Data from: Nonlinear decoding models enable music reconstruction from human auditory cortex activity

This dataset is associated with the manuscript "Nonlinear decoding models enable music reconstruction from human auditory cortex activity", and provides all preprocessed files necessary to replicate the results.

In this study, we recorded intracranial EEG data (specifically, ECoG) in 29 patients with pharmacoresistant epilepsy while they were passively listening to a Pink Floyd song.

The present dataset consists of preprocessed neural activity (High-Frequency Activity, 70-150 Hz), electrode coordinates (in MNI template space) and auditory stimulus (raw wave file, and 32- and 128-frequency-bin auditory spectrogram). HFA and both auditory spectrograms have a sampling rate of 100 Hz, and are temporally aligned (duration of 190.72 s).

The code we used to preprocess and analyze the data is hosted on GitHub, here for the manuscript and there for the predictive modeling functions.