Research On Driving Attention Detection Based On Multi-motion Features

In recent years,people's living standards have improved significantly,which has promoted the rapid development of Chinese auto industry.At the same time,the rapid growth of domestic GDP and the acceleration of domestic urbanization not only bring new opportunities to the automotive industry,but also promote the rapid development of my country's automotive industry.The continuous rise of car ownership not only brings convenience to people but also leads to the continuous rise in the number of traffic accidents.According to the statistics,distracted driving is one of the main causes among many causes of traffic accidents.Machine vision,which can effectively detect driving attention,is the main detection method of traditional driving attention.But there are still some problems,such as limited camera coverage,violation of user privacy,and large memory usage.With the rapid development of wearable technology,more and more wearable technology is applied to the field of identification.Therefore,this paper proposes a driving attention detection method based on multi-motion features.First,the wearable acquisition module is used to collect the movement data of the driver's head and hands,and the movement features characterizing driving attention are analyzed and determined;then the PCA method is used to fuse the effective feature matrix.Finally the driving attention detection model with multi-motion features was constructed.The main contributions of this article are as follows:(1)A driving attention detection method based on head pose is proposed.In order to eliminate the impact of various environments during driving and protect the privacy of users,a wearable motion collection module was used to collect the driver's head movement information,then movement information such as acceleration,angular velocity,quaternion,that characterize the head pose is determined.The corresponding time-domain and frequency-domain features of the motion information representing the head pose are extracted and transmitted to the detection model based on random forest.Experimental results show that the detection accuracy of this method reaches 84%,which verifies the effectiveness of the proposed method.(2)A driving attention detection method fusing hand motion features is proposed.In order to improve the detection accuracy and robustness,this paper proposes a driving attention detection method that fuses hand movement features.Through research and data analysis,the motion information that characterizes the hand pose is determined.That is,the acceleration,angular velocity are used to characterize the hand movement,and the gravity acceleration component and quaternion characterize the features of rest after hand movement.Then extract the relevant time and frequency domain features from the collected time series motion data of the head and hands,and finally use the random forest algorithm to detect driving attention.Experimental results show that the detection accuracy of the proposed method reaches 88.2%(accuracy rate).Compared with the driving attention detection method based on head pose,the accuracy of the detection method of fusing hand motion features is improved by 4.2%,further illustrating the effectiveness of the proposed fusion method.(3)A driving attention detection method based on multi-motion features is proposed.Aiming at the problem of "dimensional disaster" caused by too many feature dimensions and the timeliness of data processing,a driving attention detection method based on multi-motion features is proposed.First,the time and frequency domain features with high redundancy extracted from the acceleration and angular velocity of the head and hands are reduced by the principal component analysis method,and then the remaining motion features are fused.Finally,a driving attention detection model based on multi-motion features is constructed.The experimental results show that the detection accuracy of the driving attention detection method based on multi-motion features has been reduced,but the time required for the detection has been reduced a lot,which proves the effectiveness of this method.