ECG Signals Classification Model Based on Frequency domain Features Coupled with Least Square Support Vector Machine (LSSVM)

ECG Signals Classification Model Based on Frequency domain
Features Coupled with Least Square Support Vector Machine (LSSVM)

The electrocardiogram (ECG) is used to inspect the electrical activity of the heart through which experts can detect heart disorders. Mainly medical experts manually examine ECG patterns; however, manual inspection of ECG signals takes significant amount of time and effort as well as is prone to errors. As a result, researchers have started to design automatic models for ECG patterns classification. In this paper, we propose a novel ECG signals classification model that utilises frequency characteristics of ECG signals coupled with a least-squares support vector machine (LS-SVM). An Optimization Triple Half Band Filter Bank (OTHFB) is used to decompose ECG signals into 6 bands delta δδ, theta θθ, alpha αα, beta1 β1β1, beta2 β2β2, and gamma γγ. Then nine statistical features named {max, min, mean, mode, std, variance, skewness, rang, median} are extracted from each band and sent to the LS-SVM. The obtained results showed that the extracted features from the bands alpha αα, beta1 β1,β1, beta2 β2β2, gamma γγ gave a high classification accuracy compared bands delta δδ, theta θθ. The results showed that the proposed model achieved a high accuracy compared with the previous studies. An accuracy of 96% was obtained by the proposed model.

Files associated with this item cannot be displayed due to copyright restrictions.