Scholarlink

<p>Emotions are very valuable natural human characteristics. Emotion recognition is frequently utilized in human-computer interfaces (HCI) to aid individuals with impairments. Electroencephalogram (EEG) signals are essential for identifying emotional states, as they rapidly reflect changes in brain activity. In this work, the usefulness of the Variational mode decomposition (VMD) for detecting diverse emotions within EEG is explored. VMD partition the EEG into band-limited intrinsic mode functions (BIMFs) and extracts the bandwidth of amplitude modulation (BWAM), bandwidth of frequency modulation (BWFM), and spectral features from the BIMFs. The extracted features are fed to a grid search cross-validation (GSCV) optimized support vector machine(SVM) classifier, which identifies different emotional states. When compared to other published works, the experimental findings of the proposed approach show superior emotion recognition (ER) performance on the SEED, SEED-IV, and DEAP datasets. The proposed emotion recognition system achieved maximum accuracies of 95.8%, 95.6%, and 94.1% on these databases, respectively, surpassing the state-of-the-art (SoA) performance.</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

Emotion recognition using VMD domain bandwidth and spectral features

  • 期刊论文

  • 发表日期：

    2025年1月19日

查看全文
  • Avinash Vujji,
  • Nalini Pusarla,
  • Anurag Singh,
  • Shrivishal Tripathi

摘要

Emotions are very valuable natural human characteristics. Emotion recognition is frequently utilized in human-computer interfaces (HCI) to aid individuals with impairments. Electroencephalogram (EEG) signals are essential for identifying emotional states, as they rapidly reflect changes in brain activity. In this work, the usefulness of the Variational mode decomposition (VMD) for detecting diverse emotions within EEG is explored. VMD partition the EEG into band-limited intrinsic mode functions (BIMFs) and extracts the bandwidth of amplitude modulation (BWAM), bandwidth of frequency modulation (BWFM), and spectral features from the BIMFs. The extracted features are fed to a grid search cross-validation (GSCV) optimized support vector machine(SVM) classifier, which identifies different emotional states. When compared to other published works, the experimental findings of the proposed approach show superior emotion recognition (ER) performance on the SEED, SEED-IV, and DEAP datasets. The proposed emotion recognition system achieved maximum accuracies of 95.8%, 95.6%, and 94.1% on these databases, respectively, surpassing the state-of-the-art (SoA) performance.