基于DSP的语音识别系统研究与实现  

Realization of speech recognition system based on DSP

作　　者：孙爱中[1] 刘冰[1] 张琬珍[1] 栾前进 

机构地区：[1]中国航空计算技术研究所,陕西西安710068 [2]陆航驻西安地区军代室,陕西西安710068

出　　处：《现代电子技术》2013年第9期76-78,共3页Modern Electronics Technique

摘　　要：语音识别是人机语音通信的关键技术之一,也是难题之一。介绍了一种语音识别系统,主要介绍了该系统的语音处理流程,阐述了系统使用Mel频标倒谱参数作为特征提取的方法,采用隐马尔科夫模型算法的测度估计技术。通过严格测试,该系统达到实用化要求。该语音识别系统较好的实现了在移动电子设备上资源有限条件下方便快捷的汉字语音输入,具有重大现实意义。Speech recognition is one of the key technologies in human-machine speech communication. A speech recognition system and its voise processing program flow are introduced. The way that Mel frequency cepstrum parameters are used to extract features, and the measurement estimation algorithm based on hidden markov model （HMM） are elaborated. The strict testing results show that the system can meet the requirement of practicability. The speech recognition system realized Chinese characters voice input under the condition of resource limited in eleetrommunication equipments. This system has great practical significance.

关 键 词：语音识别 特征提取 带通滤波 隐马尔科夫模型 

分 类 号：TN911.73[电子电信—通信与信息系统]