计算机技术与发展2018，Vol.28Issue(4)：119-123,5.DOI:10.3969/j.issn.1673-629X.2018.04.025
一种新的基于DTW的孤立词语音识别算法
A Novel Isolated Word Algorithm of Speech Recognition Based on DTW
摘要
Abstract
In order to improve the recognition efficiency of dynamic time warping(DTW) algorithm in traditional isolated word recogni-tion system,we propose a new isolated word speech recognition algorithm based on DTW.It has made improvements on the contents of speech recognition phase.Firstly,the input speech signal is preprocessed,including pre-emphasis,frame-by-frame and endpoint detec-tion,and then the feature vector of preprocessed speech signal is extracted.The initial parts of speech feature vector are intercepted and re-spectively matched with all model vectors properly,then only several model vectors with less distortion can be left to continue next opti-mal path matching,repeated the optimal path matching and model vectors retaining until it is unique.The experiments show that compared with the ordinary DTW,the proposed algorithm can dramatically reduce the computational cost and improve the efficiency for the isolated word recognition system on the premise of ensuring its recognition accuracy.关键词
语音识别/动态时间规整/部分匹配/孤立词Key words
speech recognition/DTW/partial match/isolated word分类
信息技术与安全科学引用本文复制引用
周炳良,邓立新,洪民江..一种新的基于DTW的孤立词语音识别算法[J].计算机技术与发展,2018,28(4):119-123,5.基金项目
国家自然科学基金(61401227) （61401227）
