Motion Gait Recognition of Lower Limb Exoskeleton Based on Particle Swarm Optimization-Based Extreme Learning Machine Algorithm

A human gait recognition method based on the PSO-ELM algorithm is proposed in order to achieve coordinated movement between humans and lower limb exoskeletons. Ground reaction force (GRF) from the foot, and motion capture data (MCD) from two joints were collected through the exoskeleton device. The sample data were obtained through multiple experiments in different action scenarios, including standing still, walking on the flat, climbing up and down stairs, traveling up and down slopes, in addition to squatting down and standing up. The algorithm utilizes short-term posture data to recognize different posture movement patterns, with two advantages: (1) A user-friendly wearable device was constructed based on multi-source sensors distributed throughout the body, addressing multiple subjects with varying weights and heights, while being cost-effective and reliably and easily collecting data. (2) The PSO-ELM algorithm identifies key features of gait data, achieving a higher recognition accuracy than other advanced recognition methods, especially during arbitrary gait transition duration.