Self-organizing maps
Recognition of Western style musical genres using machine learning techniques
Expert Systems with Applications: An International Journal
| Hi-index | 0.00 |
In the present paper, we describe a method for extracting certain basic note sequences that should function as stable starting points for variations in Hungarian folk music. The database of our study contains melody section contours constructed from digitized notes of 2323 Hungarian folk songs. Using these contour sequences, a self-organizing map (SOM) was trained to determine typical note sequences appearing in many individual melody sections. To identify these "roots" in different parts of melody sections, we worked out a version of dynamic time warping with adaptive weights for the purpose of the distance function of the SOM. The resulting melody roots proved to be of clear and self-contained musical meaning, and they correspond to the results of classical ethnomusicology. The system of the roots can be interpreted as a condensed form of the main information of the studied musical culture.