As more and more trafc accidents were caused by driving fatigue inrecentyears,detectionofdrivingfatiguehasbecomeanimportantresearchtopic in safety driving. In the past researches of driving fatigue detection,facial video signal, body temperature, electroencephalogram (EEG) wereoften used. Diferent from these signals, drivers’ grip force on steeringwheel has higher signal noise ratio (SNR), and can be collected more eas-ily, which will not disturb driving. So we propose an unobtrusive way todetect fatigue for drivers through grip forces on steering wheel. At frst,we introduce the simulated driving platform and the system of detectinggrip force. Fatigue-related features are extracted from wavelet coefcientsthroughwavelettransformationandflteredbymovingaveragemethodandlineardynamicalsystem. Wedesignaperiodicalaudiotask,andusethere-actiontimetomeasuresubjects’vigilance. Intheend,Wecomparetheper-formance of k-nearest neighbors, linear discriminant analysis, and supportvector machine (SVM) on the task of discriminating drowsy and awake s-tates. SVM with radial basis function reaches the best accuracy,79.1%onaverage. |