Driver fatigue detection by fusing multiple cues
Rajinda Senaratne, David Hardy, Bill Vanderaa, Saman Halgamuge, DR Liu (ed.), SM Fei (ed.), ZG Hou (ed.), HG Zhang (ed.), CY Sun (ed.)
ADVANCES IN NEURAL NETWORKS - ISNN 2007, PT 2, PROCEEDINGS | SPRINGER-VERLAG BERLIN | Published : 2007
A video-based driver fatigue detection system is presented. The system automatically locates the face in the first frame, and then tracks the eyes in subsequent frames. Four cues which characterises fatigue are used to determine the fatigue level. We used Support Vector Machines to estimate the percentage eye closure, which is the strongest cue. Improved results were achieved by using Support Vector Machines in comparison to Naive Bayes classifier. The performance was further improved by fusing all four cues using fuzzy rules. © Springer-Verlag Berlin Heidelberg 2007.
Authors wish to thank Mr. Peter Sparkes andMr. Cameron Joss for their support given throughout the project.This project is partially funded by the Australian Research Council.