Suppression of Artifacts for Mobile ICG Using Nonlinear Adaptive Algorithms

Abstract

Impedance cardiography (ICG) is advantageous to identify various heart diseases. In recent times, mobile ICG has grabbed the attention of researchers and analysts for real-time supervision and automatic diagnosis. However, the attainment of ICG’s survey system was degraded due to many interrupts created by subject movement, which paves to wrong diagnosis. Several attempts have been performed to abolish the noise from clinical ICG signal using distinct digital signal processing techniques. Those approaches are not directly synced to be used for the mobile ICG environment and regular noises. Basically, motion artifact still is an open problem in mobile ICG. In this paper, an advanced process of adaptive artifact elimination proposed through impedance cardiography (ICG) signals. This is a composite exemplary based on wavelets and adaptive filter. The prime aspect of this methodology is the realization of adaptive noise canceller (ANC) without any reference signal. In the real-time medical environment during critical conditions due to heartbeat disorders, the filter coefficients become negative. This convergence unbalance leads to low filtering capability. In order to solve this issue, one may incorporate NN adaptive algorithms in the suggested ANC. To enhance the attainment of ANC, error normalization is adapted to change filter coefficients automatically. Again, in order to minimize computational complexity and to avoid overlapping of data samples at the input stage of the filter, a hybrid version of nonnegative and sign sign-based algorithms is considered for implementation. The resulting hybrid versions are exponential normalized nonnegative least mean square (eN3LMS) algorithm, exponential normalized nonnegative sign regressor LMS (eN3SRLMS) algorithm, exponential normalized nonnegative sign error LMS (eN3SELMS) algorithm, and exponential normalized nonnegative sign sign LMS (eN3SSLMS) algorithm. Finally, various ANCs are developed using these algorithms, and attainment measures are calculated and compared. Several implemented ANCs are verified on real impedance cardiogram signals.