强噪声环境下语音增强算法的比较研究

介绍语音增强的原理和从强噪声背景中提取语音信号的方法,并对基于减谱法的增强算法、基于自适应滤波法的增强算法和基于小波变换的增强算法进行对比研究。鉴于语音增强算法的两个目标即增强语音的清晰度与理解度并不是相关联的,有时甚至相互矛盾,因此任何一个语音增强算法都是根据不同的应用做适当的选择和折衷。基于此,对三种算法进行仿真实验的比较研究,实验结果证明应用小波变换的语音增强算法比其它的方法更有效。

Comparative Study on Speech Intensification Algorithms in Strong Noise Background

The principle of speech enhancement and the method of extracting speech signals from strong noise background are introduced. Three kinds of speech intensifying algorithms based on spectrum subtraction algorithm, wiener filtering and wavelet transform, respectively, are comparatively studied. However, the two purposes of the speech enhancement algorithms, to enhance articulation and to enhance the understandability, are not related each other, and even in contradiction sometimes. Therefore, a good speech enhancement algorithm is in fact a compromise between the two purposes according to the requirement of application. In this paper, results of numerical experiment of the three algorithms are compared and studied. The results demonstrate that the speech enhancement algorithm using wavelet ransform is more efficient than the others.