摘要
针对目前传统人脸表情识别算法存在特征提取复杂、表情识别率低等问题,提出一种基于混合注意力机制的ResNet人脸表情识别方法。该方法把通道注意力模块和空间注意力模块组成混合注意力模块,将混合注意力模块嵌入ResNet残差学习分支中。针对CK+人脸表情数据集过小问题,采用数据增强策略扩充数据集。实验结果表明,改进后的ResNet在CK+数据集上表情识别准确率为97.04%,有效提高了表情识别准确率。
The traditional facial expression recognition algorithm has the problems of complex feature extraction,low expression recognition rate and so on,in order to solve these problems,a facial expression recognition method based on the combination of hybrid attention mechanism and ResNet is proposed.The method combines the channel attention module and the spatial attention module into the hybrid attention module,and then embeds the hybrid attention module into the ResNet residual learning branch.For the CK+facial expression data set is small,the data augmentation strategy is adopted to expand the data set.The experimental results show that the expression recognition accuracy of the improved ResNet is 97.04%on the CK+data set.This method improves the accuracy of expression recognition.
作者
高健
林志贤
郭太良
Gao Jian;Lin Zhixian;Guo Tailiang(College of Physics and Information Engineering,Fuzhou University,Fuzhou 350116,China)
出处
《信息技术与网络安全》
2020年第1期59-62,99,共4页
Information Technology and Network Security
基金
国家重点研发计划课题(2016YFB0401503)
福建省科技重大专项(2014HZ0003-1)
广东省科技重大专项(2016B090906001)
广东省光信息材料与技术重点实验室开放基金资助项目(2017B030301007)
