Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
June 6, 2026Journal of the Korea Society of Computer and Information

EEG Super-Resolution in PSD Feature Space: Bridging Wearable and Clinical-Grade Emotion Recognition

View Full Paper
Ask AI
Bookmark
Share

Key result

Transformer-based EEG super-resolution with uniform channel selection outperforms standard strategies for emotion recognition.

  • P<0.01

Why the study?

High-density EEG systems offer superior spatial resolution for emotion recognition but are impractical due to high cost, complexity, and discomfort.

Population

SEED-IV dataset

Comparison

Uniform channel selection strategy vs Clinical Standard and Hardware Realistic strategies

Authors

DBDuyong BaekSLSeok-Won Lee

Discussion

Loading...

Member takes

Implication

May enable low-channel wearable EEG emotion research; leaves open clinical validation before practice change.

Key Points

  • This research aims to develop a super-resolution EEG paradigm using power spectral density features to enhance emotion recognition capabilities.
  • Proposed a transformer-based architecture for super-resolution in the PSD feature space.
  • Evaluated on SEED-IV dataset with varying channel inputs (8, 14, 16, 32 channels) to restore high-density PSD features (62 channels).
  • Conducted statistical analysis comparing uniform channel selection with clinical and hardware strategies.
  • Achieved an accuracy retention rate of 78.45% for 32 channels and 77.92% for 16 channels against the baseline.
  • Demonstrated a maximum accuracy retention of 97.56% under the proposed method.
  • Uniform channel selection strategy significantly outperformed clinical standards (p<0.01, Bonferroni adjusted).

Structured PICO

P
Population
SEED-IV dataset (EEG data for emotion recognition)
I
Intervention
Transformer-based EEG super-resolution architecture operating in power spectral density (PSD) feature space to reconstruct 62-channel high-density PSD features from 8, 14, 16, or 32 sparse channels
C
Comparator
Clinical Standard and Hardware Realistic channel selection strategies
O
Outcome
Accuracy preservation rate of emotion recognitionsurrogate

Main Result

p-value: p=<0.01

A Transformer-based EEG super-resolution model operating in PSD feature space enables clinical-grade emotion recognition using low-channel wearable EEG data without requiring subject-specific high-channel data collection.

Cite This Study

Baek et al. (2026) studied Emotion recognition. Transformer-based EEG super-resolution in PSD feature space vs. Clinical Standard and Hardware Realistic strategies was evaluated on Accuracy preservation rate (p=<0.01). Transformer-based EEG super-resolution using a Uniform channel selection strategy significantly outperformed Clinical Standard and Hardware Realistic strategies for emotion recognition (p<0.01).

synapsesocial.com/papers/6a23baa771a5da9775e7650dhttps://doi.org/10.9708/jksci.2026.31.05.041
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1EEG emotion recognition method integrating improved dynamic graph convolutional neural networks and bidirectional simple recurrent units2026
  2. 2Modular Flexible 80-dB-DR Artifact-Resilient EEG Headset with Distributed Pulse-Based Feature Extraction and Multiplier-Less Neuromorphic Boosted Seizure Classifier2024 · 5 citations
  3. 3Effects of feature reduction on emotion recognition using EEG signals and machine learning2024 · 17 citations
  4. 4EEG-based emotion recognition using super-resolution superlet transform and self-attention convolutional neural network2026
  5. 5Real-Time Seizure Detection Using Behind-the-Ear Wearable System2024 · 9 citations