The actual demographic trend predicts a significant increasing percentage of elderly people in the German society until 2050. According to this trend, it must be assumed that the number of elderly persons who wants to live self-sufficiently in their habi tual environment will rise as well. In this paper a human centered assistance system is tightly descri bed and proposed as a resolve for the coherent challenges of this structural, societal changes which relate to home care and geriatrics; thereby the focus of consideration lies on the statistical analysis of gath ered process data in order to derive an intelligent emergency detection based on probabilistic modeling. Using room automation states and events to design and train Hidden Markov Models for position tracking, complemented with the stochastic evaluation of telemedical devices a nd contactless sensor networks, the ma in issue is to achieve a solid, robust situation recognition mechanism. Due to the knowledge of the hidd en user states or i.e. situations, this approach offers the opportunity to detect emer gency situations instantl y and to prevent harmful aftermath for the user through interventions like emerge ncy calls. This pa per illustrates the architectural concept of the assistance system with its implemented techniques and demonstrates the approximation of user situations through a clear example. Simulation re sults are discussed and finally, a first conclusion with an outlook on further progress of research closures this article.