Particle Swarm Optimization of Morphological Filters for Electrocardiogram Baseline Drift Estimation

Abstract

Electrocardiogram (ECG) is the most vital biosignal of the body. It contains a variety of important clinical pieces of information and it is the fastest approach to asses the health condition. However, ECG is highly susceptible to noise and low-frequency interference. The low-frequency interference in ECG appears in the form of baseline drift that as an essence of preprocessing prior to any clinical observation or computer-aided analysis, it should be estimated and removed. A strong classical approach to ECG baseline drift estimation and cancelation is Morphological Filtering (MF). The two main variants of MF algorithms have been presented in the literature for ECG baseline estimation known as 1-stage MF and 2-stage MF algorithms. The one-stage MF deploys a Structuring Element (SE) for morphological smoothing of the ECG. The length of SE deployed half of the ECG period. Since the ECG period estimation is quick and easy, the presetting of 1-stage MF baseline estimation is fast. The 2-stage MF algorithm is more accurate with a more efficient estimation of baseline. The 2-stage MF has two stages of ECG morphological smoothing and the length of SE for each stage is according to the length of ECG features, therefore, an ECG feature estimation is an essential need for the setting of 2-stage. While both classical MF baseline estimations require MF parameter setting by prediction of ECG features, this chapter investigates a faster technique of parameter setting in 1-stage morphological ECG baseline estimation through Particle Swarm Optimization (PSO).