Study On Facial Expression Recognition Of Automobile Drivers

The emotions of car drivers will affect the driving operation.When the drivers have emotional fluctuations due to their own pressure or the external environment,they are easy to produce dangerous driving behaviors,thus threatening the traffic safety.Therefore,real-time monitoring of drivers’ emotions can reduce the possibility of traffic accidents by timely reminding or adjustment and relief when their emotional state is abnormal.The drivers’ emotional state can be analyzed through facial expressions.This paper studies the related problems of facial expression recognition of automobile drivers,and the main research contents and results are as follows:(1)In view of the driving task will affect the driver’s emotional expression and the facial expression dataset obtained in the natural state could not highlight the driver’s emotional characteristics.In this paper,we collected the images in film and television works that reflect the six basic emotions of drivers,as well as seven types of expressions in a state of calm.Appropriate naming rules and emotion determination methods were formulated,and 388 images were selected to establish the driver’s facial expression dataset after emotion determination of the intercepted facial images.Considering that the more complex background in the image may affect the recognition rate of the driver’s expression recognition,a Multi Task Convolutional Neural Network(MTCNN)was used to detect and alignment the face images.At last,adjust the images to a uniform size.(2)In view of the problems such as low accuracy of expression recognition and network overfitting caused by insufficient and unbalanced dataset,data enhancement is needed.This paper proposed an improved CycleGAN network model,it is added in the original network category constraints in order to realize more category expression image transformation,at the same time the discriminant classifier is used to replace the discriminant in the network,in order to improve the robustness of the model,a gradient penalty mechanism is added to the loss function of the discriminator.The improved Gycle GAN model was tested on driver facial expression dataset and CK+ dataset.The results show that the proposed algorithm based on improved CycleGAN can effectively improve the recognition rate of driver facial expression.