In this work, we propose an efficient health-care monitoring system for the daily home activity of persons. We intend to combine a good optimization of the resources (e.g. network and energy) and an automatic evaluation of the person’s dependency while ensuring a high accuracy for detecting unusual behaviors. The proposed system considers the person’s context and predicts the health condition based on the usual behavior and energy consumption for each daily activity. The proposed system requires a minimum set of sensed data with short training periods for predicting the person’s behavior changes.