Driver Data Collection And Pattern Analysis

With the rapid development of automobile industry and technology,the shortcomings of conventional cars are more and more exposed.Intelligent vehicles are widely accepted by people for improving the problems of conventional cars and are developing rapidly.More and more attention has been paid to the driver-centered active safety system of intelligent vehicles.Therefore,the development of the driver model is particularly important for the research of intelligent vehicles.By collecting data from the human driver and analyzing its driving mode,we can study the driving habits and characteristics of the human driver,and establish a better driver model,which helps to develop the research of intelligent assistant driving technology.This paper is supported by the national key research project' Electric vehicle intelligent auxiliary driving technology research and development and industrialization'.Driver data collection and drive mode analysis problem is studied,mainly research on the following several aspects:1)A multi-source information collection and analysis system for drivers is designed and built.We designed a multi-source information acquisition system structure and summarized the general state of the driver in the driving process.Then we built a driving simulator and signal acquisition system,using the scene simulation software to simulate the driving environment,and collecting different kinds of signals from different sensors.It provides a platform and framework for the experiment and analysis of data acquisition below.2)The driver's physiological signal,included EEG and leg surface EMG,are collected.We extracted the characteristics of the signals and analyze the driving state of the driver under different driving conditions.By using the sample entropy of EEG beta rhythm wave and the high frequency wavelet coefficients of surface EMG signal,we analyze the driver's states in different stages.Finally,by comparing the front vehicle's speed,spacing and beta rhythm sample entropy,we analyze the factors affecting the driver's state change,and acquire related rules of driver's state change in general driving conditions and driving assistance rules.3)Collect the driver's front and side image information,on the basis of target recognition,the eye opening in the frontal image and the front angle of the body in the side image are recognized and extracted.Then analyze the difference of features,and fuse the multi-channel visual information to analyze the driver's related driving behavior and state change.4)We selected the drivers' high frequency wavelet coefficients of surface EMG,EEG beta rhythm sample entropy and eye opening of visual signal as the fusion features,Through the combination of fuzzy comprehensive evaluation and DS theory,the features are fused on decision layer.According to the fusion results,the driving mode was determined,and the experiment verifies the effectiveness of the fusion analysis compared with the single feature analysis.