基于MATLAB的英语口语语音识别技术的研究信号与信息处理专业论文.docxVIP

Based on the MATLAB spoken English speech recognition technology research Abstract With the development of computer technology, It has become a new trend to develop a software with voice recognition.by combining speech recognition technology and multimedia to assist the spoken English teaching and learning. English language teaching softwares lack an effective feedback on the result of spoken English learning whereas,effective evaluation is the key point.Based on the features,using DTW algorithm with MATLAB7.0 as the development software,we provided a spoken language practice scoring system to help learners improve their pronunciation timely. This paper introduces the principles and compositions of the speech recognition system, the key part of the speech signal pre-processing, endpoint detection, MFCC cepstrum calculation, the theoretical analysis, and by using the DTW algorithm established a standard template in the template library and compared template librarytest voice, the voice matches the given fraction.The software has a voice signal endpoint detection, feature extraction, speech recognition and other functions,is easy to operate a good interactive interface. Test results through data analysis confirmed the feasibility of the program, provides a theoretical basis for research and development of voice recognition technology in the hardware or software. Keywords:speech recognition,MFCC,the DTW algorithm,score 中北大学学位论文 中北大学学位论文 I I 目 录 1 绪论 1.1 课题的提出及意义 1 1.2 语音识别技术的发展 2 1.3 技术难点 3 1.4 研究的主要目的及内容 4 2 语音识别系统的原理及其组成 2.1 语音识别系统的原理 5 2.1.1 英语语音发音特点 5 2.1.2 语音识别的分类 6 2.1.3 语音识别的基本方法 7 2.2 语音识别系统的组成 13 2.2.1 语音信号预处理 13 2.2.2 加窗分帧 15 2.2.3 端点检测 17 2.2.4 语音特征提取 21 3 DTW 的识别算法 3.1 DTW 算法的匹配原理 28 3.2 DTW 模板训练和识别 31 3.3 一种改进的高效 DTW 算法 32 4 语音识别系统的设计 4.1 设计思想 36 4.2 结构设计 36 4.3 运行环境和开发软件 37 4.3.1 运行环境 37 4.3.2 开发软件 38 II II 4.4 系统实现过程 39 4.4.1 数据采集 39 4.4.2 语音信号的预处理 41 4.4.3 语音信号的特征及其端点检测 43 4.4.4 语音特征信号的提取 49 4.4.5 模板训练与识别 52 4.4.6 语音识别