Scholarlink

<p>Accurate recognition of human emotions and mental activities is essential for affective computing, healthcare, and human-computer interaction. This study introduces EEG-EmotionActivityNet, a deep learning framework that simultaneously classifies emotions and human mental activities from EEG signals. A novel Convolutional Neural Network (CNN) extracts discriminative spatial-temporal EEG features, while an ensemble classifier combining Transformer and Bidirectional Long Short-Term Memory (Bi-LSTM) captures both global and sequential patterns for robust classification. The EEG-EmotionActivityNet was trained and tested on the DEAP dataset. The findings demonstrate that EEG-EmotionActivityNet achieved better results for emotion classification (Valence: 92.1%, Arousal: 92.5%, Dominance: 91.5%, Liking: 93%) and for human activity classification across High Arousal High Valence (HAHV), High Arousal Low Valence (HALV), Low Arousal High Valence (LAHV), and Low Arousal Low Valence (LALV) tasks. In conclusion, EEG-EmotionActivityNet provides a unified and effective solution, bridging the gap between human emotion and activity classification.</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-EmotionActivityNet: a deep learning approach for emotion and human activity classification

  • 期刊论文

  • 发表日期：

    2025年12月19日

查看全文
  • Sheetal Patil,
  • Rudragoud Patil

摘要

Accurate recognition of human emotions and mental activities is essential for affective computing, healthcare, and human-computer interaction. This study introduces EEG-EmotionActivityNet, a deep learning framework that simultaneously classifies emotions and human mental activities from EEG signals. A novel Convolutional Neural Network (CNN) extracts discriminative spatial-temporal EEG features, while an ensemble classifier combining Transformer and Bidirectional Long Short-Term Memory (Bi-LSTM) captures both global and sequential patterns for robust classification. The EEG-EmotionActivityNet was trained and tested on the DEAP dataset. The findings demonstrate that EEG-EmotionActivityNet achieved better results for emotion classification (Valence: 92.1%, Arousal: 92.5%, Dominance: 91.5%, Liking: 93%) and for human activity classification across High Arousal High Valence (HAHV), High Arousal Low Valence (HALV), Low Arousal High Valence (LAHV), and Low Arousal Low Valence (LALV) tasks. In conclusion, EEG-EmotionActivityNet provides a unified and effective solution, bridging the gap between human emotion and activity classification.