KUL-AAD-NFB: Dataset of Auditory Attention Decoding with Neurofeedback, KU Leuven

If you use this dataset for your research, please cite this Zenodo repository, the accompanying article describing this dataset [1] (please check the most updated version at the time of citing), as well as the media broadcaster of the audio stimuli used [2]. [1] Rotaru, I., Geirnaert, S., Heintz, N., Bertrand, A., & Francart, T. (2026). When feedback backfires: investigating neurofeedback effects in a closed-loop auditory attention decoding paradigm. Submitted. bioRxiv, 2026-04 https://doi.org/10.64898/2026.04.28.721343 [2] Universiteit van Vlaanderen. Science outreach presentations. https://www.universiteitvanvlaanderen.be/college Overview This dataset contains electroencephalogram (EEG) data collected during a neurofeedback auditory attention decoding (AAD) experiment. The research was conducted at ExpORL, Dept. Neurosciences, KU Leuven and Dept. Electrical Engineering (ESAT), KU Leuven (Belgium), with the goal of investigating short-term neurofeedback effects in a closed-loop interaction between a user and a real-time AAD system. Dataset Summary: Participants: 19 Flemish-speaking adults with normal hearing 2/19 subjects were re-tested in a second session only 11/19 subjects are included in this online dataset (cf. data sharing consents) Stimuli: audio podcasts from Universiteit van Vlaanderen (Flemish) Total Duration: Subjects 1-10, ses-01 and subjects 8-9, ses-02: 56 or 64 min of EEG data / subject Subjects 11-19, ses-01: 64 min of EEG data + 20 min calibration data Data: 64-channel raw EEG (.bdf files) at 1024 Hz 64-channel preprocessed EEG (64 Hz), parameters and metadata (.npy files) Audio stimuli (.wav files) Behavioral responses (.csv file) Overview of participants (.xlsx file) The full dataset contains EEG data collected from 19 normal-hearing subjects, during a competing listening auditory attention decoding (AAD) task, where the subjects were instructed to focus on one of two competing speech signals while they received real-time feedback based on their decoded brain activity. Note that 9 subjects were excluded from the online repository due to not consenting to sharing their data in a public database (cf. signed informed consents approved by KU Leuven Ethical Committee). EEG recordings were conducted in a soundproof, electromagnetically shielded room at ExpORL, KU Leuven. The BioSemi ActiveTwo system was used to record 64-channel EEG signals at 1024 Hz sample rate. Experimental protocol The experimental trials were split into 2 blocks. Each block consists of the following sequence of conditions including different feedback modalities (visual or auditory): OL, CLV, CLA, psCLA. The conditions order was pseudo-randomized across participants. Each trial lasted for 8 min. Below is a brief overview of the conditions: OL: open-loop, no feedback was received (baseline) CLV: closed-loop with visual feedback (moving slider) CLA: closed-loop with auditory feedback (changing gains for both attended/unattended speakers) psCLA: closed-loop with auditory pseudo-feedback (the feedback was computed based on EEG data from another trial) For subjects 11-19, four additional OL-CAL trials (4 x 5 min) were recorded for calibration, i.e., to train a subject-specific decoder. All trials (except OL-CAL) contain either 1 or 3 switches in attention enforced at predetermined times, whereby the participant had to shift their attention from the L to the R speaker or vice-versa, as cued by a visual arrow displayed on the screen. The full description of the experimental conditions and protocol can be consulted in [1]. Further remarks: The steps for data preprocessing are described in detail in [1]. Due to technical problems with triggers during data collection, the following trials are missing from the dataset: sub-01, ses-01, CLV trial6 sub-08, ses-01, CLA trial10 sub-12, ses-01, psCLA trial 11 sub-17, ses-01, psCLA trial 14 sub-19, ses-01, CLV trial 7 Acknowledgements This research was funded by the Research Foundation Flanders (SBO mandate 1S14922N for I. Rotaru, junior postdoctoral mandate 1242524N for S. Geirnaert and FWO projects G081722N, G026026N), the Flemish Government (AI Research Program), Internal Funds KU Leuven (projects C3/25/017, IDN/23/006 and C14/25/108), and the European Research Council (ERC) under the European Union’s research and innovation program (grant agreement No 101138304). We also thank the participants for their time and effort in the experiments. Contact Information Executive researcher: Iustina Rotaru, [e-mail] Led by: prof. Tom Francart, [e-mail]