High ECG diagnosis rate using novel machine learning techniques with Distributed Arithmetic (DA) based gated recurrent units

Abstract

A classifier for patient adaptability is vital to overcome instances of inter and intra-patient variance using the automatic arrhythmia diagnosis approach is currently a high focus. The design of an automated cardiac arrhythmia diagnosis system to detect and classify the heartbeats is an active research topic. The paper describes an efficient heart monitoring system in remote areas where cardiologists are scarce and therapies are highly-priced. The proposed technique incorporates the pre-processing of the ECG (electrocardiogram) signals, detecting the Quick Reaction Strike (QRS) beats, characteristics extraction from ECG, identifying abnormality in ECG waveforms and categorizing the ECG beats with a profile curve. A mathematical procedure adopted to filter out unwanted sounds that adjust the window length depending on the proximity. The shifting window mean technique eradicates high-frequency noise that includes interference from a power line and noise from electromyography. The Fast Fourier Transform (FFT) eliminates the low-frequency noise baseline drifting and motion artefact. The most distinguishable waveforms noticed in ECG are QRS beats. Hence, the detection of the complexes in QRS is vital for the analysis of an ECG. The suggested work showcases high rate accuracy, minimal sensitivity, and high specificity.