Context-Dependent Variations in Gait and Head Motion During Real-World Walking

In recent years, gait analysis has increasingly shifted its focus toward real-world observations, which allow for a more comprehensive evaluation of individual motor performance. However, the influence of daily-life activities on such performance, and the consequent cognitive load, are still scarcely investigated. To address this gap, this study aims to investigate differences in gait parameters based on the activities performed in real-world conditions. Considering that navigation in real-world environments requires significant exploratory trunk and head movements to manage multiple concurrent visual and auditory stimuli, this work also includes the quantitative observation of head, trunk, and lower back movements. A total of 11 young adults (60% females, 20-31 years) wore a multi-sensor system, including pressure insoles, distance sensors and 5 magneto-inertial measurement units positioned on the head, chest, lower back, and feet (100 Hz), along with smart glasses for video recording. Data from the feet and lower-back sensors were used to segment strides. Video data was used to identify and annotate six real-world walking situations: 1) road crossing, 2) texting, 3) phone calls, 4-5) navigation through fixed and moving obstacles and 6) visual search within a supermarket. Each subject completed four straight walking trials in a lab setting and a 2.5-hour real-world session. Walking parameters along with head, trunk and lower back parameters were estimated on a stride-by-stride basis, covering five characteristic domains: i) Amplitude, ii) Attenuation, iii) Symmetry, iv) Smoothness and v) Regularity. A comprehensive dataset was built from the collected data, associating information on concurrent activities to quantitative information on walking and trunk motion parameters on a stride-by-stride basis. For each of the six annotated types of activity, the computed stride-by-stride parameters were averaged per subject. Real-world situations were compared with each other and with lab walking for all parameters with a one-way repeated measures analysis of variance. Analysis showed that statistically significant differences are found in all the evaluated domains except for Smoothness measures. Gait parameters (e.g., stride duration) and amplitude measures (e.g., relative yaw angle between head and chest) showed the greatest number of significant differences among activities. For instance, stride duration distributions during visual search (median: 1.378 s, inter-quartile range: 0.268 s ) significantly differed from texting (median: 1.029 s, inter-quartile range: 0.094 s), while yaw angle distributions during road crossing (median: 20.762°, inter-quartile range: 11.612°) differed from straight walking (median: 9.925°, inter-quartile range: 1.168°). These findings demonstrate that gait and trunk and head motion parameters vary substantially depending on the specific activity performed in real-world contexts. As such, aggregating movement data across heterogeneous daily-life situations risks obscuring critical behavioral adaptations. This underscores the importance of incorporating contextual information into real-world gait analysis to ensure accurate, ecologically valid interpretation of motor performance.