In this paper, we propose to model the evolution of data sensors during the driving situation encountered by a driver, using the hidden Markov Model formalism. We then use this modeling to identify in real time the current driver's aim. We tested the capacity of this modeling in a first experiment where we were able to categorize with an 80% success rate the driver's actions from their initial preparatory movements. Moreover, this formalism could give us information on the driver's behavior in certain situations. First we will describe the value of using this type of modeling in the field of transport. Then we will present the methodology we used to collect and analyse data on the driver's behavior. In the last section, we will present and discuss our results. (Author/publisher)
Abstract