DIODEM -- A Diverse Inertial and Optical Dataset of kinEmatic chain Motion

Abstract: Real-world applicability and adoption of inertial motion tracking technology depends on the development of advanced algorithms that overcome challenges like magnetometer-free and sparse sensing, sensor-to-segment alignment, and motion artifact reduction. This publication and the accompanying dataset foster this development by providing researchers with the ability to systematically study and explore various combinations of these challenges in a well-controlled setting. The dataset comprises 46 minutes of optical and inertial data (20 reflective markers and ten, rigidly and foam-attached IMUs) of five-segment kinematic chains that feature different joint types and are moved in a large indoor tracking range such that they perform motions of various speeds and characteristics. Due to the large number of segments, as well as the different types of joints, motions, and sensor attachments, this dataset allows researchers to identify and push the boundaries of current inertial motion tracking solutions and to develop and validate novel methods for application areas ranging from biomechanics to autonomous systems.