基于多分辨率小波阈值去噪的响度补偿方法

A Loudness Compensation Method Based on Multiresolution Wavelet Threshold Denoising

进行多通道响度补偿时构建数字滤波器比较繁琐,并且产生的噪声不易消除,在增强了语音的同时,噪声也被放大。为了解决这些问题,文中提出利用小波阈值去噪和多分辨率分析的方法在小波域对消噪后信号进行响度补偿,并且对补偿后的声强进行控制,最后利用小波的完美重构特性对信号进行重构。一方面小波多分辨率分析能够满足多通道响度补偿对频带的非等宽划分,并且小波的阈值去噪可以有效防止噪声的放大;另一方面直接在小波域对补偿后声强进行控制,可以使声强始终处于听觉的舒适域内。经实验仿真,达到理想的效果。

It is complex to construct digital filter to reaiize multi-bands loudness compen^tion. Also, the noise will be amplified when en hance person＇s voice because the noise is hard to eliminate. In this paper the method of using wavelet threshold denoising and multi-reso lution analysis will be used to compensate the loudness of voice and control the gain of intensity of voice. At last,reconstruct the signal u sing new wavelet coefficients. On the one hand, multi-resolution analysis of wavelet can meet the requirement of dividing different band width easily, also, it can avoid the noise amplification effectively by setting a thread of wavelet. On the other hand, using wavelet can con trol the gain of voice in wavelet domain directly and make the intensity of voice comfortable to listeners. According to experiment and simulation, it can achieve the ideal effectively.