Cueless EEG Imagined Speech for Subject Identification: Dataset and Benchmarks
A recent study presents an EEG-based imagined speech paradigm for biometric identification that does not rely on cues, requiring participants to visualize saying semantically meaningful words. The research includes a dataset featuring more than 4,350 trials collected from 11 individuals over five sessions. It assesses different classification techniques, such as traditional machine learning methods (SVM, XGBoost) and deep learning models, setting benchmarks for this innovative method.
Key facts
- The study introduces a cueless EEG-based imagined speech paradigm for subject identification.
- Subjects imagine pronouncing semantically meaningful words without external cues.
- The dataset includes over 4,350 trials from 11 subjects across five sessions.
- Classification methods assessed include SVM, XGBoost, time-series foundation models, and deep learning architectures.
- The paper is available on arXiv with identifier 2501.09700.
- The study addresses limitations of prior methods that relied on visual or auditory cues.
- The approach requires subjects to select and imagine words from a predefined list naturally.
- The research is in the field of biometric identification using EEG signals.
Entities
Institutions
- arXiv