语音信号特征参数估计算法研究

语音信号特征参数估计算法研究(任务书,开题报告,外文翻译,论文11000字)
摘要
语音识别技术广泛应用于信号处理，模式识别，发声机制和听觉机制，人工智能等领域。中国物联网校企联盟将语音识别视为一种“机器听觉系统”。可见，语音识别在现阶段地位的重要性。它有着可观的应用背景，同时也有深远的理论研究价值。作为语音识别的第一步提取特征参数显得尤为重要。
为了提高语音识别率，本文提出了一种新的特征参数提取算法。首先提取短时平均能量、短时平均过零率和MFCC参数，然后该算法将短时平均能量、基音周期融合到MFCC特征参数中，形成新的特征参数。利用DTW算法对特征参数进行匹配，实现语音识别。使用数字0到9作为语音样本进行语音识别，仿真实验表明，新的特征参数可以提高语音识别率。但由于实验语音样本不够，实验结果客观性不足。系统的大部分工作和包装都由MATLAB完成，版本为7.0。
关键词：语音识别；特征参数；DTW；MFCC
 
Abstract
Speech recognition technology is widely used in signal processing, pattern recognition, phonation mechanism and auditory mechanism, artificial intelligence and other fields. The China Internet of Things School-Enterprise Alliance regards speech recognition as a "machine hearing system." It can be seen that the importance of speech recognition at the present stage. It has a considerable application background, but also has profound theoretical research value. Extracting feature parameters as the first step in speech recognition is particularly important.
    In order to improve the speech recognition rate, this paper proposes a new feature parameter extraction algorithm. First, the short-term average energy, short-time average zero-crossing rate and MFCC parameters are extracted. Then the algorithm combines the short-term average energy and pitch period into the MFCC feature parameters to form new feature parameters. The DTW algorithm is used to match the feature parameters to achieve speech recognition. Using the numbers 0 to 9 as speech samples for speech recognition, simulation experiments show that the new feature parameters can improve the speech recognition rate. However, due to the lack of experimental speech samples, the experimental results are insufficient. Most of the work and packaging of the system is done by MATLAB, version 7.0..
Keywords :Speech recognition; Characteristic parameters; DTW; MFCC
 

目录
摘要    I
Abstract    II
第1章绪论    1
1.1实验开发环境介绍    1
1.2语音识别的发展过程及研究进展    1
1.3语音识别的原理及组成    2
1.4语音特征参数提取在语音识别中的重要性    3
第2章语音信号的预处理    5
2.1预加重    5
2.2分帧与加窗    6
第3章特征参数提取    8
3.1 短时平均能量    8
3.2 短时平均过零率    10
3.3基音周期    11
3.4 Mel频率倒谱系数MFCC    13
3.4.1 Mel频率分析    13
3.4.2 Mel频率倒谱系数提取    14
第4章 DTW算法    16
4.1 DTW算法原理    16
4.2 DTW算法过程    17
第5章改进的特征参数    18
5.1改进的特征参数计算流程    18
5.2仿真实验及其分析    18
第6章结论    20
参考文献    21
致谢    22