Automatic Content-Based Classification of MP3 Music Objects Using Eigensong
| Autor: | 蔡易行 |
|---|---|
| Rok vydání: | 2002 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 89 As the growth of the Internet, multimedia applications on the World Wide Web are becoming popular. MPEG standards have provided compression techniques for a wide range of multimedia applications. In this paper we propose an approach for automatic content-based classification of MP3 music objects. We use 256-dimension features to represent the content of an MP3 music object. The axes feature space are called eigensongs. Our approach treats MP3 classification as a two-dimensional classification problem. Unknown MP3 objects are projected into the 256-dimensional feature space. Then, its corresponding features are represented using the linear combination of the eigensongs. Using this technique we can classify MP3 music objects into three music classes, i.e., classical, female, and male. As the experiments shown, our MP3 classification system can classify unknown MP3 objects with an accuracy rate higher than 72﹪. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|