Modified sparse representation for ECG reconstruction in telemonitoring

Presently ECG signal telemonitoring is one of the essential branch in telemedicine system. So it is highly desirable to construct a robust telemonitoring system through wireless body area network (WBAN) with consumption of less energy and less power. A number of traditional ECG reconstruction techniques have been proposed to recover the clean ECG data. However because of some specific behavior and characteristics of raw ECG data like non sparsity and heavy contamination of noise the traditional methods do not succeed in this application. This paper proposes an effective reconstruction method followed by Independent Component Analysis (ICA) in an intention to obtain the clean ECG data. The proposed framework includes context based learning adopted reconstruction method. Experimental results along with simulation results show that this framework is able to reconstruct the raw ECG recordings with high accuracy and high quality. Context based learning learns the existing context in the signal and reacts to changing context, which uses k-means clustering via singular value decomposition i.e. KSVD algorithm to recover raw ECG signal. In this paper, the proposed method shows better reconstruction performance than the traditional compressed sensing method retaining the sparsity of the ECG signal intact.