Published July 28, 2025 | Version v1
Dataset Open

Ear-SAAD dataset

  • 1. EDMO icon Katholieke Universiteit Leuven
  • 2. ROR icon Aarhus University

Description

Title: Ear-SAAD dataset: A Simultaneous Scalp, Around-Ear, and In-Ear EEG Dataset for Selective Auditory Attention Decoding

Description:

This dataset accompanies the publication:
Geirnaert S., Kappel S. L., Kidmose P., "A Direct Comparison of Simultaneously Recorded Scalp, Around-Ear, and In-Ear EEG for Neural Selective Auditory Attention Decoding to Speech," Scientific Reports, 2025. Please cite this paper and this Zenodo-dataset when using this dataset.

The Ear-SAAD dataset was designed to benchmark selective auditory attention decoding (AAD) performance across three EEG modalities: scalp EEG (29 channels + 3 EOG channels), around-the-ear EEG (19 channels), and in-ear EEG (12 channels). The goal is to support the development of neuro-steered hearing devices by comparing the decoding accuracy and signal quality of wearable EEG solutions against conventional scalp recordings.

Fifteen normal-hearing participants were recorded while attending to one of two simultaneously presented speech streams (simulated at ±60° using HRTFs), delivered via insertphones. Visual conditions (crosshair or video of the attended speaker) were balanced across trials. EEG was recorded for 60 minutes per participant using 6 10-min trials. The Danish speech stimuli consisted of lectures for high-school students from professors at Aarhus University.

This dataset was recorded at the Center for Ear-EEG, Department of Electrical and Computer Engineering, Aarhus University, in June 2024.

Dataset contents:

  • BIDS-compliant version of the raw dataset including EEG, events, and metadata (bids_dataset)

  • Preprocessed dataset used in the accompanying publication (preprocessedData)

  • A MATLAB-script that performs all preprocessing steps to go from the BIDS-compliant dataset to the preprocessed data (perprocessFromBids). Note that the output of this script might slightly differ from the provided proprocessed files due to a difference in ordering of cutting and filtering. This script can be used as an example of how to extract the different trials from the BIDS dataset.

  • An experiment manual, giving more details about the experiment and the applied procedures.

Citation:
Geirnaert S., Kappel S. L., Kidmose P., "A Direct Comparison of Simultaneously Recorded Scalp, Around-Ear, and In-Ear EEG for Neural Selective Auditory Attention Decoding to Speech," Scientific Reports, 2025.

Files

bids_dataset.zip

Files (12.6 GB)

Name Size Download all
md5:8e7231ef6c672b6a3612a55d3f886ed8
12.2 GB Preview Download
md5:26a41ea5110335e5f2b6090e2dd265f8
1.2 MB Preview Download
md5:b3cde4bf6ec08df5b8f4ed5118fcd6ba
415.6 MB Preview Download
md5:f9264b54f89b85f4a2cb8909c4a315a7
18.2 kB Download

Additional details

Funding

Research Foundation - Flanders
junior postdoctoral fellowship fundamental research for S. Geirnaert 1242524N
Research Foundation - Flanders
Travel grant for a long stay abroad for S. Geirnaert V418923N