Real-time OpenSim via IMUs for Full Body Kinematics

We present a comprehensive dataset developed as part of a study to compute real-time kinematics using a full-body wearable approach incorporating up to 12 IMUs. This dataset includes optical and inertial measurements from 22 subjects engaged in a diverse set of 9 activities: walking, running, squatting, boxing, yoga, dance, badminton, stair climbing, and seated extremity exercises. The dataset features ground truth kinematics, offline predicted kinematics, online predicted kinematics, and IMU-simulated offline predicted kinematics.

The approach leverages the tools in OpenSim to derive joint angles from IMU orientations in real time at a frequency of 20 Hz using optimization algorithms. The inverse kinematics calculation can be performed on a personal computer (PC) or on a SageMotion hub, which facilitates data exchange between IMU nodes and cellphones/PCs, enabling real-time joint angle computation in scenarios that require a larger range of motion.