Driver behavior analysis and route recognition by Hidden Markov Models

Abstract

In this investigation, driver behavior signals are modeled using Hidden Markov Models (HMM) in two different and complementary approaches. The first approach considers isolated maneuver recognition with model concatenation to construct a generic route (bottom-to-top), whereas the second approach models the entire route as a dasiaphrasepsila and refines the HMM to discover maneuvers and parses the route using finer discovered maneuvers (top-to-bottom). By applying these two approaches, a hierarchical framework to model driver behavior signals is proposed. It is believed that using the proposed approach, driver identification and distraction detection problems can be addressed in a more systematic and mathematically sound manner. We believe that this framework and the initial results will encourage more investigations into driver behavior signal analysis and related safety systems employing a partitioned sub-module strategy.