EEG SENTIMENT ANALYSIS MODEL BASED ON DEEP LEARNING FRAMEWORK AND INTERACTION DESIGN THEORY

Emotional analysis methods based on electroencephalography and interaction design theory are of great significance in the field of affective computing. It can not only achieve more accurate emotion recognition through in-depth analysis of brain activities but also lay the foundation for building a more natural emotional interaction system. This paper proposes an electroencephalogram (EEG) emotion analysis model that combines a deep learning framework and interaction design theory. The main contributions are as follows: (1) An improved Longformer model (ETLD-Transformer) is proposed, which combines the inter-segment information transfer mechanism and dynamic position coding. The strategy and multi-branch convolution mechanism significantly improve the accuracy and coherence of the model in emotion recognition. (2) By comparing with baseline models such as Longformer, the advantages of the improved model in processing long-term sequences and complex emotional tasks are verified. (3) Combined with interaction design theory, the user emotional feedback mechanism is optimized, allowing the system to dynamically adjust interactive content and improve the naturalness and responsiveness of user experience. Experimental results show that this model can not only effectively capture long-distance dependencies in EEG signals, but also exhibit excellent performance in multi-dimensional emotion classification tasks, providing efficient solutions for the fields of EEG emotion analysis and interaction design.