基于改进DTW的彝语孤立词识别研究  

Yi Isolated-Word Speech Recognition Based on Improved DTW

作　　者：余炜[1,2] 周娅[1] 万代立 杨喜敬 

机构地区：[1]昆明理工大学信息工程与自动化学院,云南昆明650500 [2]成都工业学院,四川成都611730

出　　处：《昆明理工大学学报（自然科学版）》2014年第5期47-53,共7页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：四川省教育厅重点项目(2013SZA0153)

摘　　要：彝语的语音识别与处理是语音信号处理领域的一个新方向.本文在深入分析彝语特点的基础上,对用于汉语孤立词识别的经典端点检测算法进行了改进.探讨了动态时间规整算法(Dynamic Time Warping,DTW)和高效动态时间规整算法(Efficient DTW,EDTW)在彝语孤立词识别中的应用,并提出了两种新算法:基于音节个数的动态时间规整算法(Syllable Number Based Improved DTW,SDTW)和基于音节个数的高效动态时间规整算法(Syllable Number Based Improved EDTW,SEDTW),对特定的彝语语音信号进行识别.实验结果表明,所提出的两种新算法分别在信号识别率和识别时间上比经典算法优越.The recognition and processing of Yi language is a new direction of speech signal processing field. Based on deep analysis of the characteristics of Yi, the classic endpoint detection algorithm for Chinese isolated -word are improved. The applications of dynamic time warping （DTW）algorithm and efficient dynamic time warping（EDTW） algorithm for Yi isolated -word are discussed, and two novel algorithms for this purpose are proposed, which are Syllable Number Based Improved DTW and Syllable Number Based Improved EDTW. Ex- perimental results show that the proposed algorithms have better performance on the recognition rate and time compared with classic methods.

关 键 词：语音识别 孤立词识别 彝语 MEL频率倒谱系数 DTW 

分 类 号：TN912.3[电子电信—通信与信息系统]