Wavelet and KICA based ECG Beat Classification for Cardiac Health Care

Analysis of disturbances in Electrocardiogram (ECG) signal is a challenging task for diagnosis of cardiac diseases accurately. ECG being a non stationary signal, its amplitude and time duration keeps on changing. Computer designed techniques or algorithms are effective diagnostic tools for cardiac health management. In this proposed method, five ECG beats from 'MIT/BIH' arrhythmia database have been taken The approach adopted is to decompose ECG signal using Discrete Wavelet Transform (DWT) and then to apply Kernel Independent Component Analysis (KICA) for reducing the vector dimensions and non linear feature extraction to get desired dataset. Three classifiers K nearest neighbor (KNN), feed forward neural network (FNN) and back propagation neural network (BPNN) are independently deployed to achieve classification performance. The highest accuracy of 99.8% is achieved by using KICA with KNN.