Résumé : Determining the individual transition from the 3rd to the 4th of frailty phase of life is important for both the safety of the older person and to support the care provider. We developed an automatic monitoring system consisting of cameras and different sensors that analyze human behaviors and looks for changes in their activities by detecting the presence of people, their movements, and automatically recognizing events and Activities of Daily Living (ADLs). Assessment took place in a laboratory environment (GERHOME) comprised of four rooms (kitchen, living-room, bedroom, and bathroom). Data from 2 volunteers (64 and 85 years old) were analyzed. Precision in recognizing postures and events ranged from 62-94%, while sensitivity fell in the range of 62-87%. The system could differentiate ADL levels for the 64 and 85 year old subjects. These results are promising and merit replication and extension. Considerable work remains before the complete transition from 3rd to 4th life phase can be reliably detected. The GERHOME system is promising in this respect.