Fatigue Detection Based On Texture Features Of Eyes And Identity Recognition

Fatigue,or tiredness,is defined as a phenomenon where people are disordered for some reason and their operational capability are decreased.When tired,people will not able to focus their attention,and their ability to judge and remember could be seriously influenced.Overfatigue will lead to the lack of emergency response capability and cause accidents.It is important to develop an algorithm for fatigue detection.Since fatigue detection based on EEG signal and body actions could affects normal driving or have serious lag,fatigue detection based on visual information is considered of great significance to public safety.In this paper,we proposed a fatigue detect algorithm based on texture features of eyes and face recognition.The visual images are captured through NIR devices to guarantee the application for both day and night.A single 3d face model is used for face frontalization,with face pose estimation and facial landmark localization error compensation algorithms we proposed.Probabilistic SVM model is applied for classification of HOG features extracted from eye images,and then PERCLOS measured for fatigue state decision.A special database is constructed for training and testing,with the fatigue information obtained from volunteers.Besides,Face recognition modules is introduced to realize customization of fatigue detection for different people.At the end,an actual fatigue detect system is constructed with software program and hardware devices to measure the accuracy.