A Learning-Based Model for Musical Data Representation Using Histograms

Authors:
Mehdi Naccache;Amel Borgi;Khaled Ghédira
Affiliations:
ENSI, Campus Universitaire Manouba, Manouba, Tunisia 2010;INSAT, Centre Urbain Nord de Tunis, Tunisie 1080;ENSI, Campus Universitaire Manouba, Manouba, Tunisia 2010
Venue:
Computer Music Modeling and Retrieval. Genesis of Meaning in Sound and Music
Year:
2009

Citing 4
Cited 0

C4.5: programs for machine learning

C4.5: programs for machine learning
AI and music from composition to expressive performance

AI Magazine
Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)

Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)
Using Machine-Learning Methods for Musical Style Modeling

Computer

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper we are interested in musical data classification. For musical features representation, we propose to adopt a histogram structure in order to preserve a maximum amount of information. The melodic dimension of the data is described in terms of pitch values, pitch intervals, melodic direction and durations of notes as well as silences. Our purpose is to have a data representation well suited to a generic framework for classifying melodies by means of known supervised Machine Learning (ML) algorithms. Since such algorithms are not expected to handle histogram-based feature values, we propose to transform the representation space in the pattern recognition process. This transformation is realized by partitioning the domain of each attribute using a clustering technique. The model is evaluated experimentally by implementing three kinds of classifiers (musical genre, composition style and emotional content).