EEG and Audio Dataset for Auditory Attention Decoding from 18 Subjects

Søren A. Fuglsang and colleagues at the Technical University of Denmark created a dataset containing EEG recordings from 18 subjects listening to competing speech audio streams in simulated rooms with varying reverberation. The data includes 64-channel EEG, EOG signals, and corresponding audio wav-files, organized with the COCOHA Matlab Toolbox format. It was published in associated studies from 2017 and 2018.

Use Cases

• Train auditory attention decoding models based on EEG responses to attended versus ignored speech.
• Compare temporal response function estimation methods using aligned EEG and audio time-series data.
• Study the impact of acoustic room conditions (anechoic, mild, high reverberation) on cortical speech tracking.
• Benchmark EEG preprocessing pipelines for noise-robust neural signal analysis.

Strengths

• Includes data from 18 human subjects, providing a basis for inter-subject analysis.
• Contains 64-channel scalp EEG plus 8 EOG/mastoid channels, recorded at a 512 Hz sampling rate.
• Experimental conditions are documented, including attended speaker, spatial position, and acoustic room type.
• Provides both raw and preprocessed data versions, with example scripts for alignment and preprocessing.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.
• Last update date is unknown; freshness unverified.

Provenance

Source

Technical University of Denmark (DTU), COCOHA Project.

Collection Method

Recorded in a double-walled soundproof booth using a 64-channel Biosemi system.