No abstract

Abstract-Walking speed is an important determinant of energy expenditure. We present the use of Gaussian Processbased Regression (GPR), a non-linear, non-parametric regression framework to estimate walking speed using data obtained from a single on-body sensor worn on the right hip. We compare the performance of GPR with Bayesian Linear Regression (BLR) and Least Squares Regression (LSR) in estimating treadmill walking speeds. We also examine whether using gyroscopes to augment accelerometry data can improve prediction accuracy. GPR shows a lower average RMS prediction error when compared to BLR and LSR across all subjects. Per subject, GPR has significantly lower RMS prediction error than LSR and BLR (p<0.05) with increasing training data. The addition of tri-axial gyroscopes as inputs reduces RMS prediction error (p<0.05 per subject) when compared to using only acclerometers. We also study the effect of using treadmill walking data to predict overground walking speeds and that of combining data from more than one person to predict overground walking speed. A strong linear correlation exists (rX,Y = .8861) between overground walking speeds predicted from treadmill data and ground truth walking speed measured. Combining treadmill data from multiple subjects with similar height characteristics improved the prediction capability of GPR for overground walking speeds as measured by correllation between ground truth and GP-predicted values (rX,Y = .8204 with combined data).

BackgroundThe use of embedded smartphone sensors offers opportunities to measure physical activity (PA) and human movement. Big data—which includes billions of digital traces—offers scientists a new lens to examine PA in fine-grained detail and allows us to track people’s geocoded movement patterns to determine their interaction with the environment.ObjectiveThe objective of this study was to examine the validity of the Movn smartphone app (Moving Analytics) for collecting PA and human movement data.MethodsThe criterion and convergent validity of the Movn smartphone app for estimating energy expenditure (EE) were assessed in both laboratory and free-living settings, compared with indirect calorimetry (criterion reference) and a stand-alone accelerometer that is commonly used in PA research (GT1m, ActiGraph Corp, convergent reference). A supporting cross-validation study assessed the consistency of activity data when collected across different smartphone devices. Global positioning system (GPS) and accelerometer data were integrated with geographical information software to demonstrate the feasibility of geospatial analysis of human movement.ResultsA total of 21 participants contributed to linear regression analysis to estimate EE from Movn activity counts (standard error of estimation [SEE]=1.94 kcal/min). The equation was cross-validated in an independent sample (N=42, SEE=1.10 kcal/min). During laboratory-based treadmill exercise, EE from Movn was comparable to calorimetry (bias=0.36 [−0.07 to 0.78] kcal/min, t82=1.66, P=.10) but overestimated as compared with the ActiGraph accelerometer (bias=0.93 [0.58-1.29] kcal/min, t89=5.27, P<.001). The absolute magnitude of criterion biases increased as a function of locomotive speed (F1,4=7.54, P<.001) but was relatively consistent for the convergent comparison (F1,4=1.26, P<.29). Furthermore, 95% limits of agreement were consistent for criterion and convergent biases, and EE from Movn was strongly correlated with both reference measures (criterion r=.91, convergent r=.92, both P<.001). Movn overestimated EE during free-living activities (bias=1.00 [0.98-1.02] kcal/min, t6123=101.49, P<.001), and biases were larger during high-intensity activities (F3,6120=1550.51, P<.001). In addition, 95% limits of agreement for convergent biases were heterogeneous across free-living activity intensity levels, but Movn and ActiGraph measures were strongly correlated (r=.87, P<.001). Integration of GPS and accelerometer data within a geographic information system (GIS) enabled creation of individual temporospatial maps.ConclusionsThe Movn smartphone app can provide valid passive measurement of EE and can enrich these data with contextualizing temporospatial information. Although enhanced understanding of geographic and temporal variation in human movement patterns could inform intervention development, it also presents challenges for data processing and analytics.