Related Experiment Video
Updated: May 4, 2026
04:06
Author Spotlight: Enhancing Remote Rehabilitation with Virtual Reality and Electromyography
Published on: January 12, 2024
704
Ultrathin Gallium Nitride Quantum-Disk-in-Nanowire-Enabled Reconfigurable Bioinspired Sensor for High-Accuracy Human
Zhixiang Gao1, Xin Ju2, Huabin Yu1
1iGaN Laboratory, School of Microelectronics, University of Science and Technology of China, Hefei, 230029, People's Republic of China.
Nano-Micro Letters
|September 1, 2025
Summary
This study introduces a novel bioinspired vision sensor that mimics retinal cells for enhanced human action recognition (HAR). The integrated system significantly boosts HAR accuracy by combining dynamic and high-contrast visual processing.
Area of Science:
- Optoelectronics and Neuromorphic Engineering
- Advanced Materials Science (GaN/AlN)
- Computer Vision and Artificial Intelligence
Background:
- Human action recognition (HAR) is vital for computer vision, with neuromorphic systems offering solutions for sensor-processor bottlenecks.
- Current in-sensor vision computing is limited by the lack of integrated, versatile biomimicking functionalities, restricting computational capacity and scalability.
- Existing devices often focus on specific vision tasks, failing to capture the synergistic interactions present in biological vision.
Purpose of the Study:
- To develop a bioinspired vision sensor capable of mimicking retinal cell functions for improved in-sensor computing.
- To integrate functionalities of Parvo and Magno cells, along with their synergistic activity, into a single device for enhanced HAR.
- To demonstrate a tunable photoresponse for optimizing both image quality and HAR efficiency within a single sensor.
Main Methods:
- Fabrication of a GaN/AlN-based ultrathin quantum-disks-in-nanowires (QD-NWs) array.
- Mimicking Parvo cells (high-contrast) and Magno cells (dynamic vision) using the QD-NW array.
- Achieving dual photoresponse characteristics (slow and fast) by tuning bias voltage for in-sensor vision computing.
Main Results:
- The developed QD-NW array successfully mimicked both high-contrast and dynamic visual responses.
- Tuning bias voltage enabled distinct slow and fast photoresponses, enhancing image quality and HAR efficiency.
- Synergistic interaction of the two modes within the sensor significantly increased HAR recognition accuracy from 51.4% to 81.4%.
Conclusions:
- The proposed bioinspired vision sensor effectively integrates multiple biomimicking functionalities for advanced in-sensor computing.
- This intelligent vision sensor demonstrates a promising platform for highly efficient HAR systems.
- The device opens avenues for the development of next-generation smart optoelectronics with enhanced computational capabilities.
