计算机与现代化

• 网络与通信 • 上一篇    下一篇

基于PCA串并融合的多路心音特征表征方法

  

  1. (1.南京邮电大学信息化建设与管理办公室,江苏南京210023;
    2.南京邮电大学电子与光学工程学院、微电子学院,江苏南京210023)
  • 收稿日期:2019-10-16 出版日期:2019-12-11 发布日期:2019-12-11
  • 作者简介:成雨含(1984-),女,四川资阳人,工程师,硕士,研究方向:计算机应用,网站设计,计算机动画,E-mail: chengyh@njupt.edu.cn; 张友讯(1993-),男,江苏淮安人,硕士研究生,研究方向:生物信号处理,E-mail: 495690825@qq.com; 孙科学(1981-),男,安徽界首人,副教授,硕士生导师,博士,研究方向:电子电路设计,智能信号处理,嵌入式系统与通信软件设计,E-mail: sunkx@njupt.edu.cn。
  • 基金资助:
    国家自然科学基金资助项目(61271334); 上海申康三年行动计划项目(16CR3079B)

Multi-channel Heart Sound Feature Characterization Method #br# Based on PCA Serial-merged Fusion

  1. (1. Informatization Construction and Management Office, Nanjing University of Posts and Telecommunications, Nanjing 210023, China;
    2. College of Electronic and Optical Engineering & College of Microelectronics, Nanjing University of Posts
    and Telecommunications, Nanjing 210023, China)
  • Received:2019-10-16 Online:2019-12-11 Published:2019-12-11

摘要: 多路心音信号不仅比单路心音信号涵盖更多关于总体的特征,而且能够弥补单路心音数据携带的信息量可能不充分的缺陷。利用笔者自主设计的4路心音传感器,初步建立一个小型4路心音数据库。基于这个数据库,首先阐明多路心音信号的特点,论述心杂音与听诊位置的关系;然后分别提取心音的单路和4路能量熵系数、4路心音互信息作为有效特征数据集,利用PCA对能量熵特征进行降维处理,获得串行特征;将相关性特征和互信息特征从实向量空间拓展到复向量空间,进行并行融合,获得并行特征;最后将串行并行特征再次融合成为多元优化组合特征。这种融合策略,具有针对性强,凸显差异性的优点。仿真实验结果表明,由多路心音信号获取的多元优化组合特征表征效果明显优于单路心音信号的特征表征效果,不仅有益于分类模型的构建,而且对实现先心病的快速筛查,提高分类识别率具有积极的意义。

关键词: 多路心音, 多元特征, 特征表征, 特征融合

Abstract: The multi-channel heart sound signal not only covers more general characteristics than the single-channel heart sound signals, but also can make up for the defect that the information carried by the single-channel heart sound data. Using a four-way heart sound sensor, a small four-way heart sound database was established. Based on it, firstly, the characteristics of multi-channel heart sound signals are clarified, and the relationship between heart murmur and auscultation position is discussed. Then, the single-channel and four-channel energy entropy are extracted as effective feature data by using PCA pair. The energy entropy feature is dimension-reduced to obtain serial features. The correlation features and mutual information features are extended from real vector space to complex vector space. Finally, serial parallel features are re-converged into multi-optimal combinations feature. The simulation results show that the feature representation of multivariate optimal combination obtained by multi-channel heart sound signals is better than that of single-channel heart sound signals, which is not only beneficial to the construction of classification models, but also to quickly screen for congenital heart disease and improve classification. The recognition rate has a positive meaning.

Key words: multi-channel heart sound, multi-feature, feature characterization, feature fusion

中图分类号: