Абстрактный

Automatic Classification of ECG Signals with Features Extracted Using Wavelet Transform and Support Vector Machines

Sambhu D., Umesh A. C.

Electrocardiogram (ECG) is one of the most widely used techniques for diagnosing cardio vascular diseases. Automatic beat segmentation and classification of ECG signal is paramount since scrutinizing each and every beat is a tedious job for even the most experienced cardiologist. In this paper, we have accurately classified and differentiated Normal and abnormal heartbeats such as left bundle branch block (LBBB), right bundle branch block (RBBB), atrial premature contractions (APC) and premature ventricular contractions (PVC), atrial premature beat (APB), Paced beats and Fusion beats with adequate levels of accuracy. At first the multi resolution analysis of ecg signal is done to denoised and extract 25 features. The mother wavelet used for decomposition was db4. The classification is implemented by using OAO (One Against One) SVM (Support Vector Machine). 7 SVM’s were trained and final grouping is done by maximum voting. ECG signals are obtained from the open source MIT-BIH cardiac arrythmia database. Experiments reveal that the overall classification accuracy is well above 97 % for all the classes.