A study design for physical activity reference data collection using GPS and accelerometer

Hoda Allahbakhshi, Haosheng Huang, Robert Weibel
2018
A physically active lifestyle is a key component of promoting health and well-being particularly for healthy ageing. Most sensor-based studies are focused on measuring the level (or intensity) of physical activity and use data collected using a specific study protocol under a controlled laboratory condition and are thus hardly comparable to each other and difficult to use as training or validation data for reallife studies. Therefore, it is important to have available a reference dataset for
more » ... ence dataset for physical activity type classification especially in real-life environments. The main aim of this study is to provide a study design for collecting a reference dataset that can maximize both internal and ecological validity of measuring physical activity types. To that end, we designed study protocols in three different conditions, namely: laboratory/controlled, semi-structured and real life. To collect data, a sample of 40 healthy participants (30 younger adults and 10 older adults) will participate to perform activities including: lying, sitting, standing, walking on level ground, running, cycling, walking uphill, walking downhill, walking downstairs and walking upstairs both indoors and outdoors. The activity walking on level ground will be performed at three different speeds. Additionally, GPS and GIS methods (e.g. information about slope or dominant land use) will be used to enrich the detailed information about accelerometry-based activity types and to provide the environmental information of the place where the activities will take place. The proposed reference dataset can be useful for future validation and comparison studies and for the development of new physical activity type classifications algorithms particularly under real-life conditions. Abstract A physically active lifestyle is a key component of promoting health and well-being particularly for healthy ageing. Most sensor-based studies are focused on measuring the level (or intensity) of physical activity and use data collected using a specific study protocol under a controlled laboratory condition and are thus hardly comparable to each other and difficult to use as training or validation data for real-life studies. Therefore, it is important to have available a reference dataset for physical activity type classification especially in real-life environments. The main aim of this study is to provide a study design for collecting a reference dataset that can maximize both internal and ecological validity of measuring physical activity types. To that end, we designed study protocols in three different conditions, namely: laboratory/controlled, semi-structured and real life. To collect data, a sample of 40 healthy participants (30 younger adults and 10 older adults) will participate to perform activities including: lying, sitting, standing, walking on level ground, running, cycling, walking uphill, walking downhill, walking downstairs and walking upstairs both indoors and outdoors. The activity walking on level ground will be performed at three different speeds. Additionally, GPS and GIS methods (e.g. information about slope or dominant land use) will be used to enrich the detailed information about accelerometry-based activity types and to provide the environmental information of the place where the activities will take place. The proposed reference dataset can be useful for future validation and comparison studies and for the development of new physical activity type classifications algorithms particularly under real-life conditions.
doi:10.5167/uzh-161051 fatcat:y2ome2o6lraxvjlt5og4dvcuoy