This Paper-Based Brain-Inspired Sensor Could Provide Energy-Efficient AI-Enhanced Health Monitoring

Capable of recognizing handwritten digits, this smart paper sensor could be the future of healthcare wearables.

Researchers from the Tokyo University of Science have developed a paper-based smart sensor inspired by the human brain that, they say, could pave the way for energy-efficient health monitoring devices with stand-alone artificial intelligence (AI) capabilities — and which can be disposed of by incineration when no longer required.

"A paper-based optoelectronic synaptic device composed of nanocellulose and ZnO [zinc oxide] was developed for realizing physical reservoir computing [PRC]," Takashi Ikuno, associate professor and corresponding author on the paper detailing the team's work. "This device exhibits synaptic behavior and cognitive tasks at a suitable timescale for health monitoring. This study highlights the potential of embedding semiconductor nanoparticles in flexible CNF [cellulose nanofiber] films for use as flexible synaptic devices for PRC."

Physical reservoir computing is designed to mimic the operation of the human brain using controllable physical phenomena, allowing it to perform machine learning and artificial intelligence tasks with improved efficiency over traditional electronic computing. In the case of the device built by the team, the phenomenon is light — taken as an input by a device made from nanocellulose and zinc oxide nanoparticles with gold electrodes.

Despite being, effectively, paper and a little bit of metal, the device in question has proven capable of handling machine learning tasks: in experiments, prototype devices were able to carry out recognition of handwritten digits from the MNIST dataset with an 88 percent accuracy, even after being subjected to 1,000 cycles of bending. "Furthermore," the team notes, "the device burn[s] in a few seconds, much like regular office paper, demonstrating its disposability."

The flexibility of the device, the ease with which it can be safely disposed of, and its ability to perform machine learning tasks and to exhibit a short-term memory effect mean it could be ideally suited to future healthcare wearables — delivering low power usage, comfort, and the ability to provide AI-enhanced real-time monitoring of biological signals.

The team's work has been published in the journal Advanced Electronic Materials under open-access terms.

Gareth Halfacree
Freelance journalist, technical author, hacker, tinkerer, erstwhile sysadmin. For hire:
Latest articles
Sponsored articles
Related articles
Latest articles
Read more
Related articles