简单语音识别算法研究

文档编号:JD818  文档字数:8587,页数:21

简单语音识别算法研究
摘要：语音识别的研究目标是让机器“听懂”人类口述的语言。听懂有两种含义，一种是将这种口述语言逐词(字)逐句地转换为相应地书面语言(即文字)，第二种则是对口述语言中所包含的要求或询问做出正确的响应。本次设计是说话人的确认。语音识别，主要是选择说话人的特征参数和一个合适的算法。特征提取部分就是寻找能够有效代表一个人的特征参数。本文设计是基于MATLAB软件实现的，mfcc作为特征参数，采用VQ算法编程实现。结果是在MATLAB下用文本的形式显示出来。

关键词：语音识别；特征；Mel频率倒谱系数

Simple Voice Recognition Design

Abstract: The goal of speech recognition research is to make machines “understand” the language of human oral. There are two means : one is that the machine can turn the spoken langue to written langue sentence by sentence; the other is that it can response to langue requirement correctly. This design is the recognition of the speaker. Voice Recognition ,mainly is to find characteristic parameters and a proper algorithm . The purpose of feature selection is to find the characteristic parameters representing a speaker properly.This design is implemented based on MATLAB.mfcc is characteristic parameters .It use VQ algorithm .The result is displayed in text under MATLAB.

Key words: Voice Recognition；feature；mfcc

目录
摘要 I
Abstract I
1 引言 1
2 识别过程概述 2
2.1 预处理 3
2.2 语音采样 3
2.3 预加重 3
2.4 加窗 3
2.5 短时能量 4
2.6 短时平均过零率 5
2.7 短时自相关函数 6
2.8 语音信号的端点检测 6
3 特征参数提取技术 7
3.1 特征参数提取 7
3.2 MFCC系数 7
4 系统设计 9
5 结果分析 11
6 总结 12
6.1 回顾 12
6.2 展望 12
参考文献 13
谢 辞 14
附录 15

相关论文