基于多层交叉注意力融合网络模型的人脸图像情感分析

目前大多数人脸图像情感分析方法只单方面关注图像整体或局部来构建视觉情感特征表示，忽略了二者在情感表达上的协同作用。针对此问题，提出了一种多层交叉注意力融合网络情感分析方法。该方法首先利用特征相关性分析实现最大化类的可分性；其次通过多层交叉注意力网络中的多个不重叠的注意力区域来提取整体和局部的信息；然后将整体与局部提取的注意力图进行融合，来共同训练图像情感分类器并进行情感分析。实验结果表明，提出的方法在真实数据集RAFDB上的情感分类准确率达到了88.53%,优于现有其他方法，验证了该方法的有效性与优越性。

Emotion analysis of face image based on multi-layer cross attention fusion network model

At present, most facial image emotion analysis methods only focus on the whole or part of the image to construct visual emotion feature representation, ignoring the synergistic effect of the two on emotion expression. To solve this problem, a multi-layer cross-attention fusion network sentiment analysis method is proposed. Firstly, feature correlation analysis was used to maximize class separability. Secondly, global and local information can be extracted from multiple non-overlapping attention regions in the multi-layer cross-attention network. Then the attention attempts extracted from the whole and the local were fused to train the image emotion classifier and carry out the emotion analysis. Experimental results show that the accuracy of the proposed method on real data set RAF-DB reaches 88.53%, which is better than other existing methods, and verifies the effectiveness and superiority of the proposed method.