Population aging combined to the desire to age in place for dependent persons brings a need for health at home systems. In this work, the modeling of Activities of Daily Living (ADL) and their recognition was studied in order to validate experimentally previous work conducted at LURPA Laboratory and extend results. An approach was proposed to evaluate the models' performances according to an Activity Recognition (AR) method developed previously. It was then applied to a case study, conducting to question the relevance of stochastic models compared to deterministic models, according to the desired result of the AR.