ABSTRACT
The field of personal informatics has become progressively mainstream recently with the rise of groups such as Quantified Self and the introduction of commercially available user-friendly fitness trackers and visualization software. The users of personal informatics technologies monitor aspects of their daily lives and habits for the 304purpose of self-reflection. For example, by recording workout habits, sleep patterns, and medicine intake, a person can generate concrete health-related statistics that can drive personal change toward healthier habits. We propose a solution that seeks to offer the same type of self-reflective statistics but also supply more detailed wellness indicators for caregivers and medical professionals. Our target population is the rapidly aging segment of many developed countries, who have also been the impetus for much research in the area of assistive technologies for older adults. We intend to monitor users for extended durations in their homes in order to extract and track various wellness indicators and behavioral patterns. These indicators and patterns of behaviors can then be tracked over time to look for trends and sudden changes that can indicate incipient health problems. Our processing framework is based on motion sensor data as a privacy-preserving alternative to camera data while still extracting similar observations of relevant health-related indicators. We will derive from these motion sensor data parameters such as the user’s position and trajectory, wellness indicators such as sedentary motion periods and sleep patterns, and social indicators such as guest visitations and out-of-home activities. Each of these indicators can eventually be further processed to look for anomalous behavior in the form of outliers, which can be used in conjunction with other indicators to look for potential correlations that may indicate causes of the anomalous behavior. We intend to expand our work to include data from more sensor types, including cameras and smartphones to evaluate the potential increased utility compared to a system that uses only motion sensors. In the following sections, we present our initial results in extracting indicators that we believe to be important in tracking a person’s health and well-being as well as present the various mechanisms we use to process motion sensor data.
