音樂擷取對於成功的發展音樂檢索系統是一個很重要的議題，尤其是對於多音的音樂。選擇音樂的主旋律當作索引可以減少在龐大音樂資料中的搜尋時間。在過去，多數的研究者僅專注在單音音樂上。但是，實際上音樂的主旋律可能在音軌間轉換。而也有人把音軌合併起來尋找主旋律，可是卻會多出許多不重要的音符。本論文中，提出兩種改進的方法，分別是以四小節為基礎和利用LBDM 串接的方式從多個音軌中擷取出主旋律。從實驗結果中可以看出所擷取出來的主旋律有很高的正確性。除此之外，本論文還分析各個音樂特徵值的影響力。並進一步利用組合分析找出最有可能影響判斷的特徵值組合。而本論文的目的是希望能夠從多音音樂中自動的擷取出主旋律，以利於後續的處理。

　　Melody extraction is a very important issue of research for successful development of music information retrieval system, particularly for polyphonic music. Selecting the representative melody part of songs as index can reduce the search time for large collections of music. In the past, most researchers focus on monophonic music. However, in reality, melody can move from instrument to instrument. And other people combine all channels to find melody, but it will contain extra and incorrect notes. In this thesis, two methods are proposed: "Four-Measure” and “LBDM Concatenation” to extract the melody from Multi-track music. Experimental results show that extracted melody has high perception accuracy. Comparison of these methods are discussed. Besides, to investigate the influence of each music feature, many experiments with varied combinations of music features are also carried out. Detailed analysis of nine music features are presented. The goal of this work is to extract representative melody from polyphonic music automatically.

[1] A. Ghias, J. Logan, and D. Chamberlin, “Qurey by Humming - Musical Information Retrieval in an Audio Database,” Proceedings of the 3rd ACM International Conference on Multimedia, pp. 231–236, 1995.

[2] N. Kosugi, Y. Nishihara, T. Sakata, M. Yamamuro, and K. Kushima, “A Practical Query-by-Humming System for a Large Music Database,” Proceedings of ACM International Multimedia Conference, pp. 333–342, 2000.

[3] T. Kageyama, K. Mochizuki, and Y. Takashima, “Melody Retrieval with Humming,”Proceedings of International Computer Music Conference, pp. 349–351, September 1993.

[4] S. Field, “Conceptual and Representational Issues in Melodic Comparison,” Melodic Similarity, Concepts, Procedures, and Applications, MIT press, pp. 3–64, 1998.

[5] J. L. Hsu, C. C. Liu, and A. L. Chen, “Discovering Nontrivial Repeating Patterns in Music Data,” IEEE Transactions on Multimedia, vol. 3, no. 3, pp. 311–325, 2001.

[6] C. W. Chang and H. C. Jiau, “Using Quantized Melody Contour to Discover Significant Repeating Figures in Classic Music,” Proceedings of the International Computer Symposium, pp. 1641–1648, December 2002.

[7] C. W. Chang and H. C. Jiau, “Extracting Significant Repeating Figures in Classic Music by Using Quantized Melody Contour,” Proceedings of IEEE International Symposium on Computers and Communications, pp. 1061–1066, June 2003.

[8] C. W. Chang and H. C. Jiau, “A Harmonic-Based Algorithm for Monophonic Music Chord Decision,” Proceedings of IEEE International Conference on Networking, Sensing and Control, pp. 1295–1300, March 2004.

[9] C. W. Chang and H. C. Jiau, “An Improved Music Representation Method by Using Harmonic-Based Chord Decision Algorithm,” Proceedings of IEEE International Conference on Multimedia and Expo, June 2004.

[10] A. Uitdenbogerd and J. Zobel, “Manipulation of Music for Melody Matching,”Proceedings of ACM International Multimedia Conference, pp. 235–240, September 1998.

[11] A. Uitdenbogerd and J. Zobel, “Melodic Matching Techniques for Large Music Databases,” Proceedings of ACM International Multimedia Conference, pp. 57–66, November 1999.

[12] A. Uitdenbogerd and J. Zobel, “Music Ranking Techniques Evalutated,” International Symposium on Music Information Retrieval, October 2000.

[13] M. K. Shan, F. F. Kuo, and M. F. Chen, “Music Style Mining and Classification by Melody,” IEICE Transactions on Information and Systems, vol. 1, pp. 26–29, August 2002.

[14] H. C. Chen and A. L. Chen, “A Music Recommendation System Based on Music Data Grouping and User Interests,” Proceedings of the 10th International Conference on Information and Knowledge Management, pp. 231–238, 2001.

[15] M. Tang, C. L. Yip, and B. Kao, “Selection of Melody Lines for Music Databases,”The 24th Annual International Computer Software and Applications Conference, pp. 243–248, October 2000.

[16] S. G. Blackburn, Content Based Retrieval and Navigation of Music Using Melodic Pitch Contours. A thesis submitted for the degree of Doctor of Philosophy, 2000.

[17] S. Blackburn and D. D. Roure, “Musical Part Classification in Content Based Sstems, "Proceedings of the 6th International Workshop and 2nd International Workshop on Open Hypertext Systems and Structural Computing, pp. 66–76, 2000.

[18] E. Cambouropoulos, “A Formal Theory for the Discovery of Local Boundaries in a Melodic Surface,” Proceedings of the III Journees d’ Informatique Musicale, 1996.

[19] E. Cambouropoulos, “The Local Boundary Detection Model (LBDM) and its Application in the Study of Expressive Timing,” Proceedings of the International Computer Music Conference, pp. 17–22, September 2001.

[20] C. H. Hsu, Music Retrieval Using Melody Line Extracted from Real-World Music. Master thesis, 2002.

[21] Z. Q. Wu, A Guide to Musical Forms. Mercury Publishing House, 1996.

[22] L. Stein, Structure and Style–The Study and Analysis of Musical Forms. Evanston, Ill., Summy-Birchard Company, 1962.

[23] E. G. Gutierrez, Melodic Description of Audio Signals for Music Content Processing. A thesis submitted for the degree of Doctor of Philosophy, 2002.