Adaptive Exponential Trigonometric Functional Link Neural Network Based Filter Proportionate Maximum Versoria Least Mean Square Algorithm

Abstract

Purpose The traditional proportionate-type algorithms used for sparse system identification are robust to Gaussian noise. However, real sparse systems to be identified are also affected by both nonlinearity and non-Gaussian noise environments. So, the purpose of this paper is to propose a novel AETFLN-FPMVLMS algorithm in this paper to compensate for the system's nonlinearity and sparsity. Method To overcome the issue of nonlinearity due to the presence of passive devices or due to the effect of noise or distortions, the adaptive exponential functional link neural network (AETFLN)-based input expansion is used in this paper for the proposed algorithm. The FPNLMS algorithm is used here to update the adaptive filter coefficients as it exploits the sparsity of the systems thereby enhancing the convergence speed and the steady-state behavior. Lastly, the P-MVC approach is applied to filter the proportionate normalized least mean square (FPNLMS) algorithm to compensate for the non-Gaussian noise during the sparse system identification. Result Simulation results also show that the proposed algorithm is robust in a non-Gaussian noise environment compared to other algorithms with improved performance.