Wang Feiqin, Dong Jianping
Inner Mongolia Preschool Education College For The Nationalities, Ordos City, 017000, Inner Mongolia, China.
Graduate University of Mongolia, Ulan Bator, Mongolia.
Sci Rep. 2025 Jul 6;15(1):24138. doi: 10.1038/s41598-025-10269-7.
Children's mental health has become an increasingly prominent concern in modern education. However, insufficient attention from schools and families to children's psychological and emotional issues has exacerbated the problem. This study proposes a psychological emotion recognition model for children based on an Attention-Enhanced Convolutional Neural Network (AFCNN). The model initially employs a traditional Convolutional Neural Network (CNN) to extract image features and integrates an attention mechanism to enhance focus on key features, thereby improving recognition accuracy and precision. The proposed AFCNN model achieves an accuracy of 86.5%, representing a 14.4% improvement over traditional CNN models. By incorporating a channel attention mechanism, AFCNN significantly enhances the recognition of micro-expressions, particularly excelling in cross-age generalization and real-time performance. Experimental results demonstrate that AFCNN has broad application prospects in monitoring children's mental health and offers a precise tool for emotion recognition in educational and psychological counseling contexts.
儿童心理健康在现代教育中已成为一个日益突出的问题。然而,学校和家庭对儿童心理和情感问题的关注不足,加剧了这一问题。本研究提出了一种基于注意力增强卷积神经网络(AFCNN)的儿童心理情感识别模型。该模型首先采用传统卷积神经网络(CNN)提取图像特征,并集成注意力机制以增强对关键特征的关注,从而提高识别准确率和精确率。所提出的AFCNN模型准确率达到86.5%,比传统CNN模型提高了14.4%。通过引入通道注意力机制,AFCNN显著增强了对微表情的识别能力,尤其在跨年龄泛化和实时性能方面表现出色。实验结果表明,AFCNN在监测儿童心理健康方面具有广阔的应用前景,并为教育和心理咨询环境中的情感识别提供了一种精确工具。