Research On Driver Behavior Analysis Based On Hidden Markov Model

In recent years,for the recognition and evaluation of driver behavior,people design system with the modeling theory and data analysis.However,there are inaccurate driver behavior evaluation due to the driver behavior has some stochastic characteristics and the driving scene is not considered enough.In the bus field,driver behavior identification and evaluation not only can regulate the driver behavior,but also has important relationship with the driver's income,so the accuracy need to be higher.The status of identification and evaluation of driver behavior at home and abroad is analyzed and summarized in this thesis.Based on the Hidden Markov Model and cluster technique,the multi-layer driver behavior evaluation model is proposed after driver behavior recognition.This model consists of three parts including driving operation,driving intention and driving scenario.The rapid deceleration is chosen as the research content and prove the effectiveness of the model,the research work is as follows:(1)Calculate the driving operation cut-off value.To judge the bad driving operation,the velocity-acceleration density parameter is used in DBSCAN which obtains the cut-off value between the normal deceleration and the rapid deceleration.(2)Determine the driving intention.To determine the specific driving intention in a certain driver behavior,the HMM is used to model the emergency avoidance intention which tells the emergency avoidance intention from the rapid deceleration.(3)Analyze the driving scenario.To analyze the potential cause of driver behavior in a specific scenario,time and space are taken as observation parameters of rapid deceleration.The k-means is used to mine the rapid deceleration under the specific scenario.In this thesis,the data analysis and mining of the bus database prove that the multi-layer driver behavior evaluation model has a high accurate evaluation for the rapid deceleration,and it shows that this model is feasible and effective.