Wearable consumer physical activity tracking devices1 ssuch as Fitbit®) which generate logs of step count data for users offer a simple and relatively inexpensive way to monitor ambulatory behaviour. However, for researchers, the time granularity with which incremental step count data can typically be obtained using these devices, limits their usefulness particularly for short duration events. This prevents wider use of consumer wearables for non-fitness related purposes, such as using step counts to describe specific activities of daily living. This paper describes a method to overcome this limitation by classifying movements based on patterns in the data. We propose a simple model for analysing adjacent step count data values to achieve this using pilot data on physical activity in an office setting.
|Publication status||Published - 29 Jan 2018|
|Event||2018 Australasian Computer Science Week Multiconference, ACSW 2018 - Brisbane, Australia|
Duration: 30 Jan 2018 → 2 Feb 2018
|Conference||2018 Australasian Computer Science Week Multiconference, ACSW 2018|
|Period||30/01/18 → 2/02/18|
- ACM proceedings
- Text tagging