Real-Time Limb Motion Tracking with a Single IMU Sensor for Physical Therapy Exercises

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Sensor Systems

Background:

• Physical therapy commonly uses limb exercises to enhance range of motion (RoM), strength, and flexibility.
• Current motion tracking systems for physical therapy, such as camera-based or multi-IMU systems, have limitations including occlusion, lighting issues, and inconvenience.
• Accurate and convenient motion tracking is crucial for improving therapy outcomes and reducing costs in rehabilitation.

Purpose of the Study:

• To develop a novel, single 9-axis inertial measurement unit (IMU) sensor-based system for tracking limb motion during physical therapy exercises.
• To address the challenges of data noise and multi-joint motion estimation using a single sensor.
• To enable real-time, accurate, and convenient monitoring of limb movements for improved physical therapy outcomes.

Main Methods:

• Proposed a limb motion tracking system utilizing a single 9-axis IMU sensor placed on the distal end joint (e.g., wrist, ankle).
• Developed a recurrent neural network (RNN) model to estimate 3D positions of multiple limb joints (e.g., wrist, elbow) from noisy IMU data in real time.
• Validated the system's accuracy using leave-one-subject-out cross-validation for arm motion tracking.

Main Results:

• The proposed RNN model accurately estimated 3D joint positions, achieving a median error of 7.2 cm for the wrist and 7.1 cm for the elbow.
• The single-IMU system outperformed state-of-the-art approaches by over 10% in accuracy.
• The developed model is lightweight, facilitating real-time applications on mobile devices.

Conclusions:

• The novel single-IMU sensor system with an RNN model offers a highly accurate and convenient solution for limb motion tracking in physical therapy.
• This technology has significant potential for enhancing home-based physical therapy by improving exercise monitoring and RoM measurement.
• The cost-effectiveness and mobile compatibility of the system promote widespread accessibility for rehabilitation applications.