The relative depth of objects causes small shifts in the left and right retinal positions of these objects, called binocular disparity. The binocular energy model has been proposed to model the response of disparity selective cells in the visual cortex. In this model, different disparities may be encoded either by a shift in the position or the phase between the monocular receptive field profiles of the two eyes. In the cat, disparity appears to be encoded primarily by phase. We have built and characterized an electronic implementation of the binocular energy model. Our system consists of two silicon retina representing the left and right eyes, two silicon chips containing retino-topic arrays of spiking neurons with monocular Gabor-type spatial receptive fields (RF), and circuits that c...[ Read more ]

The relative depth of objects causes small shifts in the left and right retinal positions of these objects, called binocular disparity. The binocular energy model has been proposed to model the response of disparity selective cells in the visual cortex. In this model, different disparities may be encoded either by a shift in the position or the phase between the monocular receptive field profiles of the two eyes. In the cat, disparity appears to be encoded primarily by phase. We have built and characterized an electronic implementation of the binocular energy model. Our system consists of two silicon retina representing the left and right eyes, two silicon chips containing retino-topic arrays of spiking neurons with monocular Gabor-type spatial receptive fields (RF), and circuits that combine the spike outputs to compute a disparity selective complex cell response. However, mismatch in the analog transistors implementing the Gabor-like neurons produce mismatch between the monocular receptive field profiles. Our measurements indicate that the relative responses between neurons tuned to different disparities are better matched in the phase based model than in the position based model. We performed a numerical sensitivity analysis that confirms that the relative responses of neurons in the phase model are more robust in the presence of monocular receptive field mismatch and that serendipitously, the robustness is greatest for those parameters of the receptive fields which vary the most on the chips. We conjecture that this robustness may be one factor explaining the preference for phase encoding in biological neuron.

Keywords- binocular vision, disparity, primary visual cortex, depth perception, receptive field, binocular energy model, complex cell, simple cell, disparity selective neurons, analog integrated circuits, Gabor filters, image processing, image sensors, neuromorphic engineering, address event representation