Neural Noise- A Nature-published EEG imagined-speech dataset enables direct comparison of BCI paradigm designs.1
- The dataset directly supports speech prosthesis and EEG–BCI R&D and is suitable for immediate reuse in decoding pipelines.1 1
Weekly enrichment (2026-07-20)
- The dataset (Aguilera-Rodríguez et al., 2025, Scientific Data) recorded EEG from 15 healthy volunteers (7 female, 8 male) performing imagined speech.23
- Each participant completed two paradigms—a traditional cue-based abstract paradigm and a user-centered gamified video-game (Pac-Man-style maze) paradigm—to test whether engagement reduces fatigue and boosts attention.23
- Four Spanish directional commands were imagined: avanzar, retroceder, derecha, and izquierda, framing a four-class decoding problem.43
- Recordings used a 24-electrode mBrainTrain Smarting system sampled at 500 Hz, with a trial duration of about 11.8 s and roughly 30 repetitions per prompt.3
- The release comprises on the order of 1,800 trials across two sessions per subject, with both raw and EEGLAB-preprocessed data provided.43
- The gamified paradigm added a vocalized-speech step after each imagined instance, enabling imagined-versus-vocalized comparisons; trials flagged with vocalization errors are not recommended for classifier training.4
- Questionnaires captured Internal Representations, Self-Assessment Manikin emotional state (before/after), and User Experience across attractiveness, perspicuity, efficiency, dependability, stimulation, and novelty.4
- Data are released under a CC-BY-NC-ND-4.0 license via Mendeley Data and are accessible through the Mother of All BCI Benchmarks (MOABB) library as the AguileraRodriguez2025 dataset.43