Speech decoding resources

On this page, you can find an overview of publicly available speech decoding datasets, that generally contain neural data of subjects listening to natural speech. This can be either with multiple competing talkers (to decode selective auditory attention) or with a single talker.

In case your publicly available dataset is not mentioned, please contact me.

Currently, there is about:

• 148 hours of EEG with competing speech
• 8 hours of MEG with competing speech
• 10.5 hours of EEG with polyphonic music
• 273 hours of EEG with single talker speech
• 71 hours of MEG with single talker speech

publicly available.

Multi-talker (selective attention)

Single talker