neuralinterfacinglab/SingleWordProductionDutch

Scripts to work with an intracranial EEG dataset of speech production.

SingleWordProductionDutch

Scripts to work with the intracranial EEG data from here described in this article.

Dependencies

The scripts require Python >= 3.6 and the following packages

• numpy
• scipy
• scikit-learn
• pandas
• pynwb

Repository content

To recreate the experiments, run the following scripts.

• extract_features.py: Reads in the iBIDS dataset and extracts features which are then saved to './features'

• reconstruction_minimal.py: Reconstructs the spectrogram from the neural features in a 10-fold cross-validation and synthesizes the audio using the Method described by Griffin and Lim.

• viz_results.py: Can then be used to plot the results figure from the paper.

• reconstuctWave.py: Synthesizes an audio waveform using the method described by Griffin-Lim

• MelFilterBank.py: Applies mel filter banks to spectrograms.

• load_data.mat: Example of how to load data using Matlab instead of Python