Music Note Recognition: Extraction and Structure Analysis of Primitives for Music Scores

摘要

Note recognition is the foundation and core in optical music recognition, which aims at automatically converting scanned scores into a versatile machine-readable format Firstly, this paper proposes run-length coding based algorithms for extracting primitives (stems, noteheads and beams). Secondly, introduces the concept of interaction field to describe the relationship between primitives, and defines six hierarchical categories for the structure of notes. Thirdly, proposes a novel workflow to assemble the primitives into notes, which is guided by the mechanism of giving priority to the key structures. To evaluate the performance of those approaches, we present experimental results on reallife scores and comparisons with commercial systems. The results show our approaches can recognize notes with high-accuracy and powerful adaptability, especially for the scores with high density and complexity.