This paper uses an algorithm of a genetic neural network to conduct an in-depth study and analysis of the attention relationship in pre-school education of young children. In face expression images, expression features are often concentrated and distributed in some local regions, while convolutional neural networks are homogeneous for feature extraction of facial expression images. To address the problem of the nonuniform distribution of expression features, a facial expression recognition method based on the spatial domain attention mechanism is proposed. A comparison of the differences between group and grade before the intervention using a two-way multivariate ANOVA revealed a nonsignificant interaction effect with a value of 0.90 for the group-grade interaction effect and an F-value of 0.36. The group's main effect had a value of 0.90 and an F-value of 0.36, with the experimental and control groups being equal groups on multiple dependent variables. The salient features in face expression images that are more important for facial expression recognition can be selected adaptively, and relatively high weights are assigned to these salient features. Experimental results on the image library show that the introduction of a spatial domain attention mechanism can improve the average recognition accuracy of face expressions by about 1.1% and 1.0%, respectively. A hybrid domain attention mechanism-based face expression recognition method is proposed to extract important expression features from both the spatial domain and the channel domain. Experimental results on the image library show that combining the spatial domain attention mechanism and the channel domain attention mechanism can improve the average recognition accuracy by about 2.1% and 1.5%, respectively. The use of rule games was effective in developing children's sustained and selective attention; the effect of rule games on the sustained and selective attention of older boys and girls did not show significant gender differences.
展开▼
