Age-sensitive differences in single and dual walking tasks from footprint floor sensor data

This 2017 IEEE SENSORS paper uses a footprint floor sensor and support vector machines to compare single-task and dual-task walking across age groups.

“Age-sensitive differences in single and dual walking tasks from footprint floor sensor data” was presented at the 2017 IEEE SENSORS conference in Glasgow by Omar Costilla-Reyes, Patricia Scully, and Krikor B. Ozanyan, and published in 2017. It starts from the observation that gait can provide insight into decline in executive function, the set of mental skills used to plan, focus attention, and switch between tasks. The paper presents experiments and a methodology for analyzing age-sensitive differences in walking patterns using the University of Manchester’s footprint-imaging floor sensor, with volunteers from three age groups: a young adult, an adult, and a mature adult.

A dual task asks a person to walk while also performing a cognitive task. Because both activities draw on attention, the way someone walks under this extra mental load can differ from the way they walk normally. Cognitive-load tasks were added so that their effect on spatio-temporal walking patterns, meaning how footsteps are arranged in space and over time, was captured in the recordings. Support vector machine classifiers, which learn a boundary that separates classes of examples, were applied to the raw gait sensor data in three experiments: normal versus fast walking, normal walking versus a dual task, and a three-way classification of normal walk, fast walk, and dual task.

For single tasks the paper reports a high F-score, a measure that combines how many detected cases were correct with how many true cases were found, and competitive performance on the finer-grained variability of the dual-task experiments. Measuring walking under mental load matters because gait is an observable correlate of attention and executive function, which is why a floor sensor that captures every footstep is well suited to healthcare research. The paper set out the experimental design that was taken further in a study of healthy adults published in 2020, “Age-Related Differences in Healthy Adults Walking Patterns Under a Cognitive Task With Deep Neural Networks,” which applied deep neural networks to the same question.

The paper is the third manuscript of Dr. Costilla Reyes’s PhD thesis. At the same conference he took part in a young-researcher panel discussion on PhD project management, time management, high-quality research, and motivation.