Selective Feature Coding for Cardiac Arrhythmia Detection through ECG Signal Analysis

Abstract

Detection of abnormalities in the ECG signal to achieve an automatic diagnosis of several heart related diseases has become an increased research aspect. This paper focused to develop an automatic detection system to detect abnormalities in ECG. These abnormalities results in different cardiac arrhythmias. Towards the detection of different cardiac arrhythmias, this paper analyzed the ECG signal through Dual Tree Complex Wavelet Transform (DTCWT) as a feature extraction technique and further proposed a new selective band coding technique to extract only the informative features from the sub bands obtained from DTCWT. The novelty of this proposed system is to remove the redundant information, thereby achieving a fast and accurate detection results. Multi-Class Support Vector Machine (MC-SVM) is used for classification purpose. Extensive simulations are carried out for the MIT-BIH database and the performance is measured through the performance metrics such as Accuracy, Precision, Recall, False Positive Rate, F-Measure and overall computational time. The proposed method is also compared with conventional approaches to alleviate the performance enhancement in the detection of Cardiac Arrhythmias (CAs) with less time span.