Score Informed Tonic Identification for Makam Music of Turkey - ISMIR 2013

Şentürk, S., Gulati, S., and Serra, X. (2013). Score informed tonic identification for makam music of Turkey. In Proceedings of 14th International Society for Music Information Retrieval Conference, pages 175–180, Curitiba, Brazil.

 

Tonic is a fundamental concept in many music traditions and its automatic identification should be relevant for establishing the reference pitch when we analyse the melodic content of the music.

 

In this paper, we present two methodologies for the identification of the tonic in audio recordings of makam music of Turkey, both taking advantage of some score information. First, we compute a prominent pitch and a audio kernel-density pitch class distribution (KPCD) from the audio recording. The peaks in the KPCD are selected as tonic candidates. The first method (Distribution Matching) computes a score KPCD from the monophonic melody extracted from the score. Then, the audio KPCD is circular- shifted with respect to each tonic candidate and compared with the score KPCD. The best matching shift indicates the estimated tonic. The second method (Repetitive Section Linking) extracts the monophonic melody of the most repetitive section of the score. Normalising the audio prominent pitch with respect to each tonic candidate, the method attempts to link the repetitive structural element given in the score with the respective time-intervals in the audio recording. The result producing the most confident links marks the estimated tonic.

 

We have tested the methods on a dataset of makam music of Turkey with 257 recordings, achieving a very high accuracy (94.9%) with the first method, and almost perfect identification (99.6%) with the second method. For comparison, we also modify and test the approach in (Gedik & Bozkurt, 2010) on a subset of the dataset (152 recordings) and obtained 69.74% accuracy. We conclude that score informed tonic identification can be a useful first step in the computational analysis (e.g. expressive analysis, intonation analysis, audio-score alignment) of music collections involving melody-dominant content.

 

[pdf] [bib] [results]

 

References

Ali Cenk Gedik and Barış Bozkurt. Pitch-frequency histogram-based music information retrieval for Turkish music. Signal Processing, 90(4):1049–1063, 2010.