Heartbeat Classification using discrete wavelet transform and kernel principal component analysis

In this paper, an automatic heartbeat Classification method based on discrete wavelet transform (DWT) and kernel principal component analysis (KPCA) is proposed. DWT is employed to extract time-frequency characteristics of heartbeats, and KPCA is utilized to extract a more complete nonlinear representation of the principal components. In addition, RR interval features are also adopted. A three-layer multilayer perceptron neural network (MLPNN) is used as a classifier. The MIT-BIH Arrhythmia Database was used as a test bench. In the “class-oriented” evaluation, the classification accuracy is 98.48%, which is comparable to previous works. In the “subject-oriented” evaluation, the classification accuracy is 92.34%. The Se (sensitivity) of class “S” and “V” is 62.0% and 84.4% respectively, and the P+ (positive predictive rate) of class “S” and “V” is 70.6% and 77.7% respectively. The results show an improvement on previous works. The proposed method suggested a better performance than the state-of-art method in real situation.