A Novel Design of Combined Systolic and Convex Architectures for Efficient High-Speed Adaptive Filter

This paper proposed a novel design for a high-speed adaptive filter, specifically a new filter structure employing a convex combination architecture incorporating systolic architecture. The performance of convex architecture and systolic architecture are analyzed for LMS and RLS filters in two different orders. Systolic architecture improves the speed of convergence and SNR of the filters. Additionally convex combination architecture significantly improves the overall filter performance, especially in steady-state conditions. The combination of these two filter design techniques results in significantly improved convergence speed and steady-state performance of filters. The performance of the proposed filters combined two architectures outperforms the conventional filters in several metrics for filter design, including convergence speed, SNR, Steady-state MSE, learning capability, and filtering capability in terms of residual noise. Simulation results demonstrate the effectiveness of proposed filters in denoising ECG signals from PLI noise. The suggested filter configurations undergo evaluation for the purpose of ECG noise elimination for a set of ten ECG signals, acquired from the MIT-BIH database.