Wavelet Based Data Mining For Discrimination Of Ecg Beat Characteristic Points

Abstract

In the present work, a systematic study is carried out to analyze, investigate and develop a data
mining algorithm for Electro Cardio Gram (ECG) signals. Based on the nature of the human heart beats,
the physiology cycle gets disturbed, as the normal human heart beats signal has its signature PQRST
waveform, whereas the disturbed one will have short R-R intervals, absence of and widen QRS. To
predict and classify the events happening in the heart, Wavelet Transform (WT) has been used. As the
ECG signals are non-stationary signals, which can be handled very well with basic data by various WTs,
as it adapts two-dimensional time scale processing technique. To compare the proposed technique, the
performance comparison is carried out. It results in better correct rate of the ECG signal features of
Normal Sinus Rhythm (NSR), Ventricular Tachycardia (VT) and Ventricular Fibrillation (Vfib).