Pattern activities identification in the framework of medical nursing home using infrared sensors
Résumé
This work propose a remote healthcare monitoring system designed to derive elderly's activities of daily living and distress situation in Nursing Home environments through a set of pyroelectric infrared sensor placed in the rooms. We aim to infer from the nursing home environment information as the exits and entries in the room, presence of care-givers and activities as night rounds, sleeping, eating, care-giver intervention. Thank to the sensors emplacement in the present case is possible to extract patterns regarding paths performed inside the room. Based on these patterns, we propose in a first step, a finite state model that represents the behavior of only one person moving inside the room. Through accomplishing of this model, we can derive some information as resident tracking, presence of care-givers, entries and exits. In previous work we have used the Dempster-Shafer theory (DST) based on static and dynamic model called Evidential Networks [11] in order to take into account the imperfect information (uncertainty, precision and conflict) involved by the sensors data acquisition and also providing a framework allowing the fusion of multi-sensors data yielding richer and reliable information. To this aim, the imperfect information representation by DST was considered in the proposed model. Thus, degradation of the sensor signals can be considered, becoming our model most reliable. In a second step, DST can also be used to perform the data fusion of the infrared sensors with other sensors introduced in the environment [11]. The proposed algorithms were evaluated in the framework of real data recording made in a medical nursing home in France in the framework of EMonitor'age project