DWT Based SVM Multi Classifier Approach for HR Signal Classification

The present paper proposes wavelet based entropy features and Support Vector Machine (SVM) multi classifier for Heart Rate Variability (HRV) signals classification. The Heart Rate (HR) signals are obtained from ECG signals. The HR signal is decomposed into different frequency bands by wavelet decomposition. The entropy is calculated for each wavelet sub band coefficients. The wavelet based entropy features are given to SVM multi classifier for the classification. The SVM multi classifier is implemented using One Againt All(OAA) principle. In this paper 8 different classes of HR signals are considered for the classification. The maximum classification accuracy obtained by proposed method is 99.64%, whereas the neural multi classifier is 68.12%.