EEG Dataset for 'Decoding of selective attention to continuous speech from the human auditory brainstem response' and 'Neural Speech Tracking in the Theta and in the Delta Frequency Band Differentially Encode Clarity and Comprehension of Speech in Noise'.

Update: a more accessible version of this dataset is now available: https://doi.org/10.5281/zenodo.7778289

The repository contains the unprocessed EEG data recorded for the publications [1, 2]. The stimuli are also included. 64-channel EEG was recorded from 18 participants whilst they listened to audiobooks narrated in a number of listening conditions: clean speech in their native language (English), clean speech in three levels of speech-shaped-noise, and two competing-speakers conditions (focusing on the males speaker whilst ignoring the female speaker, and vice-versa). Additionally, 12 of the participants listened to speech in a foreign language (Dutch, for which their comprehension scores were 0, as measured via behavioural experiments), in four listening conditions: noiseless conditions, and speech-in-babble-noise conditions at three different SNRs. For more information on the recording protocol and stimuli, please refer to the publications [1 ,2].

The EEG was acquired at 1 kHz via the actiCHamp amplifier (BrainProducts, Germany), and the electrodes were positioned according to the standard 10-20 system via to EasyCap electrode cap (BrainProducts, Germany). To align the stimuli with the EEG recordings, the audio was simultaneously recorded with the EEG (also at 1 kHz), using the StimTrack device (BrainProducts, Germany). The resulting sound channel was then cross correlated with the (resampled) audio data. The resulting stimulus-onset timestamps are stored as annotations in the VHDR files. The physical EEG reference was located at P04, and the ground was located on the right earlobe.

The raw VHDR files containing the EEG data for each trial are located in the folder ‘eeg’. The stimulus files are located in the folder stim. For convenience, the raw data are also provided in a format similar to the CND data format [3]. The CND-format files can be found in dataCND. The EEG data for each trial have been extracted so that they align with the raw audio data in dataCND/trialStories.mat. We provide the time-aligned broadband envelopes at 1 kHz in the file dataCND/stim.mat.

Please note some details on the raw speech-in-noise and Dutch stimuli:

For the ENGLISH speech-in-noise conditions, the babble noise amplitude increased linearly from zero to the desired SPL for 0.5 seconds. The babble noise continued to play for an additional 0.5 seconds at this SPL, before the story began. After the story finished, the babble noise was again played for 0.5 seconds, and then linearly decreased to 0 amplitude for 0.5 seconds. Therefore, the babble noise track is two seconds longer than the story track.

For the DUTCH speech-in-noise conditions, the dutch speech was played by itself for one second. Then, the babble noise was also played for one second, linearly increasing in amplitude until it reached the desired SPL. Therefore, the babble noise track is one second shorter than the Dutch story track.

The relevant tracks in dataCND/stim.mat were padded with zeros so that they are aligned and their lengths are equal.

In all of the Dutch conditions, a few English sentences were embedded in the Dutch narratives in order to help participants maintain their attention. Participants were asked behavioural questions related to the English sentences after each trial. These English sentence tracks are also time-aligned and provided in dataCND/stim.mat.

If you use this data, please cite the original publications, as well as this repository [1,2, 4].

[1] Etard O, Kegler M, Braiman C, Forte A E and Reichenbach T. “Decoding of selective attention to continuous speech from the human auditory brainstem response” 2019. NeuroImage 200 1–11

[2] Etard O and Reichenbach T. “Neural speech tracking in the theta and in the delta frequency band differentially encode clarity and comprehension of speech in noise” 2019. J. Neurosci. 39 5750–9

[3] Giovanni DL and Nidiffer, A. "The Continuous-event Neural Data structure (CND) Specifications and guidelines". 2022 Jul. https://data.cnspworkshop.net/CND_Specifications.pdf

[4] Etard O and Reichenbach T. "EEG Dataset for 'Decoding of selective attention to continuous speech from the human auditory brainstem response' and 'Neural Speech Tracking in the Theta and in the Delta Frequency Band Differentially Encode Clarity and Comprehension of Speech in Noise". Doi: 10.5281/zenodo.7086208