Retina-inspired spike processing for neuromorphic color recognition

Conventional machine vision systems are hindered by constrained adaptability, particularly in dynamic and unpredictable environments. Herein, we present a neuromorphic color recognition system inspired by the intricacies of retinal signal processing, constructed through a hierarchical bio-mimetic framework. The system integrates a broadband photosensor to emulate the spectral selectivity of cone cells and employs an ion-gel-gated oxide transistor to replicate synaptic dynamics, both of which are integral to achieving highly efficient color recognition. In addition, a dual-threshold algorithm is incorporated, enabling precise control of robotic motions. The system's event-driven architecture with a hierarchical coding strategy enhances dynamic perception, collectively rendering it highly adaptive and highly efficient for environmental interactions.