On the design of dynamic adaptive exponential linear-in-the-parameters nonlinear filters for active noise control

Adaptive exponential functional link network (AEFLN) is a recently developed linear-in-the-parameters nonlinear filter, which provides significantly better convergence performance over other traditional linear-in-the-parameters nonlinear filters. To further improve the convergence characteristics of AEFLN, a variable step-size AEFLN (VSS-AEFLN) is proposed in this paper. An adaptive exponential variable step-size least mean square (AEVSS-LMS) algorithm is developed, and the same is tested on modeling benchmark nonlinear plants. Following the above formulation, a VSS-AEFLN-based nonlinear active noise control (ANC) system is designed, and an adaptive exponential filtered-s variable step-size least mean square (AEFsVSS-LMS) algorithm is also developed for improved noise mitigation. Simulation results show that the convergence performance of the proposed algorithms, for system identification and ANC systems, is superior to the state-of-the-art linear-in-the-parameter nonlinear adaptive filters