Combination of two NLMP algorithms for nonlinear system identification in alpha-stable noise

The normalized least mean pth power (NLMP) algorithm based on adaptive Volterra filters has conflicting requirement of fast convergence rate and low steady-state error. To address this problem, a novel combination of two NLMP (CNLMP) algorithms is proposed which adaptively combines two independent NLMP filters with large and small step sizes to obtain fast convergence rate and low misadjustment in the presence of α-stable noise. Additionally, to achieve fast convergence at the transition stage, a tracking weight transfer scheme is proposed. Simulation results demonstrate that the proposed algorithm is superior to the NLMP, LMP and NLMAD algorithms for nonlinear system identification problem in terms of convergence rate and steady-state kernel error.