乐谱识别中音符结构分析方法

Structure analysis approach of music notes for optical music recognition

将物理学中的作用场概念引入到音符基元的关系表达,采用作用场的势函数描述音符基元间的位置关系,在表达方式上实现了知识性、鲁棒性和精确性三者的统一。对音符结构进行细分,定义了6个音符子结构,并对各个子结构的关键程度进行分析和排序,设计了关键结构优先定位的音符结构分析流程和算法。该方法模拟了人工识谱时突出重点、从整体到细节的思维习惯。实验结果表明,该方法对识别各种复杂音符有良好的正确性、适应性和容错性。

A structure analysis approach of music notes is proposed for optical music recognition. By introducing the physics concept ofaction field to describe the association relationship ofnote primitives, it has good qualities with knowledgablity, robustness and precision. With dividing music notes to elementary parts, six substructures are defined. On this condition, a workfiow of music notes structure analysis with priority identification for key structures is set up. The approach imitates the way human recognizes music score, which is prone to focus on point features and observe objects from whole to detail. The experiments show that the approach embodies high percentage ofaccuracy, powerful adaptability and fault-tolerance capability, especially for the situation ofcomplicated music notes recognition.