A Bio-Inspired Adaptive Retinal Processing Neuron with Multiplexed Spiking Outputs

We present an adaptive, current-mode analogue neuron circuit (Boahen, 1996), which adapts its spiking output frequency and pulse width relative to the magnitude and vector of the computation of its input nodes. By modulating the pulse width to represent the output vector signed double byte data transfer protocol along a single channel is enabled: doubling output data bandwidth. This artificial neuron consists of a four quadrant current subtraction circuit, a low power current comparator, and an adaptive spiking output circuit, implementing novel, distributed, focal plane image processing, enabling real-time operation. Each cell consumes 15 pico-joules, and takes 20 milliseconds to process an output decision: allowing it to be integrated within large scale arrays. This biologically inspired circuit imitates the spatial image processing at the Bipolar, Ganglion cell nodes of the retina, and has used these cells´ functions to define its processing algorithms.