基于隐马尔可夫模型的语言种类识别方法的研究

论文总字数：18131字

摘 要

语言是人类思想表达方式最为重要的传输媒介，是人们最高效、最便利、最自然的交流放式。语音识别技术就是让计算机接收、识别和理解人类的语言，将语言信号转换成相应数字信号的技术[1]。

语音识别技术是当今语音信号处理的一项研究热点。语音识别是将经训练算法的语音信号转化为可被计算机识别的文本或命令[2]。换言之，让机器去理解识别人类的语言。本此课题主要围绕语言种类识别进行研究，实验采用Matlab工具进行，语言种类识别是指计算机对输入的语音信号进行训练和识别，对其语言做出种类判决。语言种类识别在多种语音语言识别系统中起着重要的角色，它在教育培训、医疗卫生和国防安全等诸多方面发挥着越来越重要的作用。

语言种类识别是语音识别技术的一个分支，本课题采用的基于隐马尔科夫模型（HMM）的语言种类方法的研究，根据语言识别系统原理依次进行语音信号预处理、LPCC特征参数提取、语音训练及识别判决模块分析得到良好的测试结果。本次论文主要分为五大章节详细论述了识别系统工作原理及各个模块功能，深入学习基于隐马尔可夫模型的语言种类识别系统及其原理等。

关键词：HMM模型；语言种类识别；特征参数分析；LPC倒谱参数；语音数据库；

Research on the Method of Language Category Recognition Based on Hidden Markov Model

Abstract

Language is the most important transmission medium for the expression of human thought, and it is the most efficient, convenient, and natural way for people to communicate. Speech recognition technology is a technology that allows computers to receive, recognize and understand human language, and convert language signals into corresponding digital signals

Speech recognition technology is an important research hotspot in speech signal processing today. Speech recognition is to convert speech signal into text or command that can be read by computer. In other words, let computers understand human language. This topic mainly focuses on the research of language type recognition. The experiment is carried out using Matlab tools. Language category recognition refers to the training and recognition of the input speech signal by the computer, and the judgment of the language category. Language type recognition plays an important role in multiple speech language recognition systems, and it plays an increasingly important role in many aspects such as education and training, health care and national defense security.

Language category recognition is a branch of speech recognition. Speech recognition has a broad application prospect. The research on the language category method based on Hidden Markov Model (HMM) adopted in this topic is carried out in order according to the principle of language recognition system: speech signal preprocessing module, LPCC feature parameter extraction module, Voice training and recognition decision module and analysis to get good test results. This paper is mainly divided into five chapters to discuss in detail the block understanding of each module, in-depth study of recognition system based on Hidden Markov model and its principle.

Key Words: HMM Model Language Category Recognition，Characteristic Parameter Analysis，LPCC Parameter，Speech Database

目 录

摘 要 I

Abstract II

第一章 引 言 1

1.1语言种类识别背景 1

1.2语言种类识别的分类 1

1.3语言种类识别发展前景 2

1.3.1语言种类识别的发展 2

1.3.2语言种类识别的研究现状 2

1.4本论文内容安排 2

第二章HMM的基本介绍 4

2.1 HMM基本原理 4

2.1.1 HMM基本概念 4

2.1.2 Markov链 5

2.2 HMM基本算法 5

2.2.1向前-向后算法 5

2.2.2 Viterbi算法 6

2.2.3 Baum-Welch算法 7

第三章 语言种类识别基本原理和方法 9

3.1语音预处理 9

3.2特征参数选取 9

3.3 LPC倒谱参数 10

3.4模式匹配 11

3.5判别方法和阈值的选取 11

第四章 语言识别系统 12

4.1系统实现 12

4.2语音数据库 12

4.2.1语音数据库处理 12

4.2.2语音数据库的安排说明 13

4.3实验结果与讨论 13

4.3.1 TD测试 13

4.3.2 TI测试 14

第五章 总 结 15

5.1实验结论 15

5.2实验不足及改进 15

致 谢 16

参考文献（References） 17

剩余内容已隐藏，请支付后下载全文，论文总字数：18131字