Identification of nonlinear systems using a Trigonometric Polynomial Neural Network

A Trigonometric Polynomial Neural Network, called TPNN, is proposed in this paper. TPNN is a new scheme of neural network based on a trigonometric polynomial which is familiar in Fourier Analysis. The proposed network is linear with respect to its coefficients and the well known recursive least squares method of linear parameter estimation can be used as a learning algorithm. Using the TPNN, learning of nonlinear functions and identification of nonlinear discrete time systems are examined with some additional comments for the type of nonlinear systems. The efficiency of the proposed method is also certified by applying the identified system for control of nonlinear discrete time system.