Human Motion Tracking for Action Recognition Through Wearable Sensors: an IMU Based Approach

The development and effective implementation of Industrial Collaborative Robotic applications is one of the challenges in Industry 5.0. In this area, one of the key issues is related to the development or availability of reliable systems as well as open and light software packages for tracking human-motions and, possibly, infer intentions. In this work, an approach relying on data gathered from wearable inertial sensors is developed to track human upper body motions. After the definition of a dedicated sensor placement, a kinematic model of the human upper body with 13 joints has been developed. Forward and Inverse Kinematics problem solutions have been computed considering the right part of the human body. This model has been deployed exploiting 6 IMUs attached to the following human limbs: sacrum, chest, head, right arm, right wrist and right hand. This allowed to track the rotations of the joints: backbone, neck, right shoulder, right elbow and right wrist. A real-time data processing, exploiting the orientation data of the IMUs attached to the body, has been set-up. The whole approach has been validated through a dedicated procedure and the open-source OpenSim SW was employed for comparison with a known full-body musculoskeletal model.