Smart Physiotherapy: Monitoring Distal Carpal Kinematics Using TinyML

This work proposes an intelligent physiotherapeutic system for real-time wrist movement assessment utilizing TinyML and inertial measurement unit (IMU) sensors. The hardware setup comprises two MPU-6050 IMU sensors connected to an ESP32 microcontroller affixed to a wristband. During wrist movements, the sensors capture the accelerometer and gyroscope data. Random Forest and Decision Tree machine learning models were considered for training using a manually curated dataset, adopting TinyML techniques to predict movement correctness on the device efficiently. Timestamped predictions are transmitted to the central cloud server via Wi-Fi, enabling users to access assessments through a mobile app. This system provides an objective and efficient approach to wrist movement evaluation for physiotherapy, ensuring reliable and unbiased assessments. Additionally, an app was developed for patients' use, facilitating a user-friendly interface for tracking and assessing their physiotherapeutic movements. The dataset was recorded considering commonly encountered cases while performing wrist movements in a pragmatic scenario, with the dataset recorded from an individual acting as a volunteering patient.