基于耳蜗基底膜特性的语音特征提取方法及在智能轮椅上的应用  

Speech feature extraction method based on cochlear basilar membrane properties and its application on intelligent wheelchair

作　　者：罗元[1] 陈君[1] 张毅[1] 童开国[1] 

机构地区：[1]重庆邮电大学智能系统及机器人研究所,重庆400065

出　　处：《计算机应用研究》2012年第10期3765-3768,共4页Application Research of Computers

基　　金：国家科技部国际合作项目(2010DF12160);国家自然科学基金资助项目(51075420);重庆市攻关项目(CSTC;2010AA2055)

摘　　要：从人耳听觉特性出发,对能模拟耳蜗基底膜滤波特性的伽马通滤波器组进行了研究、修正,并以修正的滤波器组为基础,提出了一种基于耳蜗基底膜特性的GT-ERBCC(equivalent rectangular bandwidth cepstrum coeffi-cient based on GammaTone filters)语音特征提取方法。该方法能准确地表征出语音信号的特征,降低语音识别系统的难度,并将该方法应用于智能轮椅人机交互实验中。结果表明,基于耳蜗基底膜特性的特征提取方法能有效提高语音识别系统的识别率。This paper studied and modified the GammaTone filters,which could simulate the filter characters of cochlear basilar membrane based on auditory system.It proposed a GT-ERBCC（equivalent rectangular bandwidth cepstrum coefficient based on GammaTone filters） feature extraction method based on the novel GammaTone-4 filters.Such method can precisely show the characters of speech signals,reduce the difficulty of speech recognition system.Furthermore,this method was applied on human-computer interaction of the intelligent wheelchair.Experiment results show that it can reflect the characters of speech signals precisely and increase the speed recognition rate effectively.

关 键 词：伽马通 耳蜗基底膜特性 特征提取 智能轮椅 语音识别 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]