Title Classifications of Driving Patterns Using a Supervised Learning Method for Human-Vehicle Interaction
Authors Yongho Jo (KAIST)
Dong-Soo Kwon (KAIST)
Abstract Enhancing the safety of drivers is one of important issues in a driving assistant system (DAS). In order to solve this issue, the situation information of a vehicle should be recognized. In this paper, we present Discrete Hidden Markov models (DHMMs) to classify five driving patterns which are essential features for recognizing the situation information of vehicles. A virtual vehicle simulator was developed to collect the raw data of the driving operation, and each DHMM was learned from the obtained training data. The structures of the DHMMs were experimentally selected using a cross-validation process with the results showing 90% classification accuracy.
PDF [Link]