A Modified Neural Filtering Algorithm for Tracking of Chaotic Signals

In this study, a modified neural filtering algorithm is presented for tracking of chaotic signals. A multilayer neural network (MLNN) structure is used in proposed design as a nonlinear adaptive filtering tool. Initially, the MLNN is linearized using Taylor series expansion and then the weight vector update rule is designed by using Lyapunov stability theory (LST) to adaptively update the weights of the MLNN. The tracking capability of the proposed algorithm is improved by using adaptation gain rate parameter "a(k)" which is iteratively adjusted itself depending on sequential tracking errors rate. The tracking ability of the proposed algorithm is tested on two chaotic signals and compared with conventional algorithms. The simulation results have supported that the proposed neural filtering algorithm achieved better performance.