Ultra Diffusion Poser: Diffusion-Based Human Motion Tracking From Sparse Inertial Sensors and Ranging-Based Between-Sensor Distances
Researchers have developed Ultra Diffusion Poser, a new diffusion model for human motion tracking that integrates data from sparse inertial sensors and ultra-wideband (UWB) ranging. This model explicitly accounts for the geometric constraints imposed by UWB distance measurements, which were previously underutilized. By reconstructing the 3D sensor layout and using this information alongside IMU signals, the model significantly improves pose estimation accuracy, achieving up to a 22% reduction in joint position error compared to existing methods. AI
IMPACT Enhances human motion tracking accuracy by leveraging geometric constraints from UWB data, potentially improving applications in VR, robotics, and biomechanics.