Japan: Researchers develop artificial synapses with human-like color vision

A team from Tokyo University of Science has developed an artificial synapse inspired by the human eye that can distinguish colors with a precision of 10 nanometers and does not require an external power supply. The energy for signal processing is obtained directly from the light that is already used for color recognition.

Machine vision presents developers of edge computing applications with major challenges: Processing extensive image data requires considerable computing and storage capacities as well as a constant power supply. This limits its use in areas such as autonomous vehicles, drones or smartphones, where energy efficiency is crucial. In contrast to conventional machine vision systems, which have to capture and process every detail, the human brain filters information selectively, enabling more efficient visual processing with minimal energy consumption.

As the Tokyo University of Science reports, a research team led by Takashi Ikuno has now developed a promising approach. In their study recently published in "Scientific Reports", the scientists describe a self-powered optoelectronic synapse that can distinguish colors with a precision approaching that of the human eye.

The developed device combines two differently color-sensitive dye-sensitized solar cells (DSCs) in a novel configuration. The special feature of the system lies in its bipolar reaction to different colors of light: Blue light generates a positive voltage, while red light leads to a negative voltage. Green light, which is in the middle wavelength range, produces a voltage close to zero. This property enables the device to distinguish between different colors without additional filters or sensors.

"We believe that this technology will contribute to the realization of low-power machine vision systems with color discrimination capabilities close to those of the human eye, with applications in optical sensors for self-driving cars, low-power biometric sensors for medical purposes, and wearable recognition devices," explains Ikuno.

Logical operations with a single device

The researchers demonstrated that their artificial synapse can perform various logical operations such as AND, OR and even the more complex XOR – functions that would normally require multiple conventional devices. This works because the system can respond to both light intensity and wavelength changes.

In a practical application example, the scientists used their development in a Physical Reservoir Computing (PRC) system to classify various human movements recorded in red, green and blue.

The system achieved 82 percent accuracy in classifying 18 different combinations of colors and movements – using a single device instead of the multiple photodiodes required in conventional systems.

Functionality based on dye-sensitized solar cells

The team took advantage of the unique properties of dye solar cells sensitized with two different dyes (SQ2 and D131). These dyes absorb light of different wavelengths and thus generate different photovoltages. The output voltage of the device results from the sum of these individual voltages, whereby the polarity depends on the dominant wavelength.

The scientists were able to show that their device has synapse-like properties, which are crucial for processing time-based data. In tests with successive light pulses, the system showed a remarkably broad Paired-Pulse Facilitation (PPF) index, ranging from -3776 to 8075 – significantly greater than conventional artificial synapses, which typically reach values between 100 and 200. The index describes the increased response of a synapse to two stimuli in rapid succession, with negative values indicating inhibition and positive values indicating reinforcement.

According to the researchers, this property enabled the device to distinguish up to six bits of input patterns, which surpasses the classification capability of standard dye-sensitized solar cells and makes it suitable for more complex recognition tasks.

(mack)