基于哼唱的音乐检索研究
文献类型：学位论文
| 作者 | 李明 |
| 学位类别 | 博士 |
| 答辩日期 | 2005 |
| 授予单位 | 中国科学院声学研究所 |
| 授予地点 | 中国科学院声学研究所 |
| 关键词 | 哼唱 音乐检索 旋律因子 旋律搜索 |
| 中文摘要 | 随着数字音乐的迅猛发展，如何快速有效地进行音乐检索成为日益关注的研究领域之一。Google等基于文本的传统检索方式只能对有标注信息的音乐文件进行检索，基于内容的检索则不依靠标注信息，而是根据音乐中的旋律、节奏、音色等信息进行检索。基于哼唱的音乐检索是音乐检索的方式之一。本文在基于哼唱的音乐检索方面，主要做了以下工作：1、针对谐波和提取基频方法需要大量频谱计算的问题，提出了谐波和快速算法，并在理论上分析了该算法可以达到的基频分辨率。2、根据基频和其谐波在频谱上均匀分布的特点，提出了频谱归一化谐波和方法。3、提出了一种切分音符的方法，即首先依据自适应能量闽值切分发音段，然后根据谐波和突出度切分音符，并在基频的跃变处进一步切分，最后归并音符并检查每个音符的有效性。4、在N一gram索引法基础上，根据旋律时序性特点提出旋律因子的投票机制。5、提出了轮廓旋律因子方法，该方法对于哼唱中多音少音的情况有较强的容错能力。6、根据人在旋律感知上模糊性的特点，提出了旋律因子类的概念，给出了旋律因子类的自动聚类分类算法。7、提出了一个完整的旋律搜索策略，即从旋律定位，到由粗至精旋律匹配搜索的过程。8、提出了一种在MIDI文件中寻找主旋律音轨的方法。根据以上的方法，本文构建了一个基于哼唱的音乐检索系统，实验结果证明了以上方法的有效性。 |
| 英文摘要 | With the rapid adoption of digital music, quick and effective music information retrieval has been paid more and more attention by researchers in information retrieval field. Google and other text based search engines are only effective for music files with metadata. Content-based music information retrieval is to access the desired music files by melody, rhythm, timbre and other difficult-to-extract layers of significance. Query by humming is one kind of content-based music information retrieval. This thesis mainly focuses on the following research, issues on music information retrieval by humming. 1. Since sub-harmonic summation method for pitch tracking needs a lot of computation for spectrum interpolation, a rapid sub-harmonic summation method is proposed, and the theoretic pitch resolution analyzed. 2. A normalized spectrum sub-harmonic summation method is proposed, which exploits the even distribution property of pitch and its harmonics on spectrum. 3. A note segmentation method is proposed. First it locates the voice segment boundaries by the adapted energy threshold. Then notes are segmented by bumping degree method. Locations where pitch varies rapidly are labeled as note boundaries further. Last all notes are checked for validity and some notes are deleted or merged by some rules. 4. It proposed melody element voting method which is based on the melody property of sequencing. 5. A melody contour element approach is proposed which is robust for note fragmentation and consolidation. 6. Activated by fuzzy property of music perception, a melody element cluster approach with automatic classification is proposed. 7. A complete melody search strategy is proposed, which orients melody first and then matches melody from rough mode to fine mode. 8. It proposes an automatic method to find the main melody tracks in MIDI files. Based on the all above proposals, a demo system of music retrieval by- humming is implemented. The experiments demonstrate the effectiveness of the proposed algorithms. |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 78 |
| 源URL | [http://159.226.59.140/handle/311008/928]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 李明. 基于哼唱的音乐检索研究[D]. 中国科学院声学研究所. 中国科学院声学研究所. 2005. |
入库方式： OAI收割
来源：声学研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。