Electrocardiogram Analysis for Robust Optimal Thresholding System Design

Abstract

The Electrocardiogram (ECG) is a most prominent and economical technique that has been used as the standard approach for diagnosis of cardiovascular disease through the investigation of the heart rate and the morphological analysis. Presence of different noises in ECG signals designated as Baseline wander, Power-line interference, Burst noise and Electromyography degrades the perceptual quality and performance that results into false interpretation of ECG by clinicians. The denoised ECG signals and accurate event detection of signals can be used for design of automated heart diseases analysis model. newlineThe different Bio-signal Processing methods are examined in this research for removal of artifacts and event detection. For ECG denoising and signal quality improvement the Frequency domain filtering, optimal filtering and Time Frequency domain filtering techniques are used. Frequency domain filtering involves design of High pass FIR Filter using windowing techniques; Notch Filter and Low pass IIR Filter using approximation methods. The analysis of these approaches depicts that High pass FIR Filter using Blackman window is more effective for elimination of low frequency Baseline wander noise whereas Low pass IIR filters Elliptic approximation method is found to have more potent for removing high frequency electromyography noise and IIR Notch filter for suppression of Power-line interference persists at 50 Hz. Optimal filtering includes the design of adaptive filters using LMS and NLMS algorithm for removal of burst noise. An NLMS algorithm introduces less error in the signal while the weight equation of LMS algorithm continuously update to recent input data. By employing high performance filters (both frequency domain filtering and optimal filtering), a cascade digital filter design is proposed that eliminates distinct noises present in the ECG signal which exhibits significant SNR improvement of 7.75dB at the cost of more computational complexity, large memory requirement that restricts its use for analysis of non-st