When listening to music, the brain generates a neural response that follows the amplitude envelope of the musical sound. Previous studies have shown that it is possible to decode this envelope-following response from electroencephalography (EEG) data during music perception. However, a successful decoding and recognition of imagined music, without the physical presentation of a music stimulus, has not been established to date. During music imagination, the human brain internally replays a musical sound, which naturally leads to the hypothesis that a similar envelope-following response might be generated. In this study, we demonstrate that this response is indeed present during music imagination and that it can be decoded from EEG data. Furthermore, we show that the decoded envelope allows for classification of imagined music in a song recognition task, containing tracks with lyrics as well as purely instrumental tasks. A two-song classifier achieves a median accuracy of 95%, while a 12-song classifier achieves a median accuracy of 66.7%. The results of this study demonstrate the feasibility of decoding imagined music, thereby setting the stage for new neuroscientific experiments in this area as well as for new types of brain-computer interfaces based on music imagination.
bioRxiv Subject Collection: Neuroscience