Related Experiment Video
Updated: Jun 27, 2026
08:07
Assembly and Characterization of Biomolecular Memristors Consisting of Ion Channel-doped Lipid Membranes
Published on: March 9, 2019
8.3K
Neuromorphic Visual Receptive Field Hardware with Vertically Integrated Indium-Gallium-Zinc-Oxide Optoelectronic
Hyun Wook Kim1, Jin Hong Kim1, Dong Hoon Shin1
1Department of Materials Science and Engineering and Inter-University Semiconductor Research Center, Seoul National University, Seoul, 08826, South Korea.
Advanced Materials (Deerfield Beach, Fla.)
|September 25, 2025
Summary
This study introduces an artificial retina using optomemristors and neuristors for event-driven processing. This neuromorphic vision system achieves high accuracy in pattern classification by mimicking biological visual receptive fields.
Area of Science:
- Neuromorphic Engineering
- Artificial Intelligence
- Materials Science
Background:
- Conventional von Neumann systems face energy efficiency challenges in data processing.
- Neuromorphic vision systems utilizing spiking neural networks offer potential for improved energy efficiency.
- Biological retinas perform efficient pre-processing of visual information via receptive fields.
Purpose of the Study:
- To propose an artificial retinal neuron integrating an optoelectronic memristor and a neuron transistor.
- To develop a neuromorphic vision system inspired by biological visual receptive fields (VRFs).
- To enhance energy efficiency and feature extraction capabilities in artificial vision systems.
Main Methods:
- Vertical integration of an optomemristor (In-Ga-Zn-O) and a neuristor (Si FET) to create an artificial retinal neuron.
- Implementation of excitatory (ON-type) and inhibitory (OFF-type) cells mimicking biological VRFs.
- Development of a dual-type configuration combining both ON- and OFF-type cells for enhanced feature extraction.
Main Results:
- The artificial neuron performs in-sensor pre-processing, extracting essential image features like edges.
- ON-type cells show increased spiking frequency with light, while OFF-type cells show decreased frequency.
- The dual-type configuration achieved 99.8% accuracy in fingerprint pattern classification, significantly outperforming the single-type configuration (56.1%).
Conclusions:
- The proposed artificial retinal neuron effectively mimics biological VRFs for efficient edge information extraction.
- Event-driven processing in this neuromorphic system offers superior performance and energy efficiency.
- This technology holds promise for advanced artificial vision applications requiring high-accuracy pattern recognition.
