A hardware-based approach for implementing biological visual cortex-inspired image learning and recognition

In this paper, we report a simulation based study of large-scale image learning and recognition using neural network consisting of active pixel sensor (APS), LIF neurons, and memristive devices as synapses in crossbar array. Our studies indicate that images can be efficiently encoded into spiking-patterns using the proposed model which can be used to train the memristive devices based on spike-timing-dependent-plasticity (STDP). The proposed scheme provides a robust approach for encoding, learning, and recognizing the large-scale images using hardware-based neural-circuits.