广东工业大学学报 ›› 2017, Vol. 34 ›› Issue (05): 52-55.doi: 10.12052/gdutxb.160179

• 综合研究 • 上一篇    下一篇

基于维纳-小波分析的语音去噪新方法

郑永敏1, 鲍鸿1, 张晶2   

  1. 1. 广东工业大学 自动化学院, 广东 广州 510006;
    2. 广东外语外贸大学 信息科学技术学院, 广东 广州 510420
  • 收稿日期:2016-12-19 出版日期:2017-09-09 发布日期:2017-07-10
  • 通信作者: 张晶(1977-),女,副教授,主要研究方向为语音识别.E-mail:ha_go@163.com E-mail:ha_go@163.com
  • 作者简介:郑永敏(1991-),男,硕士研究生,主要研究方向为声音去噪与识别.
  • 基金资助:
    广东省科技计划资助项目（2013B040401015）

A New Speech Denoising Method Based on Wiener Filtering and Wavelet Analysis

Zheng Yong-min1, Bao Hong1, Zhang Jing2   

  1. 1. School of Automation, Guangdong University of Technology, Guangzhou 510006, China;
    2. Institute of Information Science and Technology, Guangdong University of Foreign Studies, Guangzhou 510420, China
  • Received:2016-12-19 Online:2017-09-09 Published:2017-07-10

HTML

PDF (PC)

3036

摘要： 小波分析具备多分辨性、低熵性和灵活选择基底的特性，是目前较常用的语音去噪技术之一.但是由于在小波分析去噪中需要设定合适的阈值消除含有噪声信号的小波系数，这会过滤部分有用的语音信号，降低去噪后的语音质量.为了解决这个问题，提出了一种将维纳滤波和小波分析进行结合的去噪新方法.该方法根据维纳滤波具备最小均方误差的特点，可以先过滤掉大部分噪声信号，再进行小波分析二次去噪.仿真实验结果表明，该新方法较小波分析去噪在信噪比和整体质量方面得到较大提升.

关键词: 维纳滤波, 小波分析, 阈值函数, 去噪

Abstract: Denoising in wavelet analysis sets the threshold to eliminate the wavelet coefficients of the noise, which filters some useful speech signals and reduces voice quality. In order to resolve the problem, a new speech denoising method based on Wiener filter and wavelet analysis is proposed. According to the characteristic of the least mean square error of the Wiener filter, the method filters out most of the noise signal, and then denoises it by the wavelet analysis. Experiments is carried out in simulation, and the results reveal that the method is more improved than wavelet analysis in terms of signal-to-noise ratio (SNR) and speech quality.

Key words: Wiener filtering, wavelet analysis, threshold function, denoising

中图分类号: 

