Industry News
Researchers Achieve Human Colour Vision in Self-Powered AI Chip for Edge Devices
In a significant leap forward for edge computing and artificial intelligence (AI), researchers at Tokyo University of Science (TUS) have developed a pioneering self-powered artificial synapse capable of advanced colour recognition. This development addresses longstanding challenges in machine vision, namely, the high power and computational demands that limit the deployment of vision systems in mobile and embedded devices such as smartphones, drones, and autonomous vehicles.
The breakthrough, published in Scientific Reports (Volume 15, May 12, 2025), comes from a team led by Associate Professor Takashi Ikuno from TUS’s School of Advanced Engineering. Co-authored by Hiroaki Komatsu and Norika Hosoda, the study introduces a novel device that mimics the energy-efficient processing of the human visual system while eliminating the need for external power sources.
Rethinking Machine Vision with Bio-Inspired Engineering
Traditional machine vision systems require significant hardware and energy to process the flood of visual data they collect. In contrast, the human eye selectively filters visual input, enabling efficient perception with minimal computational overhead. To emulate this, the TUS team turned to neuromorphic computing, a field focused on replicating the brain’s neural architecture in hardware.
However, two major barriers have hindered progress: the inability of artificial systems to recognise colour at human-like resolution, and the need for continuous power to operate such devices. The TUS team overcame both by integrating two dye-sensitised solar cells into a novel artificial synapse. These solar cells respond differently to specific wavelengths of light, enabling high-fidelity colour recognition while simultaneously powering the device via solar energy.
Human-Like Colour Perception and Beyond
The result is a compact, solar-powered synapse that can differentiate colours across the visible spectrum with a resolution of 10 nanometers, approaching the sensitivity of the human eye. The device also demonstrates bipolar voltage responses, producing distinct electrical signals depending on light colour (positive for blue, negative for red), enabling it to execute logical operations typically requiring multiple components.
"The results show great potential for the application of this next-generation optoelectronic device, which enables high-resolution colour discrimination and logical operations simultaneously, to low-power artificial intelligence (AI) systems with visual recognition," notes Dr. Ikuno.
Real-World Testing and Industry Applications
To showcase its practical value, the team applied the synapse within a physical reservoir computing framework, achieving 82% accuracy in classifying 18 combinations of human movements and colour inputs, using just one device. This contrasts with traditional systems that rely on multiple photodiodes and greater power consumption.
The implications for industry are wide-ranging:
- Automotive: Enhances recognition of traffic signals and road conditions in autonomous systems.
- Healthcare: Enables continuous, low-power monitoring in wearable biometric devices.
- Consumer Electronics: Opens the door to longer battery life in smart glasses, mobile devices, and augmented/virtual reality headsets.
"We believe this technology will contribute to the realisation of low-power machine vision systems with colour discrimination capabilities close to those of the human eye, with applications in optical sensors for self-driving cars, low-power biometric sensors for medical use, and portable recognition devices," remarks Dr. Ikuno.
A Step Closer to Human-Level Perception in Edge AI
As AI pushes further toward the edge, the need for intelligent, power-efficient vision systems is more critical than ever. This research represents a transformative advance, blending the best of biology and engineering to make machines that not only see the world, but see it like we do.
