Abstract
Music Information Retrieval (MIR) is an established field that provides solutions to analyze, retrieve, classify, recommend, or visualize music. From many applications, however, only a fraction will provide results that are meaningful for musicians.
Scoring functions may use descriptors that are difficult to describe for the end-user, and visualizations are often bound to signal aspects of music such as displaying waveform. We have developed a music analysis system which is based on music theory and contains visualizations meaningful for those interested in harmony aspects of music.
Music is first segmented to chords by known techniques, providing the basis that musicians understand. From there, distances between chords are evaluated by a novel approach and new descriptors are formed, based on recent music theory studies.
End-user can visualize the musical piece and find interesting sequences in a color temperature graph. While having the music visualized, user can retrieve similar musical pieces, understanding the similarity between the chord progressions.