Dynamic segmentation approach for efficient Arabian music transcription

In Arabian music transcription where the tone-to-tone distance is half of that of the Western music, the segmentation phase of the transcription process can therefore be more challenging. This paper describes an efficient dynamic segmentation approach which considers the sixteenth note as fundamental note duration. Unlike static segmentation, the proposed dynamic segmentation approach divides the entire musical piece into non-equal durations depending upon tempo of the musical piece and the repetition rate of the musical notes. Performance of the proposed segmentation algorithm is assessed experimentally by applying it to numerous human and instrumental musical pieces that are performed with Arabian quarter tones. When compared to static segmentation with various overlap periods, the proposed dynamic segmentation has demonstrated a significant reduction in computation load and improved tone-to-tone distance estimation.