A Novel Method to Characterize Walking and Running Energy Expenditure

Abstract

Background: Physical activity and corresponding energy expenditure can improve health in various ways. Existing methods to directly measure energy expenditure are currently limited to laboratory settings and/or expensive instrumentation. The purpose of this study was to evaluate accuracy of energy expenditure characterization, during walking and running, using demographic data, as well as data collected via an accelerometer and novel piezoresponsive foam sensors. Methods: 30 individuals (14 females; mass = 67 ± 10 kg; height = 1.74 ± 0.08 m; age = 23 ± 3 yrs) walked and ran at five speeds (1.34, 2.23, 2.68, 3.13, and 3.58 m/s) on a force-instrumented treadmill while wearing a metabolic analyzer and standardized athletic shoes instrumented with an accelerometer, and four novel nanocomposite piezoresponsive force sensors. Various predictive models, including demographic data and data derived from the accelerometer and force sensors, were evaluated for each gait speed. Results: The predictive models varied in ability to accurately characterize energy expenditure. For walking, the most accurate model included acceleration and body weight, and resulted in an average absolute error of 0.07 ± 0.03 kcal/min. For running, the most accurate model included sensor and acceleration data, and resulted in an average absolute error of 0.45 ± 0.14 kcal/min. Conclusions: When combined with acceleration data and body weight, the novel foam sensors can be used to inexpensively and accurately measure walking and running energy expenditure. This can be done at various speeds, outside of a traditional research laboratory. These results have application within a wide range of diverse contexts.

Bayesian Projection Pursuit Regression

“…The first seven arise from real computer and lab experiments: the "Surge" dataset is described in section 3.2 of the main text. The "EEM" dataset measures Energy Expenditure per Minute, aggregated over time for individuals with various demographic characteristics walking/running at various speeds; additional inputs include data from a shoe-based analysis system worn by the subjects during data collection (Evans et al, 2018). The inputs are similar, but more detailed, for the "GRF" dataset, where the response is maximum vertical Ground Reaction Force per stance, aggregated over many stances for a variety of individuals running at various speeds (Seeley et al, 2020).…”

Physical Activity Tracking Wristbands for Use in Research With Older Adults: An Overview and Recommendations

“…Usually, commercial activity trackers contain at least an accelerometer to calculate outcome measures (Evans et al, 2018;Yang & Hsu, 2010). Differentiating usability factors are based on interaction with the devices' hardware and software.…”