Abstract: Based on the character of humming melody extraction and qualitative mapping, combining with the music theory, a qualitative mapping model for extracting humming melody and its algorithm are given in this paper. The key points of the algorithm are the method to choose the criterion of the note segmentation dynamically and the dynamic measure to determine the pitch of a note. Experimental results show that not only a better result of choosing flexibly and reasonably the criterion of the note segmentation can be determined, but also the note pitch can be extracted accurately. The contour of humming melody can be achieved too.

Key words: qualitative mapping, attribute computing grid, gravity subdivision model, note segmentation, pitch extraction, melody code

摘要： 根据哼唱旋律提取的特点和定性映射的相关知识，结合基本乐理，提出哼唱片段旋律提取的定性映射模型和算法，包括哼唱旋律音符切分阈值的动态定性判定，以及音符音高确定的动态赋权的定性映射。实验结果表明，该算法不仅能灵活合理地确定音符切分的定性判定基准，同时能够较准确地提取音符音高，并得到哼唱片段的旋律轮廓线。

关键词: 定性映射, 属性计算网格, 重心剖分模型, 音符切分, 基音提取, 旋律编码

CLC Number: 

• TP391