Pitch content visualization tools for music performance analysis

Abstract

This work deals with pitch content visualization tools for the analysis of music performance from audio recordings. An existing computational method for the representation of pitch contours is briefly reviewed. Its application to music analysis is exemplified with two pieces of non-notated music: a field recording of a folkloric form of polyphonic singing and a commercial recording by a noted blues musician. Both examples have vocal parts exhibiting complex pitch evolution, difficult to analyze and notate with precision using Western common music notation. By using novel time-frequency analysis techniques that improve the location of the components of a harmonic sound, the melodic content representation implemented here allows a detailed study of aspects related to pitch intonation and tuning. This in turn permits an objective measurement of essential musical characteristics that are difficult or impossible to properly evaluate by subjective perception alone, and which are often not accounted for in traditional musicological analysis. Two software tools are released that allow the practical use of the described methods.