Loaded and unloaded foot movement differentiation using chest mounted accelerometer signatures

Abstract

Heavy loads often subject foot soldiers and first-responders to increased risk musculoskeletal injury (MSI). Identifying excessive loads in real-time could help identify when soldiers are at greater risk of MSI. Using Principal Component Analysis (PCA) we derived a loaded (>35 kg) versus unloaded Naïve Bayesian classification model from 22 male Soldiers (age 20 ± 3.5 yrs, height 1.76 ± 0.09 m and weight 83 ± 13 kg). Using seven-fold cross validation we demonstrated that using only one feature our model accurately classifies heavily loaded versus unloaded over 90% of the time. This technique lends itself to use in real time accelerometry sensors and shows promise for more complex gait analysis.