Student wearing a piece of equipment on her head.

Research

Neural Decoding of Imagined Music

A machine learning framework for EEG-based audio feature reconstruction

Claire Southard

Musicians with movement disorders such as Parkinson's disease or musician's dystonia often lose the fine motor control required to perform even though their ability to imagine music remains unchanged.

Master's thesis: Claire Southard, SM '25

Claire Southard’s master’s thesis explored whether imagined music could be interpreted from noninvasive electroencephalogram (EEG) brain recordings by identifying features such as rhythm and pitch directly from brain activity.

She developed computational models that learned the relationship between patterns in brain signals and musical features, allowing elements of imagined music to be reconstructed. By combining machine learning, neuroscience, and music technology, her research represents an early step toward helping musicians with physical limitations to continue creating music—if not with their hands, then with their minds.