Research On Driver’s Lane Change Intention Recognition Based On Hidden Markov Model

The driver needs to keep observing the surrounding environment and makes a judgment when driving, especially changing lanes. If he doesn’t observe sufficiently or makes a wrong decision, it may lead to accidents easily. If the driver’s lane change intention could be predicted in advance, then we can conduct risk assessment according to the following lane change and warn the driver to take correct steps, so traffic accidents can be avoided if possible. The risk is higher while driving on the freeway at a higher speed, so it is very important to conduct the intention recognition on the freeway for the road safety.This article analyzed the status of driving intention recognition at home and abroad first, then chose the driver’s visual characteristics to study the driver’s lane change intention, independent-sample T test(T-test) was used to analyze the visual difference between lane-keeping and lane-changing phases, it extracted the parameters which can characterize the rules of visual characteristics, the model of recognizing lane change intention was established based on HMM. At last, the result of the model was evaluated through test samples. The specific research contents are as follows:1. The article analyzed the lane-change behavior and divided it. First, the test scheme is designed according to the condition, such as staff acquisition and training; the standardized design of the test process and so on. All the above can provide a theoretical basis and guarantee for the accuracy.2. The acquisition and analysis of drivers’ visual data in different intention phases. First, the view field was divided into five parts. Then the change rules of visual parameters in different intention phases were analyzed from three aspects, namely fixation, saccade and head rotation, the independent-sample T test(T-test) was adopted to check the diversity of different variables. Finally, the visual indexes were determined.3. The recognition model was established based on HMM. First, the modeling idea was explained from the theory and application of HMM. Then the visual characteristic parameters were extracted for modeling, and 276 samples as the training sample from the gathered data were chosen to train the model, and the model of lane change intention recognition was established on the basis of HMM. At last, the test samples were used to verify the model.This paper studied the key technical issue which the recognition of lane change intention is facing on the freeway deeply. The model which was established has a high percentage of accuracy. The study could provide the theoretical basis and technical support for the research on the driving behaviors, thus to improve the driving safety and reduce the incidence of traffic accidents.