Bio-plausible visual neural network for spatio-temporally spiral motion perception

Abstract

Abstract Neurophysiological studies validate that the primate cerebral cortex includes spiral neurons whose visual properties respond preferentially to spiral motion patterns. However, the biological mechanism of which a vision system perceives spiral motion is unclear, while few computational models are reported to discuss the problem of spiral motion perception. In order to fill this gap, this work develops a spiral motion perception neural network in terms of the recent achievements in neurophysiology and simulates the visual response characteristics of spiral neurons. One such network, inspired by two stages of biological visual information processing includes two subnetworks- presynaptic and postsynaptic neural networks. The former comprises multiple lateral inhibition neural sub-networks for the capture of visual motion information, whereas the latter extracts different rotational and radial motion cues and synthesizes them to detect the process of the spiral motion of an object. Experimentally, the proposed neural network is sufficiently examined by different types of spiral motion patterns in non-interference or interference environments. Numerically comparative experiments show that it can effectively detect spiral motion patterns of the object and also does not respond to any non-spiral motion, which is consistent with the metaphor of spiral neurons’ performance perception.