A Tool-Based Methodology For Long-Term Activity Monitoring
Résumé
In recent years, remarkable progress has been reported in the field of activity monitoring. However, despite significant breakthroughs in recognizing activities from sensor data, there is still a great deal more to accomplish, especially compared to fields pursuing related goals, such as computer vision and speech recognition. A key factor to move activity monitoring forward, is to enable researchers to build on each other's work more systematically via reproducible research. Besides providing sensor data, reproducibility in activity monitoring requires all aspects of a result to be available to the research community, including collection, processing and interpretation of measurements. This paper presents a tool-based methodology, dedicated to monitor the activities of daily living of older adults, that supports reproducible research. This methodology covers the key steps to defining a monitoring process of these activities, from sensor measurements to actionable activity information. These steps are uniformly described with concise and high-level rules. Additionally, to allow caregivers to monitor older adults' functional decline and to determine what assisting support is needed, our methodology includes a visualization tool, dedicated to handling user activities longitudinally. The proposed approach is validated by a set of rules dedicated to monitor activities of community-dwelling, older adults in their sensor-equipped homes. A preliminary study has been conducted to evaluate the intra-and inter-participant consistency of the results produced by our methodology, using longitudinal datasets, collected over several months. Using Signal Detection Theory, it has shown that our monitoring rules mostly produced the same interpretations as an expert in activity analysis, who manually analyzed the sensor datasets.
Mots clés
Origine : Fichiers éditeurs autorisés sur une archive ouverte
Loading...