Search

This paper proposes a deep cerebellar model articulation controller (DCMAC) for adaptive noise cancellation (ANC). We expand upon the conventional CMAC by stacking single-layer CMAC models into multiple layers to form a DCMAC model, and derive a backpropagation training algorithm to learn the DCMAC parameters. Compared with the conventional CMAC, the DCMAC can characterize nonlinear transformations more effectively because of its deep structure. Experimental results confirm that the proposed DCMAC model outperforms the CMAC in terms of residual noise in an ANC task, showing that the DCMAC provides enhanced capability to model channel characteristics.

Electrocardiogram (ECG) is a non-invasive method to monitor electric activities inside the heart. The signals observed on the surface of human body have very low amplitude, and thus, ECG is highly vulnerable to noise. One of the most devastating noise is power line interference (PLI) and its harmonics, which are interlaced with ECG signal even if the ECG equipment is operated on battery. The problem is further complicated when the frequency of PLI is not static, making the conventional notch filter completely ineffective. High-resolution electrocardiogram (HRECG) is a specialized technique in which higher frequency components present in the ECG signal are observed; here, we need to eliminate the harmonics of PLI as well. In this paper, we propose an intelligent adaptive noise rejection filter, which tracks and eliminates PLI as well as its harmonics. The proposed system can estimate the frequency of PLI and tune the adaptive filter for precise elimination of PLI as well as its harmonics without the requirement of an auxiliary reference input. The proposed system is based on recursive state space model, inherited with less computational complexity and performs well in a non-stationary environment. The proposed system responds well to the ongoing variations in amplitude and frequency of PLI present in the HRECG signal as well as intracardiac signal. The proposed system does not require any reference signal for tracking the PLI and its harmonics, and it is capable to self-adjust its tracking frequency for highly precise filtration of first, third, and fifth harmonics of PLI.