Fuzzy c-means in features space of Teager-Kaiser Energy of continuous wavelet coefficients for detection of PVC beats in ECG

In this paper, a method for the detection of premature ventricular contraction (PVC) beat in electrocardiogram (ECG) signal is presented. The method adopts the minimum distance (MD) and the linear discriminant analysis (LDA) in a proposed feature space for classification. The features are extracted by continuous wavelet-transform of the ECG beat signal followed by the Teager-Kiaser Energy (TKE) operator. The significant TKEs are then selected, transformed to [0, 1] range, and used as a feature vector. The prototypes of different classes in the feature space are generated using the fuzzy c-means (FCM) clustering algorithm. To test the proposed method, the MIT-BIH arrhythmia database has been used. The method has provided total accuracy of 100% in classifying normal and PVC beats with MD classifier. The proposed method has also outperformed a counterpart one employing features extracted by the discrete cosine transform (DCT).