摘要
为实现准确的帕金森病(PD)早期诊断,文中提出基于语音信号的集成学习诊断方法。基于AdaBoost的特征筛选方式被设计来获得最优目标特征子集,其中过多的弱分类器能习得更多目标特征,而计算复杂度和更多不相关特征被习得的风险也随之增加。相反地,较少弱分类器能降低计算复杂度,然而涉及信息丢失问题。为获得最优弱分类器方案,文中基于监督学习获得最优弱分类器配置。最后,为提升所提出方法的泛化性能,基于正则化损失函数的XGBoost被开发来实现最终病情诊断。实验结果显示,所提出方案的精度(97.28%)相比其它先进算法提升了1.93%。
To achieve accurate early diagnosis of Parkinson’s disease(PD),an integrated learning diagnosis method based on speech signals is proposed.A feature selection method based on AdaBoost is designed to obtain the optimal target feature subset,where too many weak classifiers can learn more target features,but the computational complexity and the risk of more irrelevant features being learned also increases.Conversely,fewer weak classifiers can reduce the computational complexity,but involve information loss.To obtain optimal weak classifier scheme,based on supervised learning,the optimal weak classifier configuration is obtained.Finally,to improve the generalization performance of the proposed method,the XGBoost based on the regularization loss function is developed to achieve the final diagnosis.The experimental results show that the accuracy(97.28%)of the proposed approach is improved by 1.93%compared to other advanced algorithms.
作者
谭言丹
赵阳洋
赵光财
TAN Yan-dan;ZHAO Yang-yang;ZHAO Guang-cai(Haixi Institutes,Chinese Academy of Sciences,Fuzhou 350002,China;School of Electrical and Control Engineering,North University of China,Taiyuan 030051,China;School of Optoelectronic and Communication Engineering,Xiamen University of Technology,Xiamen 361024,Fujian Province,China;University of Chinese Academy of Science,Beijing 100049,China)
出处
《信息技术》
2020年第9期124-128,共5页
Information Technology
