Robust filtering employing bias compensated M-estimate affine-projection-like algorithm

This Letter presents a bias compensated M-estimate affine-projection-like algorithm for robust adaptive filtering in case of noisy input signal and impulsive noise interference. In order to mitigate the effect of bias introduced due to noisy input signal, a bias compensation vector is derived along with robust cost function minimisation employing M-estimate and accordingly the weight update learning process of the adaptive filter has been improved. In conjunction with this, a new input noise variance estimation method is introduced using the median filter. The proposed algorithm is evaluated for acoustic echo cancellation application and compared with other existing methods. Simulation results demonstrate its effectiveness in terms of enhanced convergence and better steady-state performance in case of impulsive environment and input signal contaminated with noise.