Training a spiking neural network to generate online handwriting movements

Abstract

In this paper we developed a spiking neural network model that learns to generate online handwriting movements. The architecture is a feed forward network with one hidden layer. The input layer uses a set of Beta elliptic parameters. The hidden layer contains both excitatory and inhibitory neurons. Whereas the output layer provides the script coordinates x(t) and y(t). The proposed spiking neural network has been trained according to Sander Bohet model. The trained spiking neural network has been successfully tested on MAYASTROUN data base. Also a comparative stady between the proposed spiking neural network and an artificial neural network proposed in a previous work is established.