Makes gait recognition fast enough for exoskeleton control by keeping accuracy high on short data windows. Feature selection, including recursive feature elimination, is combined with lightweight sequential models such as TCNs and GRUs to cut latency without losing accuracy.
Publications
1
A. Majumder, S. Nisar
30th Int. Symposium on Artificial Life and Robotics (AROB) 2025
2
Achieving Real-Time Gait Recognition for Lower-Limb Robotic Exoskeletons Through Feature Selection and Model Evaluation at Reduced Window Sizes Thesis
A. Majumder
Bachelor's thesis (Lab Project), Kyoto University of Advanced Science, Kyoto, Japan · Advisor: S. Nisar
