Abstract: Toward recognizing complex multi-voice scores，new methods are proposed respectively for extracting three shapes of music note primitive：first，stem extraction is accomplished by searching stem candidates roughly from vertical run length units plus checking the candidates strictly in terms of horizontal run length properties，and this method can overcome the weaknesses of poor adaptability to the touched and overlapped notes；second，a “segmentation with feature testing” method for extracting note heads is designed，which employs the prior knowledge of notes to separate the note heads that touch；thirdly，an extraction method for beams similar to treating the note heads is proposed，and this method can avoid the troublesome touching issue in the former line detection method for extracting beams.The experiments show that the methods can extract note primitives in complex multi-voice scores successfully and effectively.

Key words: optical music recognition, multi-voice scores, music notes, primitives extraction

摘要： 面向复杂版面的多声部乐谱，提出了符干、符头和符梁三类音符基元的抽取方法。提出基于垂直游程编码的粗提取、基于水平游程编码的精检测的符干抽取方法，能够有效克服基元间密集相交、粘连的干扰；在音符先验知识引导下，提出先分割、后特征检测的符头抽取方法，解决了粘连符头的切分难题；提出一种基于块状体分割和特征检测的符梁抽取方法，避开了传统的直线抽取方法所无法处理的符梁粘连难题。实验结果表明，该方法能够有效抽取复杂环境下多声部乐谱的音符基元。

关键词: 光学乐谱识别, 多声部乐谱, 音符, 基元抽取