KNN and PCA classifier with Autoregressive modelling during different ECG signal interpretation

Electrocardiogram (ECG) conveys the information of the heart. Autoregressive (AR) coefficients extract features of the ECG signal. Three different conditions of subjects, i.e. Atrial Tachycardia, Premature Atrial Contractions and Sinus Arrhythmia were examined. For classification K-Nearest Neighbor (KNN) classifier and Principal Component analysis (PCA) classifier has carried out. Autoregressive modelling is done with an ECG signal with baseline wander noise and removed baseline wander noise signal. Among KNN-classifier and PCA classifier with autoregressive modeling methods those are Yule-Walker (YW) and Burg method; KNN classifier with Burg method shows good result at model order 9 and PCA classifier with YW method shows good result at model order 8. Signal to Noise ratio (SNR) is also calculated and checked SNRs were found to be 9.81dB, 11.23dB and 8.35dB for subjects 1, 2 and 3 respectively.