Research On Driving Behavior Detection Technology Based On Wearable Equipment

Traveling by cars is quite usual in recent years,and driving safety has attracted extensive att ention.Improper driving behavior is an important cause of traffic accidents.Therefore,real-time detection of driver's driving behavior is helpful in improving transportation safety.This thesis focuses on the detection of lane changing,which is a specific risky driving beha vior,and a lane changing detection algorithm based on the driver's forearm acceleration is propos ed.In the proposed algorithm,accurate lane changing detection is achieved by using forearm acc eleration data collected from smart watches,and car speed data collection from GPS sensors.Ge nerally,the proposed algorithm works as follows.First,we select a window whose size is dynam ically adjusted according to current car speed,and transform the acceleration data within the win dow to a frequency domain using fast Fourier transform(FFT);After that,a detection interval is dynamically determined in the frequency domain and the acceleration energy within the detectio n interval is calculated;Then,the obtained acceleration energy is comparing with a dynamic thre shold to detect lane changing;Finally,the real part of the acceleration data within the detection i nterval is calculated to further determine the lane changing direction.The proposed algorithm dy namically adjusts the time-domain transform window,frequency-domain detection interval and t hreshold based on current car speed,and can realize accurate lane changing identification under various car speeds.Based on the proposed algorithm,we have implemented a lane changing detection system o n Android smart watches and smart phones,and validated the feasibility and effectiveness of the proposed algorithm through extensive road experiments.