引入注意力机制和中心损失的表情识别算法

Expression recognition algorithm introducing attention mechanism and center loss

针对表情识别领域中识别准确率不高的问题,以卷积神经网络为基础,提出一种引入特征通道注意力机制和中心损失函数的表情识别算法。以VGG11作为网络主体框架,引入特征通道注意力机制,在每一层卷积层之后增加压缩激励(SE)模块,使模型学习到不同特征通道的重要程度,从而提高模型对特征通道的敏感性,进一步提升模型的特征表示能力。同时,通过中心损失函数联合Soft Max损失函数对神经网络进行监督训练,有效增加类间距离,保证类内紧密性,提高模型的识别准确率。通过在CK+数据集上测试,实验结果表明:改进模型的识别准确率达到97. 07%,高于其他典型算法。

Aiming at the problem that the recognition accuracy is not high in the field of expression recognition,an expression recognition algorithm based on the convolutional neural network and introducing the attention mechanism of the feature channel and the central loss function is proposed. Using VGG11 as the main frame of the network,the feature channel attention mechanism is introduced. After each convolutional layer,a squeeze-andexcitation( SE) block is added to make the model learn the importance of different feature channels,which improves the sensitivity of the model to the feature channel and further enhance the feature representation ability of the model. At the same time,the neural network is supervised and trained by the center loss function combined with the Soft Max loss function,which effectively increases the distance between inter-classes,ensures the tightness of the intra-class,and improves the recognition accuracy of the model. By testing on the CK + dataset,the experimental results show that the recognition accuracy of the improved model reaches 97. 07 %,which is higher than other typical algorithms.