Autonomous detection of different walking tasks using end point foot trajectory vertical displacement data

Identifying different activities during walking is a key requirement for ubiquitous gait monitoring, particularly when engineering new falls prevention solutions. In this study, 5 healthy young individuals (aged 26 ± 2 years old) completed 6 different tasks (a) walking with preferred walking speed (PWS), (b) walking with 10 % increment in the PWS, (c) walking while holding a glass of water at self selected walking speed (PWSW), (d) walking normally without the glass of water at the same speed as in condition c (PWS W), (e) walking while wearing a pair of occlusion goggles at a different self selected speed (PWSG), and (f) walking normally, without the occlusion goggles at the same walking speed as in condition e (PWSG). Each participant carried out 5 minutes of walking for each condition on a motorized treadmill. Toe displacement data was collected using highly accurate 3D motion capture system. The standard statistical analysis shows a noticeable difference in gait kinematics collected at different walking speeds (a, d, and f). However the differences are insignificant for the conditions which were carried out at the same walking speed, though multitasking was involved (c vs. d and e vs. f). We propose an intelligent automatic gait classification system for identifying different walking activities at the same walking speed. This brings insight to gait variability due to different everyday activities and the results demonstrated that the advanced classifier was able to detect subtle variations, which were not significant in basic statistical analysis.