Associative memory algorithm for visual pattern recognition with memristor array and CMOS neuron

This paper presents the development of a neuromorphic system for visual pattern recognition. Associative memory algorithm has been used to recognize specific patterns and the method is implemented with discrete circuit elements that use memristors as the synapse. Weight of the synapses between the inputs and output neurons are adjusted by memristors. The amount of the neural block depends on how many patterns and each block contains 30 weighted synapses connected to the output. Each input corresponding to a pixel in a 6×5 pixel image generates voltage pulses according to the pixel value. The input voltage pulses weight are then maintained by the memristors and integrated by the output neurons. Each of the output of neurons is compared with a fixed threshold voltage which will determine whether the output bit is 0 or 1. Successfully demonstration of the system has been done by training and recognizing images of numbers from 1 to 8.