基于全局空间注意力的人脸表情识别

Facial Expression Recognition Based on Attention Mechanism

人脸表情识别是模式识别领域中一个重要的研究方向。传统的机器学习方法受限于需要手动提取特征,该方式会导致识别结果的泛化能力不足,且稳定性较差。针对该限制,设计了一种基于深度学习的人脸表情识别算法,该算法通过卷积神经网络提取特征,然后经过全局空间注意力模块对特征分配权重,增强并融合重要特征、抑制边缘特征,从而提升网络分类的准确性。通过在FER2013人脸表情数据集上的实验,验证了该算法的合理性与有效性,最高达到了1.014%的准确度提升。最后,将算法应用于真实场景下的人脸表情识别,同样能拥有较高的识别精度,验证了该算法在真实环境下的有效性。

Facial expression recognition is an important research direction in the field of pattern recognition.Traditional machine learning methods are limited by the need to manually extract features,which will lead to insufficient generalization ability of the recognition results and poor stability.In view of this limitation,this paper designs a facial expression recognition algorithm based on deep learning.The algorithm extracts feature through the convolutional neural network,and then assigns weights to the features through the global spatial attention module, which enhances and fuses important features,and suppresses edge features.Thereby improving the accuracy of network classification.Through the experiments on the FER2013 facial expression data set,the rationality and effectiveness of the algorithm are verified,and the accuracy is improved by up to 1.014%.Finally,applying the algorithm to facial expression recognition in real scenes can also have high recognition accuracy,which verifies the effectiveness of the algorithm in real environments.