A Bio-inspired Perceptual Decision-Making Circuit Based on the Hassenstein-Reichardt Direction Detector

Abstract

AbstractDecision-making belongs to one of the most prominent principles in the nerve system of living species. The crucial trade-off between time to decide (speed) and accuracy is essential for a creature to react to environmental stimuli and most importantly, relevant for survival. In-depth studies of coherent motion tasks in primates, based on two-choice decision experiments, led to a profound understanding of neuronal information processing and even bridged the gap to behavioural psychology. Three fundamental processing stages are needed to successfully perform a two-choice decision experiment based on a saccadic eye movement task: the neuronal representation of the sensory signal, the integration of the incoming stimuli and the comparison of the accumulated information to a threshold for a final decision. In accordance to the results of decision-making experiments with primates, an analogue electronic decision-making circuit was developed. Our concept study includes a LED-matrix showing moving light dots (i.e. the tasks screen in primate experiments), an array of photo diodes (i.e. retina of the primate, sensory evidence), a Hassenstein-Reichardt Detector (HRD) based motion detection (i.e. a correlation-type motion detector with a high biological relevance, based on an inhibitory circuit and delay lines) and finally a signal integration circuit. The latter stage represents the decision-making unit and is realised through an inhibitory mutual coupling of the accumulated signals. The entirely analogue circuit is based on integrate and fire relaxation oscillators to mimic firing neurons. We would like to emphasise that the presented real time decision-making system is closely related to biological information pathways in nerve systems. Neither a digital computer nor programming is necessary for decision-making, which might be an interesting approach for real time data processing in autonomous robots.