Human performance analysis in natural and altered gravity conditions: an application to human movement

This thesis validates and tests portable and accessible technologies to extend human movement analysis beyond laboratory settings aiming at enhancing ecological validity and real-world applicability. First, the enhancement of a markerless pipeline for home-based infant monitoring was presented, introducing a trunk-fixed reference frame to compensate for accidental camera movements and an optimization procedure for the pose estimator training. Emerging pose estimators were also preliminary evaluated. Finally, in a case study on twins with diverging health profiles, ten literature derived movement metrics were identified as able to discriminate between the infants, supporting the early detection potential of the pipeline. Second, inertial sensors were used to assess a 10-week outdoor multisport intervention for individuals with Parkinson’s disease, which resulted in significant improvements in gait speed and stride length. Third, two platforms were developed to study motor behaviour under simulated altered-gravity: 1) a mixed reality and inertial sensors system, able to elicit gravity-dependent adaptations in throwing kinematics and accuracy under visually simulated altered gravity conditions; 2) a virtual reality and robotic arm system, able to reproduce some of the motor patterns observed in real altered-gravity conditions. Finally, a mixed reality and inertial sensor platform for gait monitoring and rehabilitation was introduced. A real-time foot-tracking algorithm using inertial sensors was developed and preliminary validated, highlighting key errors such as sensor misplacement and integration drift. Then, the Hololens 2 headset tracking accuracy was preliminary validated, showing variable performances based on movement direction and task characteristics. Overall, these findings support the use of portable and accessible technologies for ecologically valid motion analysis, while highlighting the need for improved robustness and integration.