Abstract:

With this research, it is aimed to determine what kind of results the segmentation algorithms give on Turkish music works. In the research, the segmentation results of the Segment Prob, Segment Gestalt and Local Boundary Detection Model (LBDM) algorithms in the MIDI Toolbox software developed on the Matlab platform in Turkish music works were evaluated. These results were compared with expert segmentation for a previous study. As a result of the trials conducted within the scope of the research, the closest result to the expert segmentation was the LBDM algorithm with a threshold value of 0.2. Algorithm and expert segmentation were composed in different maqams and forms and tested on 10 randomly selected works. It has been seen that the algorithm gives results close to experts on these works. Although there is information in some sources about the success of the LBDM algorithm, no comparison of expert segmentation with visual examples has been found on the notation of Turkish music works and with this threshold value. This situation shows the importance of the research.

Keywords: