基于卷积神经网络的人脸表情识别研究  

Research on Face Expression Recognition Based on Convolutional Neural Network

作　　者：张璟 ZHANG Jing(School of Computer and Information Technology, Shanxi University, Taiyuan 030006, China)

机构地区：[1]山西大学计算机与信息技术学院

出　　处：《电脑知识与技术》2019年第6期212-213,215,共3页Computer Knowledge and Technology

摘　　要：人脸表情识别是计算机视觉领域的一项重要任务。针对经典VGG模型参数量巨大、训练成本较高的问题,提出了一种基于VGG模型的改进卷积神经网络。改进的模型减少全连接层的使用有效减少了参数量,加入批规范化层和dropout随机失活操作,进一步加速了模型的收敛,从而获得较好的分类效果。实验采用FER2013公开人脸表情数据集,实验结果表明,改进的模型提高了表情识别的准确率率和泛化能力,减少了时间消耗。Face expression recognition is an important task in the field of computer vision. An improved convolutional neural network based on VGG model was proposed to solve the problem of large number of classical VGG model parameters and large training time cost. The improved model reduces the use of full connection layer, effectively reduces the number of parameters, and adds batch normalization layer and dropout random deactivation operation, which further accelerates the convergence of the model and achieves better classification results. In the experiment, FER2013 was used to publish facial expression dataset, the experimental results show that the improved model improves the recognition rate and generalization ability of expression recognition, and reduces the time consumption.

关 键 词：卷积神经网络 人脸表情识别 表情分类 批规范化 卷积 

分 类 号：G642[文化科学—高等教育学]