摘要： 面向多声部乐谱，实现基于结构模式的音符识别方法。在音符基元抽取阶段，提出基于游程分析的符干、符头、符梁3类基元抽取算法，具有较好的抗相交和抗粘连干扰能力。在音符结构分析阶段，采用“作用场”描述音符基元关系，将音符分为6类子结构，可缩小基元搜索范围。在此基础上细分音符结构，遵循关键子结构优先定位原则实现音符基元重组，可降低分析复杂度且具有良好的基元冗余排错能力。实验结果表明，该方法能快速准确识别多声部乐谱中的音符，在音符排列密集、结构复杂时适应能力较强。

关键词: 光学乐谱识别, 音符识别, 基元抽取, 结构分析

Abstract: This paper realizes note recognition method based structure mode for multi-voice music score. In the procedure of primitive extraction, three kinds of algorithms based on run-length coding are proposed respectively for extracting stem, note head and beam, they have better ability for resolve troublesome anti-intersection and anti-conglutination. In the procedure of structure analysis, interaction field is introduced to describe the association relationship of primitives, and note is divided to six kinds of substructures to reduce primitive searching range. It subdivides note structure based on aforesaid procedure, note primitive is realized recombination which follows the principle of giving priority to the key substructure, it can reduce analysis complexity and has well ability of primitive redundancy debugging. Experimental results show that this method can recognize note in multi-voice with high accuracy and efficiency, and powerful adaptability, it has good adaptability at the situation of notes intensive arrangement and complicated structure.

Key words: optical music score recognition, note recognition, primitive extraction, structure analysis

中图分类号: 

• TP18