NSF REU: Computational Methods for Music, Media, and Minds at the University of Rochester. Faculty Mentors: Prof. Zhiyao Duan (ECE), Prof. Matthew Brown (Music) Collaborators: Christodoulos Benetatos, Christopher Winders, Ph. D.
Onset during polyphonic sections is artificially "louder" than monophonic due to normalization going over entire chromagram once filled. Probably need to normalize frame by frame rather than waiting until end of processing.
not all information from musicxml files with multiple parts is being processed properly. some information is missing from the final chromagram, i believe that the information is not being accessed properly (maybe stored deeper than monophonic information) and so the chromagrams are being left empty at parts where data is not reached.