A Pupillary Light Reflex Inspired Self-Adaptive Spiking Visual Neuron

The development of neuromorphic visual systems aims to address the energy-efficiency and adaptability constraints in machine vision. However, artificial visual neurons in these systems mostly encode amplitude-modulated signals and adjust the perception range through passive gate voltage modulation, resulting in low biological fidelity. A pupillary-light-reflex-inspired self-adaptive spiking visual neuron with a superior perception range and active visual adaptation is proposed. The device functionally emulates the hierarchical visual adaptation process of human eyes through the active optical regulation of a photochromic film, photoelectric conversion through an IGCdO-based transistor, and spiking encoding through a TaO_X-based memristor-based memristor). This configuration possesses a perception range of 160 dB and active visual adaptation under extreme light intensity conditions ranging from 0.2 µW cm^-2 to 1.64 W cm^-2, outperforming previous artificial visual neurons. The advantage of active visual adaptation has been validated by integration with a spiking neural network, achieving an 86% recognition accuracy in classification tasks, a 66% improvement over non-adaptive counterparts. This bio-inspired design would endow machine vision systems with a high-level of biological fidelity.