基于AdaBoost与改进CNN的驾驶员愤怒情绪识别

Recognition of driver anger emotion based on AdaBoost and improved CNN

针对驾驶员愤怒驾驶引发的道路交通安全问题,提出一种基于AdaBoost与改进卷积神经网络(convolutional neural networks,CNN)的驾驶员愤怒情绪识别模型。利用AdaBoost算法检测驾驶员人脸并采集面部表情数据,建立基于驾驶员情绪的表情数据集。针对CNN中LeNet模型的不足,提出改进方案,搭建改进LeNet模型。利用数据集对LeNet模型和改进LeNet模型进行训练,并对二者的辨识结果进行对比分析。结果表明,改进LeNet模型的正确率为95.83%,F1-Score为0.8837,损失值为0.0745,相较于LeNet模型而言,该模型对于驾驶员表情的识别性能更高。改进LeNet模型的高识别率能够在一定程度上降低愤怒情绪所带来的交通安全隐患,为辅助安全驾驶提供理论依据,进一步提高人机共驾的安全性。

Aiming at the problem of road traffic safety caused by anger driving,a driver anger recognition model based on AdaBoost and improved convolutional neural networks(CNN)was proposed.AdaBoost algorithm was used to detect the driver’s face and collect facial expression data for establishing an expression data set of drivers’emotions.For the shortcomings of LeNet model in CNN,an improved LeNet model was proposed.The dataset was built to train the LeNet model and the improved LeNet model.The results show that the accuracy,F1-Score and loss of the improved LeNet model was 95.83%,0.8837 and 0.0745,respectively.Compared with the LeNet model,the improved LeNet model has higher recognition performance for driver’s expression.The high recognition rate of the improved LeNet model can reduce the potential traffic safety hazards caused by anger to a certain extent,and provide a theoretical basis for assisted safe driving,also improve the safety of man-machine co-driving.