DISCRIMINATING RELATIONSHIP OF DIFFERENT DRIVER STATES AND DRIVING BASED ON GAUSSIAN MIXTURE MODEL

摘要

The state of a driver before starting to drive, if not usual, may contribute to traffic accidents. This study focuses on the relationship between drivers' usual and unusual states before and while driving. Based on the data measured, we proposed a method to discriminate whether there is a relationship between driver states before and while driving, or not. We used Gaussian Mixture Model that incorporates Expectation Maximization algorithm to determine relationship between usual state and unusual state. We calculated the value we define as out-degree from our measured data. Then we compared the collection of data before driving and while driving, each group with posterior probability of 0.8 or more based on the out-degree. Data that belongs to both the groups of out-degree had a correspondence ratio of 65% from our experimental results based on the proposed method. Thus, we show that there is a possibility of having relationship with the unusual data measured before and while driving.