FEATURE EXTRACTION OF ECG USING PATTERN RECOGNITION APPROACH

Abstract

Electrocardiogram is a grpahic recording or display of the voltages produced by the electrical activity of the heart. It is extenively used by the cardiologists for diagnosing various cardiac disorders. An ECG signal consist of P,Q,R,S and T wave and interwave segments, Which form the pattern vector of the ECG. The amplitude and duration of these peaks and the intervals of interwave segments are the important features of the ECG. A pattern recognition approach has been used in the present work to extract the features of ECG signal. The pattern recognition approach utilized is based on the assumption that ECG waveform are composite entities which can be decomposed into other simpler entities, then into other simpler once, untill diagnostically improtant peak patterns and segment patterns are obtained. The peak patterns are considered as primitive patterns. The recognition is acheived by first recognising the primitive pattern and then higher patterns. As a first step, the raw ECG signal is preprocessed to remove high frequency noise by using FIR low pass filtering technique. After this set of all peaks are recognised and noisy peaks are eliminated. After detection of peak boundarues incremental energy critria is used to identify QRS complex. After recognition of QRS complex, the P wave is recognised within the left search interval and T wave is recognised within right search interval by again appling the energy critera. The developed software is tested on standard CSE-ECG database. The onsets and offsets of P wave, QRS complex and T waves and their instants of occurrance are given in the results. The amplitude and durations of these peak patterns are also presneted.