基于改进型DTW的语音识别系统的设计与实现  

Design and Implementation of a Speech Recognition System Based on Improved DTW

作　　者：焦晓燕 JIAO Xiaoyan(Chengyang District Radio and Television Center of Qingdao,Qingdao 266109,China)

机构地区：[1]青岛市城阳区广播电视中心,山东青岛266109

出　　处：《电声技术》2024年第4期60-62,76,共4页Audio Engineering

摘　　要：大部分系统使用深度学习技术完成语音识别任务,并取得良好的效果。但是,基于深度学习的语音识别技术对计算机硬件算力的要求较高,同时需要大量的语音样本对语音识别模型进行训练。针对这些问题,基于改进后的动态时间归整(Dynamic Time Warping,DTW)算法设计并实现了一个语音识别系统。Most systems use deep learning techniques to complete speech recognition tasks and have achieved good results.However,deep learning based speech recognition technology requires high computer hardware computing power,and also requires a large number of speech samples to train the speech recognition model.In response to these issues,the article designs and implements a speech recognition system based on an improved Dynamic Time Warping(DTW)algorithm.

关 键 词：语音识别 动态时间归整(DTW) 小样本 

分 类 号：TN912.34[电子电信—通信与信息系统]