一种基于能零积的综合去噪方法

基于短时能零积提出一种改进的语音综合去噪方法.通过短时能零积实现噪音阈值选择和噪音功率谱的自适应调整,得到较理想的语音起止点和噪音的功率谱估计.根据噪音在语音段和无音段不同的分布特征,噪音功率在不同阶段设置不同的权重,实现更好的谱减去噪.对每连续三帧进行残留噪音处理,可进一步除去残留噪音.实验表明,本法去噪效果优于使用固定噪音功率谱估计的传统谱减去噪方法.

A multiple de-noising method based on energy-zero-product

An improved speech de noising method by modifying the spectral subtraction technique based on the energy-zero-product was presented.Firstly,the power spectrum of the noisy speech and its threshold can be adjusted via the short time energy-zero-product.So the exact begin end points as well as the accurate noise power spectrum are obtained.Secondly,during the course of de noising using the modified spectral subtraction technique,different weights are assigned to the noise power spectrum based on the different characteristics of the noisy signal in both the speech and non speech segments.Finally,the method of residual noise reduction is applied to the residual noise.Simulation experiment results illustrate that this method is more effective than the conventional spectral subtraction methods using the fixed noise power spectrum.