Identifying and Evaluating the Effects of User Scenarios on the Data Integrity of Wearable Devices

Wearable devices are able to sense, collect, and upload various physiological data. With the technological improvements and progress in power, computation complexity, and size for wearable devices, their application has rapidly expanded in various use scenarios in healthcare. The main challenge of utilizing healthcare data through wearable devices is data integrity. The accurate assessment of data integrity necessitates a comprehensive understanding of the contextual information of wearable usages in free-living conditions, such as user scenarios. To examine the effects of user scenarios on data integrity and validate the assessment method, we conducted a human-subject experiment. We collected raw acceleration and heart rate data from participants using the Apple Watch in the lab under the selected user scenarios. We implemented an anomaly detection method based on an ensemble of neural networks. Then, we associated those data compromises with user scenarios and tested if the selected scenarios influenced the data. This chapter demonstrates the ability of the proposed method to assess data integrity under different user scenarios and contributes to future work that quantifies the influence of user scenarios as potential root causes.