一种基于HHT的语音增强算法研究与仿真  

Research and Simulation of Speech Enhancement Algorithm Based on HHT

作　　者：汝振[1] 李昕[1] 陈飞[2] 杨李箭[1] 李翔[1] 

机构地区：[1]上海大学机电工程与自动化学院,上海200072 [2]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机技术与发展》2010年第8期116-119,共4页Computer Technology and Development

摘　　要：语音增强是语音识别、语音编码等语音分析系统中一个重要的环节,直接影响到语音信号的处理过程。回顾了语音增强算法的发展历程,提出一种基于希尔伯特黄(HHT)算法的语音增强算法。使用HHT对含噪语音进行EMD变换,得到各阶IMF分量和残差,对高频的IMF分量采用小波去噪的软阈值方法进行处理,把经过阈值处理的高频IMF和低频IMF相加,得到增强的语音信号。针对三种典型噪声(白噪声、babble噪声和volvo噪声)的降噪问题,应用HHT算法取得了良好的结果。Speech enhancement is a critical step in the fields of speech analysis,recognition,coding and synthesis,which directly affects the performance of the speech digital signal processing system.In this study,retrieved the history of speech enhancement and promoted a new pattern of speech enhancement based on HHT,using wavelet packet transform（WPT） to improve the pretreatment of HHT.After that got IMF derivation and residual.The high frequency IMF has been denoised by using wavelet.The high frequency and low frequency signal that have been dealt with threshold were added,then got the enhanced speech signal.As to denoise of three typical noise（white,babble,volvo）,the experiment results indicate that the proposed algorithm get good results.

关 键 词：语音增强 希尔伯特-黄 谱减法 小波 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]