Study On Driver’s Angry Emotion Based On Facial Expression And Pulse Information Fusion

With the development of our society and economy, and the continuous improvement of people’s living standard, family cars have come to innumerable families, and become an important means of transportation in people’s daily life. meanwhile, the number of taxis, buses, trucks, etc. is also increasing. However, while the increase of vehicles has brought convenience to people, it also makes the phenomenon of traffic congestion increasingly serious, and intensifies the conflicts between pedestrians and vehicles, vehicles and vehicles, vehicles and roads. As a result, traffic accidents caused by angry emotion increase year by year, and "Road Rage" has come to light. In recent years, the accident caused by "road rage" increased seriously, so how to accurately identify the anger emotion of drivers, and do effective intervention, regulation and management to angry drive behaviors has become the hotspot of nowadays research on drivers’ angry emotion.In this paper, we took the drivers’ facial expressions and pulse information as research objects, and collected the subjects’ facial expressions and pulse information under the emotion state of calm, happy and angry evoked by the simulated driving and emotion inducing experiment, then study the drivers’angry emotion using computer image processing and physiological signal processing. The major jobs of this paper are the following:(1) Build the simulated driving and emotion inducing experiment. Collect the subjects’ real-time facial expression information of video and pulse information during the experiment, and we obtain emotional information according to a questionnaire and the record of the experimenter to guarantee the effectiveness of the data.(2) Do image pre-processing to the obtained facial image, and do pupil-location by using computer image processing and mixed integral projection, then find the accurate location of human face according to the relationship between the interpupillary distance and the size of human face, finally do feature extraction and feature reduction to the image to get emotional recognition feature vectors by using Gabor wavelet and PCA.(3) Do pulse process to gain the pulse information feature vector. De-noise the pulse information based on the Stein unbiased like-hood estimation and layered threshold method. Then extract the mean and variance of main wave peak, the mean and variance of pulse information, and the approximate entropy characteristic between the main wave periods. And use these feature vectors to recognize the drivers’ angry emotion.(4) Do information fusion to the extracted facial expression feature vector and pulse feature vector, and gain the appropriate FLD classifier through simulation experiments on MATLAB platform, finally complete the drivers’ emotional recognition using the designed classifier.