Neuro-inspired optical sensor array for high-accuracy static image recognition and dynamic trace extraction

Area of Science:

• Materials Science
• Artificial Intelligence
• Edge Computing

Background:

• Edge computing demands efficient data processing for AI applications.
• Neuro-inspired vision systems require hardware capable of processing both static and dynamic visual information.
• Developing low-power, high-performance sensors is crucial for advanced artificial vision.

Purpose of the Study:

• To demonstrate a neuro-inspired optical sensor array based on 2D NbS2/MoS2 hybrid films.
• To achieve integrated sensing, memory, and contrast enhancement for static images.
• To enable in-sensor trajectory registration for moving targets.

Main Methods:

• Fabrication of a 10x10 array of NbS2/MoS2 phototransistors.
• Characterization of photo-induced conductance plasticity and energy consumption.
• Experimental implementation of in-sensor trajectory registration for moving light spots.

Main Results:

• The sensor array exhibited remarkable photo-induced conductance plasticity and low energy consumption.
• High image recognition accuracy was achieved when integrated with convolutional neural networks (CNNs).
• Accurate restoration of moving light spot trajectories was demonstrated through in-sensor processing.

Conclusions:

• The developed neuro-inspired optical sensor array offers integrated functionalities for static and dynamic visual data processing.
• This platform shows significant potential for implementing high-performance artificial vision systems in edge computing.
• The use of NbS2/MoS2 hybrid films provides a promising route for advanced neuromorphic sensing applications.