A hierarchical fuzzy decision model for driver's unsafe states monitoring

In this paper, we present a hierarchical fuzzy decision model for driver's unsafe states monitoring. In this model, we want to combine the multi-aspect information into one output parameter, which stands for the level of inattentiveness of the driver. Five parameters are calculated: eye-closing duration, mouth-opening duration, mouth state change frequency (from opening to closure or from closure to openning), face detection failure duration, and historical information. Dozens of videos collected in Beijing urban streets of one driver are used for the test of the model. Experiment results and conclusions on the performance of the model are presented.