TinyML Model Keeps an Eye on Your Knees' Health, Thanks to a Self-Powered Soft Sensor Wearable
Arduino Nano 33 BLE and STMicro STM32 NUCLEO F401RE work hand-in-hand to deliver effective long-term joint health monitoring.
Researchers from the University of Oxford, University College London, and Xi'an Jiaotong University have developed an "AI-enabled" piezoelectric wearable that, they say, represents a breakthrough in joint health monitoring — passing its data through an tiny machine learning (tinyML) model running on a microcontroller development board.
"Joint health is critical for musculoskeletal (MSK) conditions that are affecting approximately one-third of the global population," the researchers say of the focus of their work. "Monitoring of joint torque can offer an important pathway for the evaluation of joint health and guided intervention. However, there is no technology that can provide the precision, effectiveness, low-resource setting, and long-term wearability to simultaneously achieve both rapid and accurate joint torque measurement to enable risk assessment of joint injury and long-term monitoring of joint rehabilitation in wider environments."
That is, there was no technology with addressed all of these requirements — until the researchers developed it. The team's device is a flexible, soft, and lightweight wearable torque sensor based on boron nitride nanotubes (BNNTs) on a polydimethylsiloxane (PDMS) substrate, which uses the peizoelectric effect to both capture data about the wearer's knee movements and harvest its own power.
The secret to the system's success: a tiny machine learning (tinyML) model, trained on data captured from an Arduino Nano 33 BLE running and deployed to an STMicroelectronics STM32 NUCLEO F401RE development board via TensorFlow Lite and X-Cube-AI, which processes the sensor's piezoelectric outputs and maps them to physical characteristics including torque, angle, and loading — providing continuous monitoring instead of current point-in-time measurement approaches, without causing discomfort to the user.
"The proposed system has the potential to advance global efforts in joint health monitoring, the management of MSK conditions, rehabilitation, ageing disorders, and broader applications in personal healthcare," the team claims. "Future research will prioritize enhancing the system’s adaptability, scalability, and inclusivity."
The researcher's work has been published in the journal Nano-Micro Letters.