Advanced Signal Processing Methods in Walking and Body-Posture Detection in Observational Studies

• Activity Number: 571 - Statistical Signal Processing Applied to Physical Activity Research
• Type: Topic Contributed
• Date/Time: Wednesday, August 1, 2018 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Marcin Straczkiewicz* and Jacek K Urbanek and Vadim Zipunnikov and Nancy Glynn and Tamara Harris and Ciprian Crainiceanu and Jaroslaw Harezlak
• Companies: School of Public Health-Bloomington, Indiana University and Johns Hopkins University and Johns Hopkins Bloomberg School of Public Health and University of Pittsburgh Graduate School of Public Health and National Institute on Aging and Johns Hopkins University and Indiana University Bloomington
• Keywords: Raw Accelerometry Data; Wearable Computing; Physical Activity; Activity Recognition

Abstract

Accelerometers are frequently used to measure physical activity in large observational studies (e.g. NHANES, UK Biobank, WHI, etc.) because they are convenient to wear, cheap and provide objective and reproducible proxy measurements of physical activity. The outcomes of activity measurements are often provided as 24-hour activity cycle summaries, such as activity counts, vector magnitude or number of steps. However, collected data in its raw format contain a lot more information that is presently investigated. During the talk we will discuss the structure of raw accelerometry signals and provide an intuitive link between the measurements and human movements. We will describe selected distinctive features of such data that help in more objective quantification of specific types of daily activities, such as sedentary and upright body posture or walking. The proposed methodology will be validated on a population of N=45 elder participants of the Development Epidemiological Cohort Study, a part the AREA project. During 7 days of free-living all subjects were equipped with sensors on wrists and hip, while the gold standard was provided by the activPAL sensor placed on a thigh.