The field of human motion analysis is rapidly evolving, with a focus on developing innovative systems for real-time feedback and evaluation. Recent developments have centered around creating more accurate and robust methods for assessing human movement, including the use of inertial measurement units (IMUs) and machine learning algorithms. These advancements have the potential to revolutionize various applications, such as fitness training, rehabilitation, and sports performance monitoring. Notably, self-supervised learning techniques have shown promise in estimating joint kinematics and kinetics directly from IMU data, eliminating the need for ground truth labels. Furthermore, implicit modeling of human locomotion has enabled more accurate gait phase estimation, which is crucial for adapting exoskeletons to individual gait variations. The use of garment-aware diffusion models has also improved motion capture from loose and sparse inertial sensors, opening up new possibilities for human-robot interaction. Noteworthy papers include:

• A real-time feedback system for assessing isometric poses, which introduces a novel three-part metric for evaluation and has the potential to enhance personalized exercise training systems.
• SSPINNpose, a self-supervised physics-informed neural network that estimates joint kinematics and kinetics directly from IMU data, demonstrating robustness across sparse sensor configurations.