计算机应用研究2017，Vol.34Issue(11)：3223-3228,6.DOI:10.3969/j.issn.1001-3695.2017.11.005
基于一种小波核优化学习的KSPP子空间故障特征提取
Method for feature extraction in KSPP feature subspace based on wavelet kernel learning
摘要
Abstract
In the fault diagnosis of electronic system,it is difficult to extract effectively fault features.As a result,this paper presented a new feature extraction method based on self-optimization wavelet kernel sparsity preserving projection.At first,the kernel polarization criterion was extended to an improved form so that it could simultaneously encode the multiclass information and preserve the local structure of within-class data.For Mexico-hat wavelet kernel function,this paper established a new objective function based on improved kernel evaluation measurement criterion.Then it obtained the optimal kernel parameter by minimizing objective function based on particle swarm optimization algorithm.Finally,it extracted effective features from kernel feature subspace by inserting optimized wavelet kernel function into kernel sparsity preserving projection.Compared with several well-known feature extraction methods,experimental results show that the proposed method can obtain higher classification accuracy and better generalization performance.关键词
核极化/核属性约简/小波核/核稀疏保持投影/故障识别Key words
kernel polarization/kernel attribute reduction/wavelet kernel/kernel sparsity preserving projection (KSPP)/fault identification分类
信息技术与安全科学引用本文复制引用
张伟,许爱强,高明哲..基于一种小波核优化学习的KSPP子空间故障特征提取[J].计算机应用研究,2017,34(11):3223-3228,6.基金项目
国家自然科学基金资助项目(61571454) （61571454）
