Reconstruction of Imagined Melody with Relative Pitch Decoding in Electrocorticography

Area of Science:

• Neuroscience
• Cognitive Science
• Music Perception

Background:

• Music imagery involves internal auditory experiences, but decoding imagined melodies from neural data is difficult.
• Previous research focused on detecting music imagery or decoding broad musical features, not melodic structure.

Purpose of the Study:

• To investigate if imagined melodies can be decoded from human electrocorticography (ECoG) recordings.
• To explore the utility of a relative pitch framework for decoding imagined melodic content.

Main Methods:

• Ten epilepsy patients undergoing music imagery tasks with familiar melodies in various tonalities.
• Intra-subject models trained on neural features from imagery-responsive cortical sites.
• Decoding relative pitch classes at the single-note level using electrocorticography (ECoG) data.

Main Results:

• Single-note decoding accuracy exceeded chance, though modest.
• Sequential integration of decoded pitch classes reconstructed melodic contours.
• Preservation of the relative structure of imagined melodies was achieved.

Conclusions:

• Imagined melodic structure can be accessed from intracranial neural recordings using a relative pitch approach.
• Sequence-level analysis is crucial for decoding internally generated auditory content.
• Demonstrates a pathway for decoding complex imagined auditory experiences from neural signals.