Ultrafast machine vision with artificial neural network devices based on a GaN-based micro-LED array.

Area of Science:

• Optoelectronics
• Artificial Intelligence
• Materials Science

Background:

• Gallium Nitride (GaN)-based micro-light-emitting diodes (LEDs) are emerging devices with dual functionality as displays and photodetectors.
• Their photodetector capabilities open avenues for advanced machine vision applications.
• Existing systems often lack the speed and integration required for complex real-time processing.

Purpose of the Study:

• To experimentally characterize GaN-based micro-LEDs as photodetectors.
• To propose and simulate a novel artificial neural network (ANN) device utilizing a micro-LED photodetector array.
• To explore the potential of this integrated system for ultrafast imaging and machine vision.

Main Methods:

• Experimental measurement of micro-LED photodetector characteristics.
• Development of a feasible simulation for an ANN device based on a micro-LED photodetector array.
• Analysis of synaptic weight tunability via bias voltage within the array.

Main Results:

• Demonstrated ultrafast imaging capabilities at approximately 133 million bins per second.
• Achieved a high image recognition rate using the micro-LED array as a neural network.
• Confirmed that synaptic weights are tunable by bias voltage, enabling reconfigurable computing.

Conclusions:

• GaN-based micro-LEDs can be effectively utilized as photodetectors for advanced machine vision.
• The proposed ANN device offers significant potential for high-speed imaging and pattern recognition.
• The multi-functionality of micro-LEDs allows for integration with display, communication, and computing, expanding application possibilities.