Electrocardiogram signal processing for baseline noise removal using blind source separation techniques: A comparative analysis

Efficient processing of Electrocardiogram (ECG) signal is a major goal of this paper. ECG signal measured from the surface of human body is badly effected by different noises like power line noise, muscle noise, lung noise baseline noise. The main causes of baseline noise are breathing, lose sensor contact and body movements. Baseline noise is a low frequency signal. Baseline will disturb the whole cycle of ECG signal but mainly the low frequency part. T-wave is a part of ECG signal related to a phenomenon associated with increased risk of death and baseline wandering may prevent its correct detection. Different algorithm were applied for removal of this noise by different researchers like Kalman filter, Cubic spline, Moving average. We apply projection pursuit gradient ascent algorithm in a previous paper and observe the results in term of standard deviation of the error signal for different types of baseline noises. In this paper I am comparing the results of projection pursuit gradient ascent algorithm (PPGAA) and Independent Component Analysis (ICA) FASTICA algorithm for removing this noise. Results of FASTICA were also analyzed in the presence of 32db additive Gaussian noise. Then a comparative study of these algorithms was done and results were compared for baseline noise removal.