Inattentive Driving Detection And Indentification Based On Smartphones

With the continuous development of car ownership,driving safety is getting increasing social attention.Achieving real-time inattentive driving detection is the basis for improving driving safety.Existing work on driving safety fails to get a fine-grained result for inattentive driving behavior detection and identification,and usually requires the pre-deployment of hardware facilities or the placement of intelligent terminals in the car.To improve drivers' awareness of their own inattentive driving behavior and prevent potential accidents,we propose a real-time inattentive driving detection system based on smartphones by leveraging smartphone's speaker and microphone.According to the definition of inattentive driving behavior,we abstract four specific actions,including turning around,taking back,picking up and hand movement.Through analyzing acoustic data of 10-day driving traces collected from real driving environments,we find that all of the four types of driving behaviors have their unique patterns on the Doppler frequency shift.Specifically,we extract several effective features to capture the patterns of inattentive driving behavior.After that,a machine learning method,Support Vector Machine(SVM),is trained by the features and outputs a classifier which conducts fine-grained detection and identification.Using the data of extensive experiments with 8 volunteers driving for another 2 weeks in real driving environments,the average accuracy of the system is 93.72%.The system is low cost,easy deployment and use,because all the processes are completed on the intelligent terminal.