Study of ECG feature extraction for automatic classification based on wavelet transform

Electrocardiogram (ECG) feature extraction plays an important role in automatic classification and diagnosis. The current study focuses on the feature extraction of premature ventricular contraction (PVC) and normal sinus rhythm (NSR) for the discrimination purpose between them. The data in the analysis were collected from MIT-BIH database. A beat detection algorithm that was not affected by beat shape was introduced in the study. The ECG features were extracted based on wavelet transform for the analysis. Two feature sets were formed by selected wavelet coefficients and statistic parameters of wavelet coefficients for the comparative study. Support Vector Machine (SVM) algorithm was utilized to classify the ECG beats. The experimental results show that it is possible and feasible to extract ECG features with lower dimensions from wavelet coefficients in order to improve the classification results.