Today’s society and economy are developing rapidly, and the academic and employment environments faced by college students in them are also changing dramatically, and the psychological pressure is also increasing, so it is urgent to improve their psychological adaptability. In this paper, the EEG signals are transformed into four-dimensional data through the data reconstruction module, and then fed into the multi-scale 3D CNN module to extract the temporal and spatial frequency features of the EEG signals in different scales and dimensions, the spatial attention mechanism module learns the size of the contribution of the different EEG channels to the emotion generation, and assigns a higher weight to the information that has significant emotional information, and then the global temporal features are mined by using the BLSTM module, and the Emotion Classification Module to obtain the final emotion classification results using the fully connected layer and softmax function, and finally complete the emotion recognition model (3DMSCA) construction. The emotion recognition model is integrated into the mental health education of college students to obtain a mental health education model that incorporates emotion recognition technology, and a psychological adaptation enhancement path based on the mental health education model is established, which is explored and analyzed with the help of structural equation modeling. The structural equation models GFI, CFI, NFI, TLI, and RMSEA for college students’ psychological adaptation enhancement pathway meet the research requirements, with values of 0.917, 0.924, 0.901, 0.916, and 0.038, which means that the models have a better overall fit, and they can show the interaction mechanism between college students’ psychological adaptation enhancement pathway based on the mental health education model. The interaction mechanism of the model has strategic application value for students’ mental health education in the post epidemic era.
