech efficiently, an algorithm is put forward based on the cumulate energy of the wavelet transformation. But due to the wavelet sensitivity, it will present certain number of false spots. Therefore, two methods are proposed to remove the false spots. The precision of the segmentation can be promoted on the phoneme level accordingly.

关 键 词：语音分段 小波变换 累积能量 伪点 

分 类 号：TN912.3[电子电信—通信与信息系统]