DTU EEG Recordings of Attended Speech in Reverberant Rooms

18 subjects listened to competing speech streams in simulated rooms with varying reverberation. The dataset contains EEG recordings from a 64-channel Biosemi system sampled at 512 Hz, alongside the original speech audio stimuli. It is organized for the COCOHA Matlab Toolbox and was collected to study noise-robust cortical tracking of attended speech.

Use Cases

• Benchmarking temporal response function estimation methods based on aligned EEG and audio data.
• Studying the effect of acoustic reverberation on cortical speech tracking using the simulated room conditions.
• Developing auditory attention decoding algorithms using the labeled attended/ignored speaker trials.
• Analyzing EEG channel responses, including mastoid and EOG channels, to speech stimuli.

Strengths

• Includes recordings from 18 subjects, providing a moderate sample size for analysis.
• Contains both EEG data and the original speech audio stimuli, enabling multimodal analysis.
• Data was recorded in a controlled double-walled soundproof booth using a high-density 64-channel system.

Limitations

• Key metadata such as dataset size, row count, and last update date are not provided on any platform.
• The exact number of trials per subject or total recording duration is not specified.
• Column names and a complete data schema are not listed.

Provenance

Source

Technical University of Denmark (DTU)

Collection Method

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

Geography

Recorded at DTU, Denmark.