多通道皮肤听声系统中语音增强算法的应用研究  

On Speech Enhancement Algorithm for the Multi-channel Skin-Hearing System

作　　者：宫洵[1] 李建文[1] 王马川[1] 

机构地区：[1]陕西科技大学电气与信息工程学院,陕西西安710021

出　　处：《计算机工程与科学》2011年第4期164-167,共4页Computer Engineering & Science

基　　金：国家自然科学基金资助项目(60672001);陕西省教育厅专项基金资助项目(05JC03)

摘　　要：针对提高应用多通道皮肤听声系统进行语音识别的识别率,提出了基于多频带谱减法的语音增强算法。在多通道皮肤听声的实验中,有色噪声会严重降低语音质量,进而降低皮肤听声系统语音识别的识别率,因而首次将基于多带谱减法的语音增强算法引入到皮肤听声系统中以降低有色噪声。多频带谱减法将语音频带划分为多个子频带,分别在每个子频带作不同系数的谱减运算实现语音增强。通过Matlab完成了算法仿真并通过DSP硬件实现了算法并将增强后的语音信号输出给皮肤听声系统,实验证明此设计能够有效抑制有色噪声,增强皮肤听声系统的可靠性和实用性。As for enhancing the speech recognition rate of the skin-hearing system,we propose a multi-band spectral subtraction based speech enhancement algorithm.Coloured-noise can depress the speech quality sharply in the experimental test of the skin-hearing system,we embed the multi-band spectral subtraction based speech enhancement algorithm into the skin-hearing system.The multi-band spectral subtraction based speech enhancement algorithm divides the whole frequency bands to multi bands that each band can be processed by its features,and we emulate the algorithm by using Matlab and realize it by DSP hardware,which outputs the processed speech to the skin-hearing system.The experimental results prove that it can enhance the reliability and practicality of the skin-hearing system.

关 键 词：皮肤听声 谱相减 语音增强 多通道 DSP 

分 类 号：TN912[电子电信—通信与信息系统]