Synaptic rectification model equivalent to the correlation-type movement detector

Abstract

Neural models which are equivalent to the correlation-type movement detector are described. The models involve contrast-coding channels which comprise bandpass linear filters followed by synaptic retifiers. Linear, one-directional lateral interactions are assumed among the contrast-coding channels. Synaptic rectifiers convert linear spatial interaction into a multiplication-like (quadratic) interaction, which is the core of the correlation-type movement detector. One of the neural models (E-I model) well approximates the correlation model in both time-averaged and dynamic (instantaneous) responses. Possible applicability of the model to movement detection by insects is discussed.