Analysis of Biomedical Longitudinal Multisensor Data: Extracting Interpretable Features by Context

Longitudinal quotidian acquisition of personal data over weeks and months has been increasingly facilitated in the last few years. Smartwatches and smartphones serve as platforms containing a variety of sensors, able to capture a multitude of individual biomedical and behavioral aspects. This development enables new analytic pathways in medicine and health care. Due to data of this type being increasingly ascertainable, efficient analyses become crucial. For instance, determining meaningful, individualized patterns from such multimodal, longitudinal time series can be very time and resource consuming. To this end, interpretable and robust parameters need to be extracted. In this paper, we explore a general approach to context based parameter estimation and illustrate its ability to be utilized for determining individualized and interpretable biosignal and behavioral patterns.