Gradient Based Neural Network with Fourier Transform for AR Spectral Estimator

Abstract

In this paper we propose a gradient based neural network to compute the-order AR parameters by solving the Yule-Walker equations. Furthermore, to reduce the size of the neural network, we derive a compact architecture of the discrete time gradient based neural network using the fast Fourier transform. For this purpose, the product of the weights matrix and the inputs vector which constitutes the activation of the neurons is performed efficiently in () operations and storage instead of (2) in the original discrete time gradient based neural network. Simulation results show that proposed neural network architecture is faster and leads to the same results as the original method which prove the correctness of the proposed scheme.