基于跨连接LeNet-5网络的面部表情识别

为避免人为因素对表情特征提取产生的影响，本文选择卷积神经网络进行人脸表情识别的研究.相较于传统的表情识别方法需要进行复杂的人工特征提取，卷积神经网络可以省略人为提取特征的过程.经典的LeNet-5卷积神经网络在手写数字库上取得了很好的识别效果，但在表情识别中识别率不高.本文提出了一种改进的LeNet-5卷积神经网络来进行面部表情识别，将网络结构中提取的低层次特征与高层次特征相结合构造分类器，该方法在JAFFE表情公开库和CK+数据库上取得了较好的结果.

Facial Expression Recognition with Cross-connect LeNet-5 Network

In order to avoid the influence of human factors on facial expression feature extraction, convolution neural network is adopted for facial expression recognition in this paper. Compared with the traditional method of facial expression recognition which requires complicated manual feature extraction, convolutional neural network can omit the process of feature extraction. The classical LeNet-5 convolutional neural network has a good recognition rate in handwritten digital dataset, but a low recognition rate in facial expression recognition. An improved LeNet-5 convolution neural network is proposed for facial expression recognition, which combines low-level features with high-level features extracted from the network structure to construct the classifier. The method achieves good results in JAFFE expression dataset and the CK+ dataset.