A physical activity reference data-set recorded from older adults using body-worn inertial sensors and video technology?The ADAPT study data-set

摘要: Physical activity monitoring algorithms are often developed using conditions that do not represent real-life activities, the target population, or labelled to a high enough resolution capture true detail of human movement. We have designed semi-structured supervised laboratory-based protocol and an unsupervised free-living recorded 20 older adults performing both protocols while wearing up 12 body-worn sensors. Subjects' movements were synchronised cameras (≥25 fps), deployed in laboratory environment in-lab portion camera for out-of-lab activities. Video labelling subjects' was performed by five raters 11 different category labels. The overall level agreement (percentage >90.05%, Cohen's Kappa, corrected kappa, Krippendorff's alpha Fleiss' kappa >0.86). A total 43.92 h activities recorded, including 9.52 34.41 88.37% 152.01% planned transitions during scenarios, respectively. This study has produced most detailed dataset date inertial sensor data, with frame-rate fps) video data from living independently. is suitable validation existing classification systems development new algorithms.