结合注意力机制的面部表情识别方法研究

打开文本图片集
摘 要: 传统CNN对重要通道特征关注不足,制约面部表情识别准确率。文章将通道注意力机制应用到面部表情识别中,即将通道注意力模块嵌入到卷积网络中。在Fer2013和CK+表情数据集上的验证结果表明,该方法有较高的识别率。
关键词: 面部表情识别; 通道注意力机制; 卷积网络; 表情数据集
中图分类号:TP181 文献标识码:A 文章编号:1006-8228(2022)03-24-03
Abstract: Traditional CNN pays insufficient attention to important channel features, which restricts the accuracy of facial expression recognition. This paper applies the channel attention mechanism to facial expression recognition, i.e., to embed the channel attention module in the convolution network. The verification results on Fer2013 and CK + expression data sets show that the proposed method has high recognition rate.
Key words: facial expression recognition; channel attention mechanism; convolution network; expression data sets
0 引言
面部表情不仅可显示交际者的情绪状态,还可传达交际者的深层思想和情感[1]。(剩余4449字)