• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Conference on Image, Vision and Computing >Voice Pathology Detection Based on the Vocal Fold Signal and the Vocal Tract Signal Separation
【24h】

Voice Pathology Detection Based on the Vocal Fold Signal and the Vocal Tract Signal Separation

机译：基于人声折叠信号和人声信号分离的语音病理检测

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Voice pathology correlates with vocal fold problems, so extracting valid features from the vocal fold excitation signal is helpful for classifying the normal and pathological voice. A novel feature extraction method which combines wavelet packet decomposition and nonlinear feature extraction is proposed in this paper. The original speech signals are firstly decomposed into 5 layers using wavelet packet-based method, and the high frequency signals which correlate with the vocal fold are reconstructed. Then nonlinear features are extracted from the reconstructed signals. Support Vector Machine is used to classify the normal and pathological voice using the nonlinear features. The proposed method and features are evaluated on the Massachusetts Eye and Ear Infirmary databases. The second-order renyi entropy features give very promising classification accuracy of 98.21%. The highest accuracy is 99.21 % when the Hurst parameter and second-order renyi entropy features are combined. Experimental results show that the vocal fold excitation signal can express the pathological information about sound efficiently, which can be used for the automatic detection and classification of the pathological voice.
机译：语音病理学与人声折叠问题相关，因此从人声折叠激励信号中提取有效特征有助于对正常和病理性语音进行分类。提出了一种结合小波包分解和非线性特征提取的特征提取方法。首先使用基于小波包的方法将原始语音信号分解为5层，并重建与人声折合相关的高频信号。然后从重构信号中提取非线性特征。支持向量机用于使用非线性特征对正常语音和病理语音进行分类。所提出的方法和功能在马萨诸塞州眼耳医院数据库中进行了评估。二阶renyi熵特征给出了非常有希望的分类精度，为98.21％。当Hurst参数和二阶renyi熵特征组合在一起时，最高的准确度是99.21％。实验结果表明，声带激励信号可以有效地表达声音的病理信息，可用于病理语音的自动检测和分类。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号