Measuring Progress in Music Rehearsals

Practice is an essential part of learning and performing a piece of music. The goal of this thesis is to develop a system for analysing audio recordings of music rehearsals that aids the musician in their studies. We gather audio recordings of rehearsals from pianists and develop two key measures based on these recordings: 1. match distribution, which indicates how often a segment of a recording has been practised in other recordings and 2. cover analysis, a measure of similarity between two recordings. Our analysis is based on audio matching: comparing segments of audio that are similar to each other. These analyses are used to create Musicmeasure: a system where users can upload their recordings and interact with the analyses. The system is evaluated through a qualitative interview with one of the participants