Research on Driver Emotion Recognition Based on Temporal Features of Expression Transformation
Aiming at the issues of insufficient real-time recognition and low accuracy in existing driver emotion recognition methods,a driver emotion recognition method based on facial expression recognition and temporal emotion representation is proposed.Firstly,a VGG Lite driver expression recognition model is established.On the traditional VGG Net model structure,the parameters of the model are greatly reduced by modifying the convolutional layer stacking structure,and the activation function is adjusted to enhance the model's expression capability of detailed facial features,and the performance optimization layer is added to the model to improve the convergence and generalization.Secondly,the relationship between temporal transformation in facial expression and emotional states is analyzed,and the emotional expression mode of time series evolution is studied.A driver's temporal emotion recognition method including facial expression temporal transformation,expression-emotion quantization mapping and temporal emotion expression is designed.Then,the proposed VGG Lite driver expression recognition model is compared with other models by using the Fer2013 dataset.The results demonstrate that the proposed model not only maintains high recognition accuracy but also effectively reduces the number of model parameters,thereby improving the recognition speed.In addition,a high accuracy of 98.8％is achieved in recognizing facial expression using a self-made dataset,proving the effectiveness of the model in recognizing driver expression in different driving scenarios.Finally,the proposed temporal emotion recognition method is experimentally validated using the example of bus driver's emotion recognition.The results show that the method can accurately identify the complex emotional state of the driver under various facial expression transformation,with an average recognition accuracy of over 95％,surpassing single-frame emotion recognition methods by more than 5％.The emotion recognition time per frame is averaged at less than 0.03 seconds,with an average recognition rate of over 10 frames 0per second,meeting the real-time requirements of traffic driving emotion recognition.This method can timely and accurately assess the driver's emotional state,providing a more effective means to improve the overall safety and efficiency of the transportation system.
NETL
NSTL
万方数据
