Applications of two novel computational music analysis systems
Date Issued
June 7, 2019
Type
Μεταπτυχιακή Διπλωματική Εργασία
Abstract
This dissertation addresses the question of how two novel representations of chords are implemented in predefined and well-studied datasets. In this study, the aforementioned models are reconstructed and analyzed using statistical methods. Through the use of this process music information retrieval and data visualization methods are applied to those tools for the first time. The aim of this study is to add to the range of implementations of these tools, as well as explore visualization methods that can prove useful for a range of case studies, all closely related to music perception.
The first of the proposed tools, named General Chord Type (abbreviated GCT) is a novel representation of chords based on existing representation formats and music notation systems. It is a computer-understandable model that extracts and analyzes musical information, based on rules provided by the user. It has the potential to become the basis for future implementations of music generation and genre classification systems.
Here, it is used as a representation tool and a basis for genre classification.
The second tool proposed, the Directed Interval Class vector (abbreviated DIC) deals with the intervallic content between chords. In order for it to function it requires the retrieval of a significantly smaller amount of information than any of the existing models and it can be used in exploring musical analysis methods as well as genre classification. Both of these tools are idiom-independent, in contrast to already existing models.
Subjects