QRS Complex Separation from Convolutive Mixtures of Biolectrical Signals Acquired by Wearable Systems

Independent component analysis (ICA) has been widely used to remove artefacts from multichannel biomedical signal acquisitions under the hypothesis that there is statistical independence among the original sources. However, the basic ICA model does not take into account the influence on the mixing process of the different paths from the signal sources to the sensors In this study we propose a convolutive mixtures model in order to overcome the limitations of the basic ICA approach. The independent components are estimated in the frequency domain, where the convolutive model can be solved through an instantaneous mixing model. The signals are reconstructed back to the observation space resolving the ICA model ambiguities. Simulations are carried out to optimize of the proposed method for convolutive mixtures of electrocardiographic (ECG) and motion artefacts signals. The algorithm is tested on real ECG signals acquired by wearable systems in order to preserve the QRS complex when the signals are degraded by real life conditions of acquisition.