Abstract
Inertial Motion Tracking (IMT) faces critical challenges including magnetometer-free sensing, sparse sensor configurations, sensor-to-segment alignment, and motion artifact compensation. Current IMT algorithms require systematic evaluation across combinations of these challenges in controlled environments with accurate ground truth data. This paper presents DIODEM-a comprehensive dataset comprising 46 minutes of synchronized optical and inertial data from five-segment Kinematic Chains (KCs). The dataset features 20 markers and ten IMUs (both rigidly and foam-attached) across two distinct kinematic configurations: an "arm" chain with hinge and spherical joints, and a "gait" chain with hinge and saddle joints. The KCs perform diverse motions including random movements at various speeds, pick-and-place tasks, and gait-like patterns. Key technical contributions include: (1) mechanically controlled setup with known kinematics, (2) systematic inclusion of motion artifacts through foam-attached IMUs, (3) diverse joint types including 1D, 2D, and 3D joints, and (4) comprehensive motion variety supporting sparse sensing scenarios. The dataset enables researchers to systematically study individual and combined IMT challenges, facilitating algorithm development for applications ranging from biomechanics to autonomous systems.