Presenting a Data Imputation Concept to Support the Continuous Assessment of Human Vital Data and Activities

Data acquisition of mobile tracking devices often suffers from invalid and non-continuous input data streams. This issue especially occurs with current wearables tracking the user's activity and vital data. Typical reasons include the short battery life and the fact that the body-worn tracking device may be doffed. Other reasons, such as technical issues, can corrupt the data and render it unusable. In this paper, we introduce a data imputation concept which complements and thus fixes incomplete datasets by using a new merging approach that is particularly suitable for assessing activities and vital data. Our technique enables the dataset to become coherent and comprehensive so that it is ready for further analysis. In contrast to previous approaches, our technique enables the controlled creation of continuous data sets that also contain information on the level of uncertainty for possible reconversions, approximations, or later analysis.