一种基于LMS算法的语音消噪数字设计实现  

A Digital Speech Denoising Design Based on LMS Algorithm

作　　者：徐泽琨 黄明[1] 彭召敏 马栋梁 李国诚 XU Ze-kun;HUANG Ming;PENG Zhao-min;MA Dong-liang;LI Guo-cheng(North China University of Technology, Beijing 100144, China)

机构地区：[1]北方工业大学,北京100144

出　　处：《工业技术创新》2018年第3期52-57,共6页Industrial Technology Innovation

基　　金：北方工业大学电子信息工程学院电子系;北京市大学生科学研究与创业行动计划项目资助与支持

摘　　要：从最小均方算法LMS(Least Mean Square)基本原理出发,利用该原理实现语音信号消噪。研究了LMS算法不同步长对处理结果的影响,提出了LMS算法,设计了滤波器,实现了消噪的数字设计。理论分析及实验结果表明,在合适的步长下,采用LMS算法的自适应滤波器,可以实现对含有噪声干扰的语音信号中恢复实际信号,且实验结果具有良好的收敛性、较小的稳态误差。算法和设计对同行有参考和借鉴作用。Starting from the basic principle of LMS（Least Mean Square）, this principle is used to achieve denoising of speech signals. The influence of LMS algorithm on the processing result is studied. The LMS algorithm is proposed and the filter is designed. The digital design of denoising is realized. Theoretical analysis and experimental results show that the adaptive filter using the LMS algorithm can recover the actual signal in a speech signal with noise interference at an appropriate step size, and the experimental results have good convergence and small steady state error. The algorithm and design have reference and lesson for peers.

关 键 词：最小均方算法(LMS) 语音消噪 自适应滤波 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]