文档介绍：摘要
语音增强是语音信号处理的重要分支也是语音编码语音识别语音合成等技
术的重要预处理环节随着移动语音业务和人机交互通信技术的发展语音增强的研
究及其实际应用成为数字化通信中的一个重要研究方向近年来发展迅速的高速数字
信号处理器 DSP 是专门为实现数字信号处理中的算法而设计的目前已经广泛应
用于通信语音图像等多个领域本文以基于 DSP 的语音处理平台系统项目的
研发为背景对基于 DSP 的语音处理平台上的语音增强算法进行专题研究并实时实
现
本论文首先从基本的自适应滤波器理论开始对线性均方准则和基于此准则的传
统的递归最小二乘 RLS 算法的原理进行研究进而分析和研究由 RLS 算法发展而
来的两种语音增强算法基于 RLS 算法的延迟改进型噪声抵消算法和基于累积量
的 RLS CRLS 算法在掌握 TI 公司的定点数字信号处理芯片 TMS320C55x 的特性
和实际开发流程的基础上提出并实现了基于简化型 RLS 算法的延迟改进型噪声抵消
算法同时也将 CRLS 算法在 DSP 平台上实时实现最后详细测试了两种算法的滤波
性能和实时性能
实验证明基于简化型 RLS 算法的延迟改进型噪声抵消算法和 CRLS 算法这两种
算法都能满足系统的要求但两种算法的实时性能和滤噪性能则各有千秋在实际应
用中可以根据需要进行选择
关键词: 数字信号处理器语音增强自适应噪声抵消递归最小二乘算法
累积量 TMS320C55x
II
Abstract
Speech Enhancement is the most important part of speech signal processing. It is also
an important pretreatment technology of speech coding, speech recognition and speech
synthesis. With the development of mobile speech and man-machine conversation
technology, the research and application of speech enhancement es a significant
aspect. High Speed Digital Signal Processor (DSP) developed in recent years is specially
designed for algorithms of digital signal processing. Nowadays, it is widely used in many
fields such munications, speech processing, video processing, etc.. This essay
focuses on the research and real-time implementation of speech enhancement algorithms on
speech processing platform, while it takes the project of “Speech Processing Platform based
on DSP” as background.
This essay begins with the basic theory of adaptive filter. And then the principium of
Mean Square Error and the normalized Recursive Least Square (RLS) are studied. After
that, two speech enhancement algorithms based on RLS are presented: noise cancellation
system based on RLS, and Cumulant based RLS (CRLS) algorithm. Thirdly, the characters
of TI’s TMS320C55x and development flow are introduced, and the two algorithms are
simplified and optimized implemented on the platform. At last, the detailed tes