Feed-Forward Neural Network for Graphical Symbol Recognition in Additive Noise Environment

This article proposes a feed-forward backpropagation neural network architecture for the automatic recognition of graphical symbols in the presence of high-level additive Gaussian noise based on machine learning (ML). The article presents the comparison of two different backpropagation training algorithms and presents numerical results which compare the performances of a Levenberg-Marquardt backpropagation algorithm and One step secant backpropagation algorithm. The procedures of two-layer neural network architecture creating and training are designated in detail. The neural network is depicted in MATLAB and Simulink environment [1].