raei/Khanday-2026-EEGFoundationModelsSpeechDecodingBenchmark

UGR-MINDVOICE

This is an unofficial mirror of the public UGR-MINDVOICE OSF project. It is not affiliated with or endorsed by the dataset contributors, their institutions, OSF, or the authors of the benchmark paper.

Source and provenance

• Original dataset: OSF project 6sh5d
• Source snapshot: retrieved September 3, 2026; OSF project last modified December 19, 2025
• Dataset article: Cortina et al., Computer Speech & Language 100 (2026), 101964
• Benchmark paper: Khanday et al., arXiv:2607.27268v1
• Official code: owaismujtaba/mind-voice
• Mirror-generated checksums: MIRROR_SHA256SUMS.txt

The 533 source files are mirrored without modifying their contents or source directory layout. This dataset card and the checksum manifest are the only mirror-added files. The snapshot totals 22,025,659,310 bytes (20.51 GiB).

Dataset summary

UGR-MINDVOICE is a multimodal corpus of 64-channel EEG sampled at 1,000 Hz and synchronized audio from 15 native Iberian Spanish speakers across 22 sessions. Each session includes overt and covert speech in alternating mini-blocks. Stimuli cover all Spanish phonemes through 95 consonant-vowel syllables, 60 words balanced across six semantic categories and lexical frequencies, and 30 pseudowords. Overt productions include synchronized acoustic recordings.

The Data/ directory is organized according to BIDS and contains the participant and recording data. Materials/ contains the stimuli and experiment files, while Analyses and Results/ contains analysis documentation.

Human-participant data

These are human EEG and voice recordings. Users are responsible for applicable institutional, ethical, and legal requirements and must not attempt to identify participants.

License

No dataset license is declared in the OSF project metadata or supplied data files. The CC BY-NC-SA 4.0 license on the arXiv benchmark and the CC BY 4.0 license on the dataset article apply to those publications; they do not by themselves grant rights in the dataset. This mirror makes no additional reuse grant. Contact the dataset authors if permission is required for your use.

Citation

Vales Cortina, I., Khanday, O. M., Ouellet, M., Pérez-Córdoba, J. L., Rodríguez-San Esteban, P., Miccoli, L., Galdón, A., Olivares Granados, G., & Gonzalez-Lopez, J. A. (2026). UGR-MINDVOICE: A multimodal EEG-audio dataset for overt and covert Iberian Spanish speech production. Computer Speech & Language, 100, 101964. https://doi.org/10.1016/j.csl.2026.101964

@article{cortina2026ugrmindvoice,
  title   = {{UGR-MINDVOICE}: A Multimodal EEG-Audio Dataset for Overt and Covert Iberian Spanish Speech Production},
  author  = {Vales Cortina, Ibon and Khanday, Owais Mujtaba and Ouellet, Marc and P\'{e}rez-C\'{o}rdoba, Jos\'{e} L. and Rodr\'{i}guez-San Esteban, Pablo and Miccoli, Laura and Gald\'{o}n, Alberto and Olivares Granados, Gonzalo and Gonzalez-Lopez, Jose A.},
  journal = {Computer Speech \& Language},
  volume  = {100},
  pages   = {101964},
  year    = {2026},
  doi     = {10.1016/j.csl.2026.101964}
}