Research on the Multimodal Application Method Based on AT-BiGRU in College Students' Mental Health Education

Predicting the stress conditions of university students and discerning individuals with psychological abnormalities with a certain degree of precision. Integrating various modalities, the state data is fed into the Bidirectional Gated Recurrent Unit (BiGRU) network. Subsequently, the self-attention...

Full description

Saved in:
Bibliographic Details
Published in2024 4th International Conference on Computer Science and Blockchain (CCSB) pp. 159 - 163
Main Author Gui, Jingjing
Format Conference Proceeding
LanguageEnglish
Published IEEE 06.09.2024
Subjects
Accuracy
attention mechanisms
BiGRU
Brain modeling
Computational modeling
Emotion recognition
Hardware
Integrated circuit modeling
Mental health
multimodal analysis
Optimization
Physiology
Programming profession
stress states
Online AccessGet full text

Cover

More Information
Summary:Predicting the stress conditions of university students and discerning individuals with psychological abnormalities with a certain degree of precision. Integrating various modalities, the state data is fed into the Bidirectional Gated Recurrent Unit (BiGRU) network. Subsequently, the self-attention fusion process is applied to combine the three modalities-expressive, physiological, and speech-to derive low-dimensional fusion features. These features are then input into the classifier for emotion category determination. Furthermore, the primary programming model, denoted as AT-BiGRU, was deployed onto the personal computer. The system's hardware circuit and device driver, based on the Cortex-M4 kernel, were formulated to facilitate convenient, low-power multi-channel simultaneous data acquisition, and stress state analysis. Experimental outcomes demonstrate an average classification accuracy of 62.77% among subjects, surpassing some analogous methodologies. This substantiates the efficacy and general applicability of the proposed method in multimodal emotion recognition. Additionally, AT-BiGRU exhibits notable proficiency, achieving an accuracy exceeding 85% in effectively classifying three distinct stress states. This functionality enables the assessment of psychological stress in students.
DOI:10.1109/CCSB63463.2024.10735618