基于VDM和小波阈值的语音降噪方法研究  

Research on Speech Noise Reduction Based on VDM and Wavelet Threshold

作　　者：熊林林 宫翔 梁勇[1] 冉兴隆 XIONG Linlin;GONG Xiang;LIANG Yong;RAN Xinglong(Xichang Satellite Launch Center,Xichang 615000,China)

机构地区：[1]西昌卫星发射中心,四川西昌615000

出　　处：《电子质量》2023年第5期68-72,共5页Electronics Quality

摘　　要：针对带噪声的语音信号,提出了基于变分模态分解(VMD)和小波阈值相结合的去噪方法。首先,采用VMD对带噪声道信号进行分解,从而得到分解后的固有模态(IMFs)信号;然后,引入T检验对各IMFs分量进行低频和高频区分,对其中的高频分量用小波阈值进行噪声滤除;最后,对将低频IMFs分量和滤波后的高频IMFs分量进行重构,从而得到降噪后的具有较高信噪比的语音信号。通过实验验证,本方法与EMD和小波阈值联合算法相比,具有更好的去噪性能,具有一定的实用价值。A denoising method based on the combination of variational modal decomposition(VMD)and wavelet threshold is used.Firstly,VMD is used to decompose the noise channel signal,and the decomposed intrinsic mode(IMFs)signal is obtained.Then,T test is introduced to distinguish the low frequency and high frequency of each IMFs component,and the high-frequency component is filtered by wavelet threshold.Finally,the low frequency IMFs component and the filtered high frequency IMFs component are reconstructed to obtain the voice signal with high signaltonoise ratio after noise reduction.Through the experimental verification,it is confirmed that the proposed method has better demosing performance compared with the combined algorithm of EMD and wavelet threshold,and it has certain practical value.

关 键 词：变分模态分解 小波阈值 语音降噪 T检验 

分 类 号：TN912.3[电子电信—通信与信息系统]