一种新的基于并行分段裁剪的DTW算法  

New DTW algorithm based on segment cutting in parallel

作　　者：安镇宙[1,2] 杨鉴[1] 王红[1] 余映[1] 

机构地区：[1]云南大学信息学院 [2]玉溪师范学院计算机科学系,云南玉溪653100

出　　处：《计算机工程与应用》2007年第15期35-36,88,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60265001) ;云南大学中青年骨干教师基金。

摘　　要：通过对DTW算法的研究,提出了一种并行分段裁剪的新方法,在减少DTW算法运算量方面有显著效果,并将其用于一个地名识别系统中,经测试,可以明显缩短识别时间,具有很强的实时性,有较高的识别率,适合作为小型语音识别产品的主要算法。Through the research to Dynamic Time Warping （DTW） algorithm,a new algorithm named segment cutting in parallel about reducing DTW algorithm operation quantity is described.It has the remarkable effect in the place name recognition system. After tests,we find it may greatly reduce the recognition time and has strict real time requirement and high recognition accuracy. The arithmetic is suited for designing a small speech recognition production.

关 键 词：语音识别 DTW 模板匹配 并行分段裁剪 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]