| 朱宗宝,王坤侠,肖玲玲,刘文静.一种基于小波包主成分分析的语音情感识别方法[J].安徽建筑大学学报,2017,25(5):35-39 |
| 一种基于小波包主成分分析的语音情感识别方法 |
| A Method of Speech Emotion Recognition Based on Wavelet Packet-Principal Component Analysis |
| |
| DOI：10.11921/j.issn.2095-8382.20170508 |
| 中文关键词: 特征提取 主成分分析法 小波包变换 支持向量机 |
| 英文关键词: feature extraction Principal Component Analysis Wavelet Packet Transform Support Vector Machine |
| 基金项目:安徽省自然科学基金面上项目（1708085MF167）；中国科学院自动化研究所模式识别国家重点实验室开放课题（201700014） |
|
| 摘要点击次数: 7079 |
| 全文下载次数: 0 |
| 中文摘要: |
| 在语音情感识别中，由于特征参数的提取直接影响到最终的识别效率，从原始语音信号中提取特征参数是非常重要的。但是本文中提取的特征维数太多，导致特征匹配时过于复杂，消耗系统资源，不得不采用特征降维的方法。本文主要是研究一种在小波包变换的基础上通过特征降维来提高语音情感识别效果的方法，为此本文在德国库EMODB的基础上，通过小波包变换提取出语音的情感特征参数，然后利用主成分分析法对特征参数进行降维，最后利用支持向量机进行训练和测试。通过实验，获得了较好的识别效果。 |
| 英文摘要: |
| In speech emotion recognition, the extraction of feature parameters has a direct impact on the final recognition efficiency. It is very important to extract feature parameters from the original speech signal. But in the paper,there is too much extracted feature dimension is too much, which often leads to the complexity of feature matching, and consumes the system resources, so we have to adopt feature dimension reduction method. This paper is to improve the effect of speech emotion recognition based on a transform of wavelet packet by feature dimension reduction. So the paper present a emotional feature extraction method of speech by transforming wavelet based on the German Database (EMODB), and then adopt principal component analysis to reduce the dimensionality of feature parameter,We finally use Support Vector Machine for training and testing. Good recognition results are obtained through experiments. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
