EEG-based detection of the locus of auditory attention with convolutional neural networks

This repo contains the code for EEG-based detection of the locus of auditory attention with convolutional neural networks.

License

See the LICENSE file for license rights and limitations. By downloading and/or installing this software and associated files on your computing system you agree to use the software under the terms and condition as specified in the License agreement.

Requirements

• MATLAB 2016b
• MatConvNet 1.0-beta25

Setup

• Download MatConvNet 1.0-beta25
• Extract matconvnet-1.0-beta25 to the base directory (see Directory Structure)
• Install MatConvNet 1.0-beta25 (see Installing and compiling the library)
• Download the "Auditory Attention Detection Dataset KULeuven" dataset from https://zenodo.org/record/3377911 (Skip the stimuli - this model only uses EEG)
• Extract the dataset to data/raw/AAD2015_128hz/

How to run

Run src/main.m

• Trained models are saved in models/<name of experiment>/
• Results are saved in results/<name of experiment>/

Directory Structure

• data/
  • raw/
    • AAD2015_128hz/: dataset (not included)
      • S1.mat
      • ...
      • S16.mat
• matconvnet-1.0-beta25/: MatConvNet source files (not included)
• models/: trained models, in MatConvNet format. Each MAT-file contains a struct with weights and with various metadata.
  • paper/: models used to generate the results in the paper
• results/: experiment results
• src/
  • +data/: data loading and preprocessing code
  • +model/: model initialization, training, and testing code
  • +utils/: various utility functions