ECG Signals Classification using Statistical and Wavelet Features

Abstract

The technique called signal processing is vitally used as a choice in the case of real-time examination of ECG (electrocardiography) signals. Classification of arrhythmic beat is primarily utilized in the detection of abnormalities normally found in electrocardiogram (ECG) and thus help in identifying problem occurred in heart. This work intentionally carryout, signal preprocessing of electroencephalography, feature extraction using statistical and wavelet and SVM-RFE established classification for arrhythmic is achieved to differentiate regular and irregular constituents of ECG. FFT technique, is used initially in order to remove the noise, to detect R-peaks later then the thresholding technique and windowing technique are consumed. The features of wavelet (such as the information of RR-interval) are calculated and sequenced to features of statistical to organize the final set of feature, which is then consumed to characterize and classify with SVM-RFE for ECG signals. The anticipated classification method arrhythmia is pragmatic to inputting electrocardiography signals gained from the database of MIT-BIH Arrhythmia, along with several international databases of ECG signal. There is no cross validation is required since SVM-RFE works healthy and thus it gives an outstanding performance, it will be beneficial for long-term electrocardiography beat analysis and classification. Results recommend that though the duration of the findings of the recording are of short, with the anticipated model which is able to analyze and classify the idiosyncrasies of heart from an ECG measuring unit with a single lead. The results obtained from the experiment shows an overall accuracy of high performance, sensitivity, and98.97% specificity in contrast with the current approaches mentioned in the literature.