Feature Extraction of ECG Signal by using Deep Feature

The analysis and classification of Electrocardiogram (ECG) signals have become very important tool to diagnose of heart disorders. Computer-aided techniques are generally used to classify biomedical application areas. In this paper, we aim to feature extraction and classification of ECG signals. Accordingly, an open access ECG database in Physionet was employed in order to separate normal and abnormal of ECG records. Deep feature approach which is based on Convolutional Neural Network (CNN) was applied to taking out important features of heart recordings. Afterward, Extreme Learning Machine (ELM) was applied to the ECG records. The average precision value metric was used to the performance of the classification performed. In this content, it was noticed classification success values were achieved to accuracy % 88.33, sensitivity %89.47 and specificity % 87.80 with ELM.