Heart sound signal classification using fast independent component analysis

KOÇYİĞİT Y.

Turkish Journal of Electrical Engineering and Computer Sciences, vol.24, no.4, pp.2949-2960, 2016 (SCI-Expanded, Scopus, TRDizin) identifier

  • Publication Type: Article / Article
  • Volume: 24 Issue: 4
  • Publication Date: 2016
  • Doi Number: 10.3906/elk-1409-123
  • Journal Name: Turkish Journal of Electrical Engineering and Computer Sciences
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus, TR DİZİN (ULAKBİM)
  • Page Numbers: pp.2949-2960
  • Keywords: Heart sound classification, Independent component analysis, Linear discriminant analysis, Naive Bayes, Principal component analysis, Support vector machines, Wavelet transform
  • Manisa Celal Bayar University Affiliated: Yes

Abstract

The analysis of heart sound signals is a basic method for heart examination. It may indicate the presence of heart disorders and provide clinical information in the diagnostic process. In this study, a novel feature dimension reduction method based on independent component analysis (ICA) has been proposed for the classification of fourteen different heart sound types; the method was compared with principal component analysis. The feature vectors are classified by support vector machines, linear discriminant analysis, and naive Bayes (NB) classifiers using 10-fold cross validation. The ICA combined with NB achieves the highest average performance with a sensitivity of 98.53%, specificity of 99.89%, g-means of 99.21%, and accuracy of 99.79%.