Evaluation of Letter Reproduction System Using Cellular Neural Network and Oxide Semiconductor Synapses by Logic Simulation

Abstract

A letter reproduction system has been developed by a cellular neural network and oxide semiconductor synapses and evaluated by logic simulation. A cellular neural network is utilized because it is suitable for large-scale integration of electronic devices, and oxide semiconductor devices are utilized as synapse elements because the characteristic deterioration can be employed as strength plasticity of synaptic connection with modified Hebbian learning. In this article, first, the structure and operation of the letter reproduction system are explained. Next, it is evaluated by logic simulation, where the dependence of the correction accuracy of the letter reproduction on the variation of deterioration rate of the oxide semiconductor synapses is analyzed.