Development of Signal Processing Algorithms for De-noising of ECG Signals in Wearable Systems and Testing for Arrhythmia using CNN Model

Wireless Data Communication is an emerging prospect of technology in the current world scenario. Processing an ECG signal to drive out informative insights needs the signal to be noise free. If the noise persists, it can damage the complete signal and it cannot be used for healthcare processes. De-noising techniques are employed to reduce the noise levels in the signal such that it can be further used for diagnosis. The various types of noise in an ECG signal are: Powerline Interference, Electrode Contact noise, Muscle Contraction noise, Baseline Wander and Instrumentation noise. The paper will give a detailed explanation of the different types of noise and their reduction techniques implemented in MATLAB R2017a which are compared on various parameters. Further, using image dataset of normal and abnormal heartbeat, a CNN model is proposed which classifies the heartbeat from ECG signal as arrhythmia or normal with an accuracy of 92.62%. This algorithm will de-noise the signal as well as check for arrhythmia in a feasible and accurate way.