MIND-EEG: Multi-granularity Integration Network with Discrete Codebook for EEG-based Emotion Recognition

TL;DR

This work tackles EEG-based emotion recognition by addressing the need to model dynamic, multi-scale spatial relationships among EEG nodes. It introduces MIND-EEG, a multi-granularity integration network that jointly models global brain state, intra-regional functionality, and inter-regional interactions, all built on discrete codebooks to vector-quantize learned network structures. A unified integrative loss promotes discriminative yet robust representations by coupling classification with codebook-consistency terms. Experiments on SEED-IV, SEED-V, and MPED under both subject-dependent and subject-independent settings demonstrate state-of-the-art performance and reveal diverse codebook embeddings that underpin the learned connectivity.

Abstract

Emotion recognition using electroencephalogram (EEG) signals has broad potential across various domains. EEG signals have ability to capture rich spatial information related to brain activity, yet effectively modeling and utilizing these spatial relationships remains a challenge. Existing methods struggle with simplistic spatial structure modeling, failing to capture complex node interactions, and lack generalizable spatial connection representations, failing to balance the dynamic nature of brain networks with the need for discriminative and generalizable features. To address these challenges, we propose the Multi-granularity Integration Network with Discrete Codebook for EEG-based Emotion Recognition (MIND-EEG). The framework employs a multi-granularity approach, integrating global and regional spatial information through a Global State Encoder, an Intra-Regional Functionality Encoder, and an Inter-Regional Interaction Encoder to comprehensively model brain activity. Additionally, we introduce a discrete codebook mechanism for constructing network structures via vector quantization, ensuring compact and meaningful brain network representations while mitigating over-smoothing and enhancing model generalization. The proposed framework effectively captures the dynamic and diverse nature of EEG signals, enabling robust emotion recognition. Extensive comparisons and analyses demonstrate the effectiveness of MIND-EEG, and the source code is publicly available at https://anonymous.4open.science/r/MIND_EEG.

Figures (6)

• Figure 1: Two key questions for utilizing spatial information in EEG signals.
• Figure 2: The framework of the proposed MIND-EEG. The Global State Encoder captures the global representation of the entire EEG signal. The Intra-Regional Functionality Encoder and Inter-Regional Interaction Encoder divide the EEG signal into regions, extracting functional features within regions and modeling inter-region interactions. Multi-granularity spatial information is integrated for emotion recognition, with all graph networks in the model constructed using a discrete codebook approach.
• Figure 3: Structure of the MAGCN.
• Figure 4: Discerte Codebook Construction.
• Figure 5: The parameter analysis of our proposed method on the SEED-IV dataset with subject-dependent scene. (a) Analysis of the sizes of three types of codebooks. (b) Analysis of the weight of the three codebook loss terms.