Automatic Makam recognition using chroma features

Abstract

This work focuses on the automatic makam recognition task for Turkish Makam Music using chroma features. Chroma features are widely used for music identification and tonal recognition tasks such as key estimation or chord recognition. Most of prior work on makam recognition largely rely on use of pitch distributions. Due to the imperfection of automatic pitch extraction for non-monophonic audio, use of chroma features is an alternative that has been showed to be effective in a previous study and we follow the same approach. Our work does not propose a new architecture but rather considers parameter optimization of chroma based recognition for makams. In our tests we use an open-content dataset and perform comparisons with previous studies. As a result of parameter optimization a better performance is achieved. All resources are shared for ensuring reproducibility of the presented results.