Scholarlink

<p>Emotion recognition using electroencephalogram (EEG) signals is pivotal for elucidating the functional dynamics of the human brain. Two primary approaches dominate the emotion recognition: computational dimensional emotion models and methods based on machine learning (ML) and deep learning (DL). This study proposes a novel framework integrating fuzzy linguistic summarization (FLS) with a modified version of Russell’s circumplex model augmented by fuzzy sets. The proposed method utilizes four frontal alpha/beta-based formulas to compute arousal and valence, and applies three types of fuzzy membership functions—triangular, trapezoidal, and gaussian- to facilitate emotion recognition. These fuzzy emotion labels are incorporated into FLS, which enables classification based on their degrees of truth. Furthermore, a modified multi-granular trend detection (GTD) algorithm is employed during preprocessing to enhance result robustness. The methodology was evaluated&#xa0;using the SEED (SJTU Emotion EEG Dataset)<b>,</b>&#xa0;achieving a peak accuracy of 80% with Formula <InternalRef RefID="Equ3">3</InternalRef> and the triangular membership function. This performance outperforms most ML-based approaches in the literature and remains competitive with DL models while mitigating challenges such as high computational cost, data dependency, and limited interpretability<b>.</b> As a pioneering study, applying a computational dimensional model to the widely used SEED from a FLS perspective, this work is poised to inspire future research in EEG-based emotion recognition.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

EEG-Based Emotion Classification Driven by Fuzzy Linguistic Summarization: A Study on SEED Dataset

  • 期刊论文

  • 发表日期：

    2025年10月24日

查看全文
  • Ümran Kaya,
  • Diyar Akay

摘要

Emotion recognition using electroencephalogram (EEG) signals is pivotal for elucidating the functional dynamics of the human brain. Two primary approaches dominate the emotion recognition: computational dimensional emotion models and methods based on machine learning (ML) and deep learning (DL). This study proposes a novel framework integrating fuzzy linguistic summarization (FLS) with a modified version of Russell’s circumplex model augmented by fuzzy sets. The proposed method utilizes four frontal alpha/beta-based formulas to compute arousal and valence, and applies three types of fuzzy membership functions—triangular, trapezoidal, and gaussian- to facilitate emotion recognition. These fuzzy emotion labels are incorporated into FLS, which enables classification based on their degrees of truth. Furthermore, a modified multi-granular trend detection (GTD) algorithm is employed during preprocessing to enhance result robustness. The methodology was evaluated using the SEED (SJTU Emotion EEG Dataset), achieving a peak accuracy of 80% with Formula 3 and the triangular membership function. This performance outperforms most ML-based approaches in the literature and remains competitive with DL models while mitigating challenges such as high computational cost, data dependency, and limited interpretability. As a pioneering study, applying a computational dimensional model to the widely used SEED from a FLS perspective, this work is poised to inspire future research in EEG-based emotion recognition.