Integration of Fcm, Pca and neural networks for classification of Ecg arrhythmias

Abstract

Abstract — The classification of the electrocardiogram (ECG) into different patho-physiological disease categories is a complex pattern recognition task. In this paper, we propose a scheme to integrate fuzzy c-means (FCM) clustering, Principal Component Analysis (PCA) and Neural networks (NN) for ECG beat classification. The PCA is used to decompose ECG signals into weighted sum of basic components that are statistically mutual independent. In addition, FCM clustering is among considerable techniques for data reduction. A back propagation neural network (BPNN) is employed as classifier. ECG samples attributing to six different beat types are sampled from the MIT-BIH arrhythmias database for experiments. In this paper comparative study of performance of four structures such as FCM-NN, PCA-NN, FCM-ICA-NN, and FCM-PCA-NN are investigated. The fuzzy self organizing layer performs the pre-classification task and it is among considerable techniques for data reduction. The aim of using fuzzy C-Means (FCM) is to decrease the number of segments by grouping similar segments training data. The features of obtained clustered training patterns are extracted using principal component analysis (PCA).This layer performs elimination of inconsiderable features with PCA. The test results suggest that FCM-PCA-NN structure can perform better and faster than other techniques.