ABSTRACT:- Image identification using memristor in neuromorphic system is a specific application to demonstrate the usefulness and popularity of memristor in the future by the outstanding features such as storage capacity, high integration density, low power energy. This CMOS-memristor hybrid architecture identifies 10 images, each image is a 5×6 pixel array, each pixel is a signal to the system. There are 30 signals into the system and these signals pass through 300 memristors, divided into 10 memristor arrays, each array will respectively have 30 memristors. The system has 10 signals going out from 10 memristor arrays then continues into the integrated block. The integrated block has two components which are neurons and switch controller. Image identification performs two modes in training and test. An on-chip training mode is implemented by a simpler circuit controller in this paper. The recognition rate is evaluated in consideration of additive noise.

Keywords – Neuromorphic, memristor, CMOS, crossbar array, image identification