基于深度卷积和自编码器增强的微表情判别

为了判别微表情种类，提出基于深度卷积神经网络和迁移学习的微表情种类判别网络MecNet. 为了提高MecNet在CASME II、SMIC和SAMM联合数据库上的微表情种类判别准确率，提出基于自编码器的微表情生成网络MegNet，以扩充训练集. 使用CASME II亚洲人的微表情样本，生成欧美人的微表情样本. 设计卷积结构实现图像编码，设计基于子像素卷积的特征图上采样模块实现图像解码，设计基于图像结构相似性的损失函数用于网络优化. 将生成的欧美人的微表情样本加入MecNet训练集. 实验结果表明，使用MegNet扩充训练集能够有效地提高MecNet微表情种类判别准确率. 结合MegNet、MecNet的算法在CASME II、SMIC和SAMM组成的联合数据库上的表现优于大部分现有算法.

Micro-expression classification based on deep convolution and auto-encoder enhancement

A micro-expression classification network named MecNet was proposed based on the deep convolutional neural network and transfer learning in order to classify the types of micro-expressions. MegNet was proposed to expand the training set in order to improve the accuracy of micro-expressions classification of MecNet on the joint database of CASME II, SMIC and SAMM. MegNet is a micro-expression sample generation network based on the auto-encoder. Asian micro-expression samples of CASME II were used to generate western micro-expression samples. A convolution structure was designed to encode images, and a feature map upsampling module was designed based on the sub-pixel convolution to decode images. A loss function based on the structural similarity of images was designed to optimize the network. The generated western micro-expression samples were added to the training set of MecNet. The experimental results show that the accuracy of micro-expression classification of MecNet can?be?effectively?improved?by?using?MegNet?to?expand?training?sets. The algorithm combining MegNet and MecNet performs better than the most existing algorithms on the joint database composed of CASME II, SMIC and SAMM.