基于小波变换的语音去噪阈值函数的研究

A Study of Speech Denoising Threshold Functions Based on Wavelet Transformation

目的小波去噪可以提高语音的信噪比,其中通过放宽阈值限制、加强小波系数的连续性,以及对小波系数较高的语音信号实现噪声与语音的分离可以增强语音去噪效果。方法基于一段信噪比为10dB的语音信号,采用不同的阈值函数,对该段语音进行小波去噪处理,比较各种阈值函数得到不同的去噪语音和各个阈值函数的特点。结果传统的硬阈值和软阈值得到的信噪比分别为11.57dB和10.91dB,软硬阈值折衷法得到的去噪语音信噪比12.70dB,模平方法得到的去噪语音信噪比为12.20dB,改进以后的软硬阈值折衷法和改进后的模平方法得到的去噪语音的信噪比分别为11.38dB和13.40dB,用降低阈值的方法得到的去噪语音信号的信噪比为11.68dB,新的阈值函数得到的去噪语音的信噪比为13.49dB。结论相比较软硬阈值折衷法,改进的模平方法和其它传统的方法,新阈值函数语音去噪可以取得更好的效果。

Objective Wavelet de-noising can improve the signal-to-noise ratio of speech,which can enhance the speech de-noising by relaxing the threshold,increasing the continuity of the wavelet coefficients and separating the noise and speech of the higher speech coefficients.Methods Based on a speech signal with a signal-to-noise ratio of 10 dB,different threshold functions are used to perform wavelet denoising on the segment speech,and different denoising and different threshold functions are compared.By summing up the conclusions,this paper selects the effective threshold functions,and proposes a new threshold function.Results The signal-to-noise ratios of traditional hard threshold and soft threshold are11.57 dB and 10.91 dB,and the results of the compromise of hard and soft threshold is 12.70 dB,and the signal-to-noise ratio of the square of amplitude is 12.20 dB,and the results of improved compromise method and the improved square of amplitude method are 11.38 dB and 13.40 dB,respectively.By using the lower threshold function,the denoised speech is got with,the signal-to-noise of which is 11.68 dB.And at last the result of the new threshold function is 13.49 dB.Conclusion Compared with the compromise of soft and hard threshold method,the improved square of amplitude and other traditional methods,the new threshold function can achieve better results in speech denoising.