Traffic vehicle behavior prediction using Hidden Markov models

Abstract

This paper intends to focus on vehicles steering by analyzing the trajectory. The target values we need to measure are the vehicle's speed and turning angle. These two values are needed to be quantified to certain levels. To create the Hidden Markov Model, HMM learning algorithm and the two values above are used. In HMM, turning left, going straight and turning right are the hidden states and data from the video are used to compute the parameters. HMM can be used to analyze vehicles' driving and to predicate the probable steering in time. The experimental results show that in the case of getting good vehicle trajectory, it is pretty suitable to use HMM to predicate vehicle behavior. © 2012 Springer-Verlag.