A Bioinspired Optic-Neural Device with Neuromorphic Spatiotemporal Visual Perception for Dynamic Object Recognition

Area of Science:

• Materials Science
• Neuroscience
• Computer Science

Background:

• Neuromorphic vision aims to overcome latency and storage issues in machine vision by integrating sensing, memorizing, and computing.
• Simulating the human visual system for spatiotemporal perception in a single device remains a significant challenge.

Purpose of the Study:

• To demonstrate a bioinspired optic-neural device capable of dynamic object recognition.
• To achieve spatiotemporal visual information perception mimicking the human vision system.

Main Methods:

• Utilized molybdenum disulfide (MoS2) for simultaneous encoding and sensing functions.
• Leveraged optical synaptic plasticity based on charge trapping and recombination.
• Arranged optic-neural devices in an array for nonlinear encoding and mapping of light inputs.

Main Results:

• The device demonstrated excellent optical synaptic plasticity, remembering historical visual information.
• Achieved dynamic object recognition with high classification accuracy (up to 96.3%) on five datasets.
• Showcased high energy efficiency and low computational load in the neuromorphic vision system.

Conclusions:

• The developed optic-neural device successfully mimics the human vision system's spatiotemporal information processing.
• This research holds significant implications for advancing machine vision systems.
• The device offers a promising solution for efficient and high-performance dynamic object recognition.