A wide-field direction-selective aVLSI spiking neuron

We describe an analog Very-Large-Scale Integrated (aVLSI) wide-field direction-selective spiking neuron that responds robustly to both natural and artificial images. The circuitry is based on a model of the wide-field direction-selective neurons in the fly visual system. The architecture of the chip closely follows the anatomical layout of the fly visual system. The circuit has a one-dimensional array of 37 elementary motion detectors (EMDs); each of which provides local motion information to the wide-field spiking neuron. Normalisation and adaptation are employed in the circuits at each processing stage. The inputs to the EMDs saturate with high contrasts so that the motion output is not confounded with the signal contrast. The outputs of the EMDs are aggregated in a nonlinear way to produce a motion output that is independent of the stimulus size. The global aggregated signal is then used to drive an integrate-and-fire neuron. All the computation is analog except for the output of the spiking neuron. We show results from the fabricated circuit.