A Bias-Compensated Robust Affine Projection Algorithm Using Variable Matrix-type Step Size

We propose a novel system identification algorithm using adpative filter, which demonstrates superior performance compared to existing robust Affine Projection Algorithms (APAs). Through systematic simulations, we validate the effectiveness of our algorithm in handling impulsive noise and input noise, crucial issues in practical applications of adaptive filters. Our approach incorporates variable matrix-type step sizes, derived through meticulous analysis of Mean Squared Deviation (MSD) curves, to enhance adaptability to diverse input vector characteristics. Furthermore, we introduce a compensation vector to mitigate bias induced by input noise, ensuring accurate estimation even in noisy environments. The algorithm proposed in this study demonstrates fast convergence and low steady-state error. This indicates the effectiveness of the proposed algorithm in performing system estimation.