Abstract
This paper explores the vocal and non-vocal music classification problem within popular songs. A newly built labeled database covering 147 popular songs is announced. It is designed for classifying signals from 1sec time windows. Features are selected for this particular task, in order to capture both the temporal correlations and the dependencies among the feature dimensions. We systematically study the performance of a set of classifiers, including linear regression, generalized linear model, Gaussian mixture model, reduced kernel orthonormalized partial least squares and {K-}means
on cross-validated training and test setup. The database is
divided in two different ways: with/without artist overlap
between training and test sets, so as to study the so called
‘artist effect’. The performance and results are analyzed in
depth: from error rates to sample-to-sample error correlation. A voting scheme is proposed to enhance the performance under certain conditions.
Original language | English |
---|---|
Title of host publication | Proceedings of the Ninth International Conference on Music Information Retrieval (ISMIR2008) |
Publisher | ISMIR |
Publication date | 2008 |
Pages | 121-126 |
ISBN (Print) | 978-0-615-24849-3 |
Publication status | Published - 2008 |
Event | 9th International Conference on Music Information Retrieval - Philadelphia, PA, United States Duration: 14 Sept 2008 → 18 Sept 2008 Conference number: 9 http://ismir2008.ismir.net/ |
Conference
Conference | 9th International Conference on Music Information Retrieval |
---|---|
Number | 9 |
Country/Territory | United States |
City | Philadelphia, PA |
Period | 14/09/2008 → 18/09/2008 |
Internet address |
Keywords
- Pop music database
- Music retrieval
- vocal segment classification