Recognition of Electrocardiogram Signal using Multi-class Kernel Support Vector Machine

In this proposed work, an electrocardiogram (ECG) signal is classified using multi-class kernel function based support vector machine (SVM) approach with optmized hyperparameter using bayesian optmization (BO). Measured ECG signal is denoised using butterworth filter. Then the signal is converted into scalogram patterns and features are extracted using fractal image image compression technique for recognition. For validation, the propsoed algorithm comprised of three different input patterns like, people having cardiac arrhythmia (ARR), people having congestive heart failure (CHF), and people having normal sinus rhythms (NSR). The proposed algorithm attained the recognition rate of 99.25% compared to other machine learning approaches signifcant improvement is found. Additional beneifit of the proposed algorithm is required very less amount of memory with fast response.