A New Approach to Extract Features from ECG Signals

Electrocardiogram (ECG) is the record of electrical and muscular movement of the heart over a period of time which is useful for diagnosing many cardiac diseases. Due to the increase in mortality rate associated with cardiac diseases, automatic detection of these diseases is highly required. Hence, feature extraction and de-noising of ECG signals are highly useful in cardiology. This paper describes a novel approach for feature extraction in order to recognize trusty heart rhythm. This method consists of two components: data preprocessing and feature extraction. At pre-processing stage, baseline wander removal and power line inference have been used for removing baseline wander and de-noising the signal respectively. At the second stage, R wave is detected and then Q, S, T and P waves are extracted based on R wave. In this paper, Wavelet transformation is used for power line inference. The experimental studies show that the proposed method provides 96.79% accuracy and 5 features recognition with reduced computational complexity.