摘要
共振峰是语音信号的一个重要特征,对提高耳聋患者的语言识别率具有重要意义。然而,目前数字助听器领域常用的响度补偿算法(多通道响度补偿和宽动态压缩)均对共振峰结构具有一定的破坏性,对患者听懂语音十分不利。本文结合共振峰检测,提出一种基于共振峰提取的多通道响度补偿算法,在原有多通道响度补偿的基础上,通过对滤波器组的重新设计并加入共振峰提取模块对共振峰予以保护。仿真结果证明,该算法对4类常见患耳均能达到满意的补偿效果,同时,与多通道响度补偿和宽动态压缩两种方法比较,该算法在保护共振峰结构完整性方面性能更优。
Formant is one of important features for speech signal.It is very significant to enhance the speech intelligibility（SI） for deafness patients.At present,the commonly used loudness compensation algorithms for digital hearing aid,for instance multi-channel loudness compensation（MCLC） and wide dynamic range compression（WDRC）,have strongly negative effects to the formant and the results are poor for patients＇ SI.In this study,a new MCLC algorithm based on formant extraction combining formant detection method is presented here.The algorithm can play protective effect for formant by redesigning of filter bank and joining a formant extraction module.Results show the proposed algorithm can have a suitable compensation effect for 4 common deafness types while achieving more excellent performance than MCLC and WDRC on protecting formant structural integrity.
出处
《信号处理》
CSCD
北大核心
2012年第3期352-360,共9页
Journal of Signal Processing
基金
国家自然科学基金(11074202)
西北工业大学研究生种子创业基金(Z2011100)资助课题
关键词
数字助听器
响度补偿
临界带
共振峰
digital hearing aid
loudness compensation
bark bands
formant
