Researchers at Dongguk University in South Korea have developed a versatile, battery-free digital machine that may harvest power from human motion and use it to energy neuromorphic sensing and studying capabilities.
The expertise may finally be utilized in wearable health-monitoring methods, digital pores and skin, sensible prosthetics, human-machine interfaces and clever motion-monitoring units with out requiring standard batteries or an exterior energy provide.
The system reminds us of the Seiko Kinetic watch (primary picture), first launched in 1988, which additionally harvested power from the motion of the wearer.
Led by Professor Sejoon Lee of Dongguk College’s Division of System Semiconductor, the analysis workforce mixed a triboelectric nanogenerator, or TENG, with a versatile graphene-channel ion-gel-gated transistor.
The TENG converts mechanical stimuli akin to physique motion, contact or vibration into electrical indicators. These indicators straight function the unreal synaptic capabilities of the machine, which means it could sense motion and course of info and not using a separate supply {of electrical} energy.
Lee says: “In human tactile notion mechanoreceptors sense even minute mechanical disturbances and convert them into neural spikes.
“To duplicate this course of electronically, we built-in a triboelectric nanogenerator with a g-IGT that converts mechanical stimuli into electrical indicators that straight regulate synthetic synaptic conduct with out requiring exterior energy.”
The researchers demonstrated a number of types of synthetic reminiscence, starting from sensory reminiscence lasting round 70 milliseconds to short-term reminiscence lasting 0.2-0.45 seconds.
Repeated stimulation may additionally transfer the system towards a longer-term reminiscence state lasting greater than two seconds.
The machine was moreover examined utilizing a man-made neural community designed to acknowledge six human actions, together with strolling, sitting, standing and climbing stairs.
Utilizing the experimentally measured conduct of the machine, the system achieved 88.05 p.c accuracy in classifying the actions whereas the versatile machine was beneath bending.
The researchers say the expertise may finally allow wearable AI methods combining sensing, reminiscence, studying and knowledge processing whereas lowering or eliminating their dependence on batteries.
Lee says: “Our analysis may contribute to a brand new era of wearable synthetic intelligence methods that function with minimal reliance on batteries or exterior computing assets.
“Extra broadly, our work factors towards self-powered neuromorphic electronics with built-in sensing, reminiscence, studying, and knowledge processing in a single versatile platform.”
The analysis was printed in Advanced Materials in July 2026.
