一种新的哼唱音符音高划分方法

A New Method of Note Segmentation for Humming Music

哼唱音符音高的准确划分，对哼唱音乐检索系统识别率的提高起着很大的作用。目前，大部分的哼唱音乐检索系统都是采用能量划分的方法，在很大程度上并不能对哼唱波形文件顺利完成单音切割，因此，论文提出的一种新的音符音高划分方法，在基于一般能量划分的基础上，采用基于倍音列的音高识别模型对划分结果进行二次划分、规整，最终实现哼唱音符音高的划分。实验表明，该划分方法能够有效地实现哼唱音符音高的准确划分。

The accuracy of note segmentation of humming music is very important for the recognition rate of humming music retrieval system. Currendy, most humming music retrieval systems, adopting the method of energy segmentation, can not successfully complete the single-tone segmentation of humming music. Therefore, a new method＇s proposed in this paper, which based on the traditional energy segmentation method and overtone series model. The experiments show that the new method proposed in this paper can effectively completes the single-tone segmentation of humming music.