摘要
针对单一特征对嗓音疾病分类识别效果不佳和特征组合随机性的问题,文中提出相关互补原则组合非线性特征方法,有效提高了嗓音疾病的分类识别率。应用小波包分解对嗓音疾病信号进行非线性特征提取及主成分分析,对所提取的特征进行分层降维组合,使用SVM分类器对嗓音疾病进行分类识别。实验结果表明,分层降维特征按相关互补原则组合相较于原始特征随机组合在相同的分类器下的准确率提高了6.16%,极大地提高了嗓音疾病的识别率。
To address the problems of poor classification and recognition of voice diseases by single features and randomness of feature combination,this paper proposes a method of combining nonlinear features by relevant complementary principles to effectively improve the classification and recognition rate of voice diseases.Wavelet packet decomposition is applied to extract nonlinear features and principal component analysis of the voice disease signal,the extracted features are combined in a hierarchical manner with dimensionality reduction,and the SVM classifier is used to classify and recognize voice diseases.The experimental results show that the accuracy of the hierarchical downscaled features is 6.16%higher than that of the original random combination of features with the same classifier,which greatly improves the recognition rate of voice diseases.
作者
陈益
武倩文
姜羽菲
何若男
曹辉
CHEN Yi;WU Qianwen;JIANG Yufei;HE Ruonan;CAO Hui(School of Physics and Information Technology,Shaanxi Normal University,Xi’an 710119,China)
出处
《电子设计工程》
2024年第21期18-22,共5页
Electronic Design Engineering
基金
国家自然科学基金项目(11374199,12374440)。
关键词
嗓音疾病
非线性特征
小波包分解
特征组合
voice disease
nonlinear features
wavelet packet decomposition
features combination