  • TN912.35
[1] Mallat S. A theory for multiresolution signal decomposition:the wavelet representation[J]. IEEE Trans On Pattern Analysis and Machine Intelligence, 1989, 11(7):674-693. [2] 宋知用. MATLAB在语音信号分析与合成中的应用[M]. 北京:北京航空航天大学出版社, 2013:195-199. [3] 李战明, 尚丰. 一种基于语音端点检测的维纳滤波语音增强算法[J]. 电子设计工程, 2016, 24(2):42-44.LI Z M, SHANG F. A wiener filtering speech enhancement algorithm based on speech endpoint detection[J]. Electronic Design Engineering, 2016, 24(2):42-44. [4] 张植奎. 基于LabVIEW的语音信号小波去噪分析[D]. 天津:南开大学生物物理学院, 2009. [5] 朱宗明, 姜占才. 小波递归最小二乘语音自适应增强[J]. 电子设计工程, 2016, 24(1):69-72.ZHU Z M, JIANG Z C. Algorithm for speech adaptive enhancement of wavelet recursive least square[J]. Electronic Design Engineering, 2016, 24(1):69-72. [6] ZHOU Y, LAI S, LIU L, et al. An improved approach to threshold function de-noising of mobile image in CL2 Multi-wavelet transform domain[C]//Mobile Technology, Applications and Systems, 20052nd International Conference,[S.l.]:[s.n.], 2005:1-4. [7] 刘凤山, 吕钊, 张超, 等. 改进小波阈值函数的语音增强算法研究[J]. 信号处理, 2016, 32(2):203-213.LIU F S, LYU Z, ZHANG C, et al. Research on speech enhancement algorithm based on modified wavelet threshold function[J]. Journal of Signal Processing, 2016, 32(2):203-213. [8] 林颖, 常永贵, 李文举, 等. 基于一种新阈值函数的小波阈值去噪研究[J]. 噪声与振动控制, 2008, 1(1):79-81.LIN Y, CHANG Y G, LI W J, et al. A de-nosing algorithm of wavelet threshold based on a new thresholding function[J]. Noise and Vibration Control, 2008, 1(1):79-81. [9] 何文. 基于小波变换与模糊理论的医学超声图像增强与去噪方法研究[J]. 电子设计工程, 2015, 23(12):101-104.HE W. Enhancement and denoising method of medical ultrasound image based on wavelet transform and fuzzy theory[J]. Electronic Design Engineering, 2015, 23(12):101-104. [10] TUFEKCI Z, GOWDY J N. Feature extraction using discrete wavelet transform for speech recognition[J]. Department of Electrical and Computer Engineering, 2000, 4(7-9):116-123. [11] DANIEL V, DAVID O, JEFERSON S, et al. Comparison analysis between rigrsure, sqtwolog, heursure and minimaxi techniques using hard and soft thresholding methods[J]. IEEE Program of Electronics Engineering, 2016, 1(16):01-05. [12] 陆真, 裴东兴. 基于连续小波阈值函数的语音增强技术[J]. 山西电子技术, 2016, 1(1):40-42, 57.LU Z, PEI D X. The algorithm of image denoising based on the optimized wavelet thresholding function[J]. Shanxi Electronic Technology, 2016, 1(1):40-42, 57. [13] 秦叶, 陈新度, 冯文贤. 小波基的选择与振动信号分析[J]. 广东工业大学学报, 1997, 14(4):61-64.QIN Y, CHEN X D, FENG W X. Selection of wavelet basis and analysis of vibration signal[J]. Journal of Guangdong University of Technology, 1997, 14(4):61-64. [14] 韦力强. 基于小波变换的信号去噪研究[D]. 长沙:湖南大学电气与信息工程学院, 2007. [15] 雷鸣. 基于小波神经网络的车辆发动机故障诊断[J]. 制造业自动化, 2012, 34(6):24-26.LEI M. The fault diagnosis of vehicle engine based on wavelet neural network[J]. Manufacturing automation, 2012, 34(6):24-26. [16] 袁野, 王夏天, 张子辰, 等. 基于小波变换和改进的瞬态独立成分分析融合算法的心电信号降噪方法[J]. 中国医学物理学杂志, 2016, 33(4):415-422.YUAN Y, WANG X T, ZHANG Z C, et al. Electrocardiogram noise reduction based on fused algorithm of wavelet transform and improved independent component analysis[J]. Chinese Journal of Medical Physics, 2016, 33(4):415-422.
[1] 邓杰航,吕灼荣,肖灿辉. 胸部CT DICOM图像自动分割的研究与实现[J]. 广东工业大学学报, 2013, 30(2): 37-41.
[2] 王振友，郑少杰，何远才，龙洁丽，苏娜，林洁芝. 基于PCA方法的PET图像多示踪剂分离[J]. 广东工业大学学报, 2013, 30(2): 42-46.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2017 《广东工业大学学报》编辑部
地址：广州市东风东路729号广东工业大学 邮编：510090
电话：020-37626653,020-37626139,020-37626396
邮箱：xbzrb@gdut.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发