Aristotle University Of Thessaloniki (AUTH)
Ανάκτηση μουσικών πληροφοριών για το ραδιόφωνο
Abstract
dc:descriptionRecent developments in computer science and signal processing methods have increased enormously the number of music information distributed. As music databases grow in size and number new techniques to organize, manage and search music information from large audio collections become necessary. The present thesis investigates the automatic multi-label classification of music into emotions. The primary aim of the thesis is the analysis and understanding of emotions and their relation to music. For that purpose, several emotional models are studied and compared. Experiments are conducted on a set of 593 songs with 6 clusters of emotions based on the Tellegen-Watson-Clark model. Next, several feature extraction methods are described for the representation of music and various features are used to model musical signals. Furthermore, a new multi-label feature ranking method was proposed, which seems to perform better than existing methods. Feature ranking may assist researchers working on feature extraction by providing feedback on the predictive performance of current and newly designed individual features. It also improves the performance of multi-label classification. An evaluation of five multi-label classification algorithms was performed on the described music collection. Among these algorithms, RAKEL was the most effective and is proposed for emotion categorization. The overall predictive performance was high and encourages further investigation of multi-label classification methods.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2009
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Τροχίδης, Κωνσταντίνος
- Trochidis, Konstantinos
Subjects
dc:subject × 6Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/19740