A Distracted Driving Detection Model Based On Driving Performance

A large body of evidence on traffic accidents indicates that distracted driving is a major cause accidents. To investigate the relationship between driver performance in both normal and states to discriminate which state based performance, simulator was used simulate drivers an urban road environment, distraction task designed motivate enter cognitive state. Data were collected from 80 each two states, database built. Deep neural network (DNN) multilayer perceptron structure, relative defects other classical algorithms with unclear classification thresholds, DNN can be process high-dimensional data through its node node-linked structure. Combining Gray Wolf algorithm (GWO) for initial weights thresholds as whole, four-layer structure built predict test samples, while support vector machines ( SVM) random forest controls. The results show accuracy predicting samples 95.13%, SVM 77.56%, 72.32%. F1 score model higher than SVM, detection effect better applied detect status drivers.