基于MATLAB的孤立次识别.doc

摘 要 语音识别(Speech Recognition)是让机器通过识别和理解过程把语音信号转变为相应的文本或命令的技术。语音识别技术以语音信号处理为研究对象，涉及语言学、计算机科学、信号处理、生理学、心理学等诸多领域，是模式识别的重要分支。该技术有非常广阔的应用前景。 论文主要研究语音识别的基本原理和不同的识别算法，并着重研究基于MATLAB的语音识别算法。在此基础上论文构建了一个基于MATLAB的孤立词语音识别系统。该系统完成了语音识别的主要过程，包含预处理、端点检测、特征参数提取和模式匹配。预处理主要进行了预加重和加窗分帧，端点检测采用了双门限方法，特征参数提取采用的是线性预测分析方法，选取了美尔倒谱系数(MFCC)作为主要的特征参数，利用DTW算法完成了几个孤立词“前进、后退、停”的训练和识别。多人实验的结果表明，该系统的识别率还算可观，对简单孤立词语音识别非常实用，可以满足系统的设计要求。 关键词:matlab；语音识别；孤立词 Abstract Speech Recognition (Speech Recognition) is to allow machines to identify and understand the process of the voice signal into the appropriate text or order of the technology. Voice recognition technology to voice signal processing for the study, involving linguistics, computer science, signal processing, physiology, psychology and other fields, is an important branch of pattern recognition. The technology has a very broad prospects.  Major research papers speech recognition and the basic principles of different algorithms to identify and focus on MATLAB-based speech recognition algorithm. On this basis, build a thesis based on MATLAB isolated word speech recognition system. The system completed the process of the main speech recognition, including pretreatment, the endpoint detection, feature extraction and pattern matching. Pretreatment a major pre-emphasis and processing window-frames, endpoint detection using a dual-threshold method, feature extraction using a linear prediction analysis, the United States and Seoul has been selected inverted coefficients (MFCC) as the main parameters, the use of DTW algorithm completed a few isolated words forward, rewind, stop the training and recognition. More than the results of the experiment shows that the system of recognition rate still considerable, the simple words in isolation speech recognition is useful, to meet the design requirements. Key words: matlab; speech recognition; isolated words 目 录 1 绪 论 1 1.1 语音识别概述 1 1.1.1语音识别研究的重要意义 1 1.1.2 语音识别的发展现状 2 1.2 语音识别研究的现状及发展趋势 3 1.2.1语音识