Multi-sensors acquisition, data fusion, knowledge mining and alarm triggering in health smart homes for elderly people

We deal in this paper with the concept of health smart home (HSH) designed to follow dependent people at home in order to avoid the hospitalisation, limiting hospital sojourns to short acute care or fast specific diagnostic investigations. For elderly people the project of such a HSH has been called AISLE (Apartment with Intelligent Sensors for Longevity Effectiveness). For this purpose, system having three levels of automatic measuring (1) the circadian activity, (2) the vegetative state, and (3) some state variables specific of certain organs involved in precise diseases, has been developed within the framework of a 'Health Integrated Smart Home Information System' (HIS2). HIS2 is an experimental platform for technologic development and clinical evaluation, in order to ensure the medical security and quality of life for patients who need home based medical monitoring. Location sensors are placed in each room of the HIS2, allowing the monitoring of patient's successive daily activity phases within the patient's home environment. We proceed with a sampling in an hourly schedule to detect weak variations of the nycthemeral rhythms. Based on numerous measurements, we establish a mean value with confidence limits of activity variables in normal behaviour permitting to detect for example a sudden abnormal event (like a fall) as well as a chronic pathologic activity (like a pollakiuria), allowing us to define a canonical domain within which the patient's activity is qualified to be 'predictable'. Alerts are set off if the patient's activity deviates from a predictable canonical domain. Moreover, we can follow the cardio-respiratory state by measuring the intensity of the respiratory sinusal arrhythmia in order to quantify the integrity of the bulbar vegetative system, and we finally propose to carefully watch abnormal symptoms like arterial pressure or presence of plasma proteins in the expired air flow for early detecting respectively hypertension or pulmonary oedema.