Summary: Researchers have created a self-powered artificial synapse that can recognize colors with near-human precision. Unlike conventional systems that require external power and heavy data processing, this device mimics aspects of biological vision and generates its own electricity using dye‑sensitized solar cells.
Capable of distinguishing color differences as small as 10 nanometers across the visible spectrum, the device also supports light-wavelength-based logic operations. This breakthrough opens the door to energy-efficient, high-performance machine vision for edge devices such as smartphones, wearables, drones, and autonomous vehicles.
Key Facts:
- Self-powered vision: The synapse integrates dye‑sensitized solar cells to generate power from light without external energy supplies.
- High color resolution: It distinguishes color wavelengths with roughly 10‑nanometer precision across the visible range.
- Edge-ready AI: The device supports efficient on-device visual recognition and logic operations in resource‑limited environments.
Source: Tokyo University of Science
Context: As artificial intelligence and smart devices evolve, machine vision becomes central to many technologies. Yet existing vision systems struggle with the volume of visual data: processing continuous streams requires substantial power, bandwidth, storage, and compute—constraints that complicate deployment on edge devices.
The human visual system offers an instructive model: instead of processing every detail, biological vision filters and compresses input, delivering efficient perception with minimal energy. Neuromorphic computing—systems inspired by neural structures—aims to reproduce those advantages. Two major challenges remain: matching human-like color discrimination and eliminating reliance on external power.
A research team led by Associate Professor Takashi Ikuno in the Department of Electronic Systems Engineering at Tokyo University of Science has tackled both problems. Their paper, published in the 15th volume of Scientific Reports on May 12, 2025, presents a self-powered optoelectronic artificial synapse that combines high-resolution color recognition with on-device logic capabilities.
Co-authors Hiroaki Komatsu and Norika Hosoda contributed to the study. The device integrates two distinct dye‑sensitized solar cells that respond differently to wavelength. By converting incident light into electrical signals without external power, the synapse operates as a self-contained sensing and computing element—well suited for edge computing where low energy consumption is essential.
Extensive experiments demonstrate wavelength discrimination on the order of 10 nm across the visible spectrum—an accuracy approaching what the human eye can perceive. The device also generates bipolar responses: it delivers positive voltage under blue illumination and negative voltage under red illumination. This polarity tunability enables complex logic operations—such as AND, OR, and XOR—inside a single physical device, reducing the need for multiple discrete components.
To show real-world potential, the team used the device in a physical reservoir computing framework to classify human movements recorded under red, green, and blue lighting. With a single synaptic device, the system reached 82% accuracy when distinguishing 18 different combinations of colors and movements—performance that typically requires multiple photodiodes or larger sensor arrays in conventional setups.
Potential applications span industries. In autonomous vehicles, such devices could improve recognition of traffic signals, signs, and obstacles while lowering power draw. In healthcare, low‑power wearable sensors could monitor metrics like blood oxygenation with longer battery life. Consumer electronics, including smartphones and AR/VR headsets, could incorporate these synapses to extend runtime while retaining advanced visual recognition features.
“The results show strong potential for next‑generation optoelectronic devices that combine high-resolution color discrimination with logic functions in low‑power AI systems,” says Dr. Ikuno. He highlights applications in optical sensors for self-driving cars, low‑power biometric sensors for medical devices, and portable recognition systems.
Overall, this research represents a meaningful step toward bringing biologically inspired machine vision to edge devices—allowing them to perceive color and perform logic closer to how humans do, but with far lower energy requirements.
Funding: This work was partially supported by JST through the establishment of university fellowships for science and technology innovation (Grant Number JPMJFS2144) and JST SPRING (Grant Number JPMJSP2151).
About this visual neuroscience and AI research news
Author: Yoshimasa Iwasaki
Source: Tokyo University of Science
Contact: Yoshimasa Iwasaki – Tokyo University of Science
Image: The image is credited to Neuroscience News
Original Research: Open access.
“Polarity-tunable dye-sensitized optoelectronic artificial synapses for physical reservoir computing-based machine vision” by Takashi Ikuno et al. Scientific Reports
Abstract
Polarity-tunable dye-sensitized optoelectronic artificial synapses for physical reservoir computing-based machine vision
Conventional machine vision must handle vast streams of time-series visual data, which creates major obstacles for edge-device deployment due to bandwidth, storage, and energy limits. Inspired by biological vision, optoelectronic artificial synapses emulate synaptic responses to reduce data processing and energy demands, but matching human-like color discrimination and avoiding external power remain challenging.
This study reports a self-powered optoelectronic artificial synapse that integrates dye‑sensitized solar cells to achieve wavelength discrimination down to 10 nm. The device demonstrates synaptic responses to light pulses and produces polarity-tunable, wavelength-dependent signals. That bipolar behavior enables fine separation of inputs—supporting six-bit resolution with 64 distinct states—and allows multiple logic operations (AND, OR, XOR) within a single physical device.
Using the device as a physical reservoir computer, the researchers classified color-coded human motion under red, green, and blue illumination with 82% accuracy. These results advance optoelectronic artificial synapses toward precise, human-eye-like color discrimination in low-power machine vision systems designed for edge applications.