16 channel Three speaker dynamic switch cEEGrid Auditory Attention Decoding Dataset Nanjing University

The preprocessed EEG data are stored in .npy format, each trial a file, with the naming 'dataset-001-subject-xxx-trial-yyy' The metadata to describe each trial is stored in a single pickle file, in which a dict with key 'dataset-001-subject-xxx-trial-yyy' to the corresponding values The original audio stimuli and the raw trial info are provided in .wav files and .mat files. You shall derive the enveloep from the .wav files, with the stimuli file name specified in the raw trial info. Alternatively, you may contact us for transfering the original stimuli, alongside with the processed envelope and mel spectrogram through Internet or ship a hard disk to you. In the later case, you shall pay the ship fee and provided an unpacked brand new hard disk with at least 250GB to us. The processed stimuli file (compressed) is approximately 200GB (that is to say, uploading to zenodo is almost impossible, and transfering through Internet is also time-consuming, depend on the network status). The EEG data is preprocessed, and other information is well formatted. Anyone with basic MATLAB and Python programming skill and common knowledge of EEG can directly look up files to find out trial infomation. If you struggle to find out some, please first take a care look or ask experienced persons around you for help. If still cannot find, please feel free to contact us. Furhter information about data collection and preprocessing will be available once preprint is ready. We strongly recommend you using our developed python package, SuperHugeAAD, and SuperPrepare, to start your AAD pipeline immediately. EEG data of some subjects are not provided and reserved for other purpose. These data will be published in future, and is temporarily unavailable in any cases.