| 基于改进型DTW算法和MFCC的语音识别 |
| |
| 引用本文： | 陈孟元. 基于改进型DTW算法和MFCC的语音识别[J]. 安徽工程科技学院学报：自然科学版, 2014, 0(1): 53-57 |
| |
| 作者姓名： | 陈孟元 |
| |
| 作者单位： | 安徽工程大学安徽省电气传动与控制重点实验室; |
| |
| 基金项目： | 安徽高校省级自然科学研究重点基金资助项目(KJ2013A041);芜湖市科技计划基金资助项目(芜科计字[2012]95号) |
| |
| 摘 要： | 采用模式匹配的识别技术，建立孤立词语音识别系统，基于MATLAB环境对O～9这10个数字语音进行仿真实验．在提取MFCC的基础上，整合差分倒谱参数作为语音的特征参数，并对现有的DTW算法加以改进，节省了系统匹配的计算时间，使其具有一定的鲁棒性．分别采集普通话语音和湖北、闽南、安徽3地方言的语音数据，体现了数据的完备性和系统的适用性．实验结果表明，基于改进型DTW算法和MFCC的语音识别系统具有较高识别率，取得了良好效果．
|
| 关 键 词： | 语音识别 改进型DTW 差分倒谱参数软件 MATLAB |
The speech recognition system based on improved DTW algorithm and MFCC |
| |
| CHEN Meng-yuan. The speech recognition system based on improved DTW algorithm and MFCC[J]. Journal of Anhui University of Technology and Science, 2014, 0(1): 53-57 |
| |
| Authors: | CHEN Meng-yuan |
| |
| Affiliation: | CHEN Meng-yuan (Anhui Key Laboratory of Electric Drive and Control, Anhui Polytechnic University, Wuhu 241000,China) |
| |
| Abstract: | A speech recognition system of isolated word is established, adopting the recognition technique of template matching, and applied to recognize digital speech ＇0＂ to ＂9＂ based on MATLAB. During the course of extracting MFCC, taking the derivative coefficient of cepstrum into consideration and impro- ving the existing DTW algorithm,makes the system have a shorter operation time and a certain robust- ness. Speech data containing mandarin pronunciation and dialects from Hubei, Minnan and Anhui are gathered, which embodies the completeness of database and the applicability of system. The experimen- tal results indicate that, the speech recognition system based on improved DTW algorithm and MFCC has a high recognition rate and achieves a good effect. |
| |
| Keywords: | speech recognition improved DTW algorithm derivative coefficient of cepstrum MATLAB |
| 本文献已被 CNKI 维普 等数据库收录！ |
