OM-LSA和小波阈值去噪结合的语音增强

计算机科学与探索 ›› 2011, Vol. 5 ›› Issue (6): 547-552.

OM-LSA和小波阈值去噪结合的语音增强

刘凤增¹, 李国辉^1,2, 李博¹

1. 国防科学技术大学信息系统与管理学院, 长沙 410073
2. 国防科学技术大学信息系统工程重点实验室, 长沙 410073

收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-06-01 发布日期:2011-06-01

Speech Enhancement with OM-LSA Incorporating Wavelet Thresholding

LIU Fengzeng, LI Guohui, LI Bo

1. School of Information System and Management, National University of Defense Technology, Changsha 410073, China 2. Key Laboratory of Information System Engineering, National University of Defense Technology, Changsha 410073, China

Received:1900-01-01 Revised:1900-01-01 Online:2011-06-01 Published:2011-06-01

摘要/Abstract

摘要： 针对OM-LSA(optimally modified log-spectral amplitude estimator)算法产生的残留噪声, 提出了一种结合OM-LSA和小波阈值去噪的语音增强算法。首先, 进行语音对数幅度谱估计; 然后, 估计残留噪声, 利用带噪语音第一级小波系数和语音不存在时的增益函数进行估计, 解决了常规方法对增强后语音噪声估计不准确的问题; 最后, 在小波域利用软阈值法对语音信号进行阈值处理。实验结果表明, 提出的算法有效地去除了OM-LSA算法中的残余噪声, 在分段信噪比(segmental signal-to-noise ratio, SegSNR)和对数谱失真(log-spectral distortion, LSD)等指标评价上有较大的提高。

关键词: 小波阈值去噪, 残留噪声, 最优改进对数谱幅度估计(OM-LSA), 语音增强

Abstract: Aiming at the residual noise produced by OM-LSA (optimally modified log-spectral amplitude estimator) in speech enhancement, the algorithm of OM-LSA incorporating wavelet thresholding is proposed. First, log-spectral ampli-tude of speech is estimated. Second, the estimation of residual noise is done by the gain function during speech absence and the wavelet coefficients of noisy speech, which solves the problem that general methods can’t get exact noise estimation from enhanced speech. Finally, the signal is processed by soft-thresholding method in wavelet domain. Experimental results show that, compared with OM-LSA, the proposed algorithm can lead to significant reduction of the residual noise, improve SegSNR (segmental signal-to-noise ratio) and reduce LSD (log-spectral distortion).

Key words: wavelet thresholding, residual noise, optimally modified log-spectral amplitude estimator(OM-LSA), speech en-hancement

刘凤增1 , 李国辉1,2 , 李博1 . OM-LSA和小波阈值去噪结合的语音增强[J]. 计算机科学与探索, 2011, 5(6): 547-552.

LIU Fengzeng, LI Guohui, LI Bo. Speech Enhancement with OM-LSA Incorporating Wavelet Thresholding[J]. Journal of Frontiers of Computer Science and Technology, 2011, 5(6): 547-552.