Music identification from brain activity during listening

Music identification from brain activity measured while a participant listened to music.

First author; training-signal design, PyTorch training/evaluation, writing, code release, and model-weight release.

0.859 best single-model accuracy and 0.887 three-model ensemble accuracy on NMED-T song identification.

PyTorch, PyTorch Lightning, Transformer encoder for EEG signals, masked prediction, MuQ, MusicGen, NMED-T, NVIDIA RTX A6000.