Fusion of multiple time-frequency representation techniques and classifiers for ECG and PPG signal analysis

Time-frequency representation (TFR) techniques play a vital role in signal processing by revealing inherent characteristics of complex data. This study explores the fusion of 11 TFR techniques for analysing electrocardiogram (ECG) and photoplethysmogram (PPG) signals. Signal analysis involves noise removal, characteristic point detection, and feature extraction, followed by classification. Models were trained to classify ECG signals into six arrhythmia classes and PPG signals into normal and abnormal (hypertension) categories. Several transforms achieved over 90% test accuracy, with ensemble classifiers performing exceptionally well, achieving up to 95.87% accuracy for ECG signals. Ensemble classifiers using continuous wavelet transform (CWT), S-transform, Wigner-Ville distribution (WVD), and synchrosqueezed wavelet transform (SSWT) demonstrated a 96.6% testing accuracy. Additionally, combining CWT and SSWT, the K-nearest neighbour classifier achieved 81.48% accuracy for PPG signals. The proposed approach fuses information from multiple TFR techniques, boosting ECG classification accuracy to 99.75% and PPG accuracy to 82.52%.