MIDI Melody track recognition

Description

A melody can be defined "melody" as a 'cantabile' sequence of notes, usually what a listener can remember of a song after hearing it, but this definition is not computable. MIDI files are organized by tracks containing one different voice each. A reliable system to identify the track containing the melody of a MIDI file is very relevant in music information retrieval for indexing and comparing music pieces.

4 files are provided, three of them for particular music genres and a forth one merging the other three, in order to study the possible specificities of melody in different music genres:

• CL200: classical music. 703 instances (16 unlabeled)
• JZ200: jazz. 769 instances (11 unlabeled)
• KR200: karaoke (mainly pop music). 1668 instances (338 unlabeled)
• MEL200.arff: the merge of the 3 files.

3140 samples are provided, 365 unlabeled.

Quality

Source

Pattern Recognition and Artificial Intelligence Group - University of Alicante (PRAIg-UA)

Ground Truth Annotation

Each track is described as a 34 dimensional vector stored together with class labels and four other tags (the four first attributes are metadata, see below, and the last one is the boolean label 'IsMelody').

Ground truth has been Manually annotated.

• Name of the MIDI file the track was extrated from.
• Path name, corresponding to the genre of the particular MIDI file.
• Song number (starting form zero for each genre) the track belongs to.
• Instance number, starting from 1 for each MIDI file.

Features

Features are both statistical descriptors of the track content (e.g. average note pitch) and how the track is related to the others in the file (e.g. track relative duration). ARFF Weka format is used.

Licensing / Copyright

Public use.

Citation

A Pattern Recognition Approach for Melody Track Selection in MIDI Files. Rizo D., Ponce de León P. J., Pérez-Sancho C., Pertusa A., Iñesta J. M. (2006). In: Proc. of the 7th Int. Symp. on Music Information Retrieval ISMIR 2006, pp. 61-66, Victoria, Canada.

Towards a human-friendly melody characterization by automatically induced rules. Pedro J. Ponce de León, José M. Iñesta, David Rizo, Proceedings of the 8th Int. Conf. on Music Information Retrieval, ISMIR 2007, Simon Dixon, David Bainbridge, Rainer Typke, Austrian Computer Society, 437--440, 2007

External Links

http://grfia.dlsi.ua.es/cm/projects/prosemus/database.php