摘要
简谱是大家非常熟悉和常用的乐谱之一,但是在目前光学乐谱识别领域中对它的研究几乎空白,研究的焦点都集中在五线谱识别上。在深入分析简谱特征的基础上,提出了一整套简谱识别的实现方法。输入的光学简谱经预处理后,首先通过每行的小节线特征提取出简谱部分,然后通过投影法和种子填充算法定位出简谱符号基元的位置,并由此采用不同的识别算法识别出每个简谱符号基元的类型,最后通过组装把各个简谱符号组装成音乐特征符,形成数字化乐谱。实验表明,这套方法对印刷乐谱的识别达到了令人满意的效果,是一项有意义的研究。
Numbered Musical Notation（NMN） is one of the most familiar and common music scores,but it is seldom that anyone studies about its recognition in Optical Music Recognition（OMR） field at present,and everyone focuses on the research of Staff Score Recognition.In the base of deeply study of the features of NMN,this paper proposes an approach of fulfilling the recognition.After pre-dealing the inputted optical MNM,extracts the NMN part according to the bar line feature at first,then locates primitive objects through projections and flood-fill algorithms,and uses diverse algorithms to recognize the type of every primitive object,lastly assembles them to musical features form digital musical score.The result of the experiment indicates the satisfying effect that achieved by using this method to recognize the NMN.
出处
《计算机工程与应用》
CSCD
北大核心
2006年第32期204-206,225,共4页
Computer Engineering and Applications
基金
国家自然科学基金资助项目(60473106
60273060
60333010)
浙江省教育厅科研项目(G20030433)。
关键词
简谱识别
定位
符号基元
类型识别
组装
numbered musical notation recognition
locate
primitive object
type recognize
assemble
