|
|
Computational Methods for Comparative Music Performance Analysis
Ištvánek, Matěj ; Róka,, Rastislav (oponent) ; Jirásek, Ondřej (oponent) ; Smékal, Zdeněk (vedoucí práce)
The availability of digital music content and various interpretations of musical pieces is increasing rapidly. Simultaneously, even though the computational methods of Music Information Retrieval (MIR) are evolving at a quick pace, they are not always reflected in related fields such as Music Performance Analysis (MPA). The main topic of this dissertation is the utilization of computational methods and digital music processing for the goals of MPA. It aims to combine MIR principles to analyze and compare differences in musical performances and their parameters. The thesis examines the limitations of conventional and machine learning-based onset and beat detectors to correctly estimate the ground-truth data under the effect of input audio degradations, sampling rate reductions, or in complex musical structures. Furthermore, this work shows the possibilities of music synchronization, parameter extraction, and feature selection application on novel string quartet data to provide a semi-automated strategy for binary classification of performers’ origin. Finally, it demonstrates a software tool for comparative music analysis by combining a comfortable user environment for playback, navigation, and visualization of music performance data with the computation methods of MIR.
|
host ::
RSS.