Application-Specific Discriminant Analysis of Cardiac Anomalies Using Shift-Invariant Wavelet Transform

Automatic arrhythmia detection in electrocardiogram (ECG) using supervised learning has gained significant considerations in recent years. This paper projects the performance analysis of classifiers such as support vector machine (SVM), extreme learning machine (ELM), and k-nearest neighbor (KNN) with efficient time utilization showing multi-classification for specific medical application. The wavelet double decomposition is used to show the shift-invariant use of dual-tree complex wavelet transform for noise filtering and beat segmentation is done to extract 130 informative samples. Further, the linear discriminant analysis is applied to dimensionally reduce and elite the 12 most relevant features for classifying normal and four abnormal beats collected from MIT/BIH ECG database. The proposed executed system distinguishes SVM, ELM, and KNN with percentage accuracy of 99.8, 97, and 99.8 having classifier testing time as 0.0081s, 0.0031s, and 0.0234s, respectively. The simulated experimental outcomes in comparison with existing work yields adequate accuracy, and computational time.