改进的DTW算法在实时语音辨识系统中的应用  

Application of Improved DTW Algorithm in Real-time Voice Recognition

作　　者：曹茂俊[1] 尚福华[1] 

机构地区：[1]大庆石油学院计算机与信息技术学院,大庆163318

出　　处：《科学技术与工程》2010年第7期1652-1655,共4页Science Technology and Engineering

基　　金：中国博士后科学基金(20080440923);黑龙江省自然科学基金(F2007-11);黑龙江省教育厅科研课题(11521005)资助

摘　　要：改进了传统的动态时间弯曲算法结构,将其应用到实时语音辨识系统中,并在计算机上进行了仿真。实验结果表明采用改进后的DTW算法有效地降低了待处理数据量和识别时间,提高了系统的运行速度。随着待识别语音信号数量的增多,该算法优势更加明显。The traditional DTW algorithm＇s structure is improred and applied, and the algorithm in real-time voice recognition. The results from the simulation show that the improved DTW algorithm can reduce the amount of data to be processed and recognizing time effectively and provide a better performance. The advantage of the improved algorithm will be more obvious with the increasing number of the voice signal to be recognized.

关 键 词：动态时间弯曲 端点检测 美尔倒谱系数 语音辨识 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]