ECG Abnormality Classification and Analysis with SVM Classifier

Electrocardiogram (ECG) is a primary noninvasive tool for monitoring cardiac conditions. Computer-assisted tools are the cutting edge technology to identify minute changes in ECG waveform. This paper presents an ECG classification model to detect seven types of arrhythmia using the MIT-BIH arrhythmia database. The proposed method incorporates time-domain statistical feature and wavelet-based feature extraction separately for classification using a support vector machine. The classifier considers various combinations of ECG time-domain and wavelet features to obtain higher accuracy of classification. Achieved classification accuracy is 97.7 percent within a short computation time. The proposed classifier model has been validated using Monte Carlo Simulation. The variance reduction in accuracy is achieved with the Monte Carlo technique for various ECG input features.