Datasets for the EEG-Driven Brain–Computer Musical Interfaces for Emotion Self-Induction Ph.D. Thesis

This thesis contributes four publicly available datasets collected from 82 participants across sequential BCMI experiments. Each dataset corresponds to a distinct emotion prediction paradigm. The AFAH dataset (23 participants, 14 min) contains dual-channel EEG recorded at 1000 Hz alongside valence– arousal predictions. The MLP dataset (23 participants, 20 min) includes labeled training data and trained models. The EEGNet dataset (26 participants, 30 min) provides six synchronized data streams including EEG at 100 Hz, real-time emotion predictions, and subjective self-reports. The LDA dataset (33 participants, 30 min) follows a similar structure with EEG at 1000 Hz. In total, the datasets comprise more than 50 hours of synchronized recordings, including pre- and post-experiment questionnaires, and also python scripts to extract the data from XDF files and R scripts to analyze the data.

The software used to capture the data is available at https://github.com/pamonroy/aBCMI