Abstract
Given the rich amount of information provided by accelerometers and gyroscopes in capturing linear and rotational movement, sensors have been used in a wide range of healthmonitoring related applications, including gait change monitoring. This study focuses on posture detection in a proof-of-concept wearable wireless sensor system that alerts users in real time when poor posture is detected. The Seeed Studio XIAO nRF52840 (Sense) development boards, which feature a built-in Inertial Measurement Unit (IMU) and Bluetooth module, were used as our wearable sensor node. The Edge Impulse platform was selected to train a custom-defined neural network for motion classification using accelerometer and gyroscope data. The custom-defined motion includes lower-body-generated movements, such as leg crossing, that create tension in the lower extremities and imbalance in the pelvis and spinal cord, leading to upper-body strains in the neck and shoulders. Preliminary results indicate a high accuracy rate in target motion (leg crossing) detection, with a 10 -second delay between motion detection and Bluetooth communication. For ease of use and a less intrusive user experience, two sensor placement locations: an ankle (primary location) and a knee (secondary location) were tested. By leveraging edge computing capabilities, this study accurately detected the target posture and transmitted it from two peripheral nodes to a central node via Bluetooth 5.0 wireless communication.