Motion artifact reduction in PPG signals

Abstract: The aim of this thesis was to investigate methods for artifact removal in PPG signals and to implement and evaluate a few existing algorithms claiming that the amplitude information is recovered when removing motion artifacts from photoplethysmographic signals (PPG) captured from pulse oximeters. We developed a new proposed method that uses a two-stage based approach with singular value decomposition and fixed fast ICA algorithm in order to generate a PPG-correlated reference signal that is used in adaptive noise cancellation. The results were promising and our proposed method is easy to implement and converges quickly with good extraction performance. It has a few design parameters and only needs the estimated period of the PPG signal. Our method could be used in a clinical routine for prediction of intradialytic hypotension. However it should be mentioned that although our method has great potential the simulations were only conducted on two healthy males. Further studies on a larger dataset might be needed in order to establish a full value of the efficacy of our method.