Comparison of Backpropagation and Hopfield Model in De-noising of Speech Signal

Abstract

In this project work, we have used two algorithms of Neural Network (NN): Backpropagation and Hopfield NN to de-noise speech signal. The backpropagation algorithm is found suitable to remove random noise but very poor in the removal of awgn (Additive White Gaussian Noise). The Hopfield NN shows completely reverse performance i.e. suitable for awgn but very poor for random noise. The performance of both algorithms is measured graphically with an original recovered signal, MSE, the convergence of regression and error histogram.