Please use this identifier to cite or link to this item: http://hdl.handle.net/11455/8759
標題: 以最大調性輪廓相關法為基礎的音樂調性演算法
The music key finding algorithm based on the maximum key-profile correlation
作者: 李思源
Lee, Sz-Yuan
關鍵字: music key;音樂調性
出版社: 電機工程學系所
引用: Reference [1]J. Pedro Ponce de Le′on and Jos′e M. I˜nesta, “Pattern Recognition Approach for Music Style Identification Using Shallow Statistical Descriptors”, IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART C: APPLICATIONS AND REVIEWS, VOL. 37, NO. 2, MARCH 2007. [2]張智星、陳若涵,”以音樂內容為基礎的情緒分析與辨識”,2006。 [3]Carol L. Krumhansl, “Cognitive Foundations of Musical Pitch”, New York: Oxford University Press, 1990. [4]Carol L. Krumhansl and E. J. Kessler, “Tracing the dynamic changes in perceived tonal organisation in a spatial representation of musical keys”, Psychological Review, 89:334–368, 1982. [5]David Temperley, “What’s Key for Key? The Krumhansl-Schmuckler Key-Finding Algorithm Reconsidered”, Music Perception, 17(1):65–100, 1999. [6]Pauws, S, “Musical key extraction from audio”, Proceedings of International Conference on Music Information Retrieval, 2004. [7]Chuan, Ching-Hua and Chew, Elaine, “Polyphonic Audio Key-Finding Using the Spiral Array CEG Algorithm”, Proceedings of International Conference on Multimedia and Expo, Amsterdam, Netherlands,July 6-8, 2005. [8]Zhu, Y., Kankanhalli, M. S. and Gao, S, "Music Key Detection for Musical Audio", Proceedings of the 11th International Multimedia Modelling Conference, Melbourne, Australia, 2005. [9]Katy Noland and Mark Sandler, ”Key Estimation Using a Hidden Markov Model”, In Proceedings of the 7th International Conference on Music Information Retrieval (ISMIR), pages 121–126, Victoria, Canada, October 2006. [10]Geoffroy Peeters,“Musical key estimation of audio signal based on hidden markov modelling of chroma vectors”, In Proceedings of the 9th International Conference on Digital Audio Effects, Montreal, 2006. [11]ZENZ V., RAUBER A.,“Automatic chord detection incorporating beat and key detection”, IEEE International Conference on Signal Processing and Communications (ICSPC 2007). [12]S. T. Madsen and G. Widmer,“Key-Finding with Interval Profiles”, In Proceedings of the International Computer Music Conference (ICMC), Copenhagen, Denmark, August 2007. [13]Ching-Hua Chuan and Elaine Chew,“Audio Key Finding: Considerations in System Design and Case Studies on Chopin’s 24 Preludes”, EURASIP Journal on Advances in Signal Processing, 2007. [14]M. Robine, T. Rocher, and P. Hanna, ”Improvements of Key-Finding Methods”, Proceedings of the International Computer Music Conference (ICMC), Belfast, Irlande du Nord, 2008. [15]http://ppt.cc/R6Bc [16]http://zh.wikipedia.org/zh-tw/%E5%A4%A7%E8%AA%BF
摘要: 
摘要
隨著網際網路快速的發展以及多媒體壓縮技術的日益精進,現今已有非常大量的多媒體資料在網路上無國界的傳遞著,由於數位化多媒體資料的應用日益普遍,多媒體資料的分類與查詢技術之需求也漸漸受到大家的重視,因此,以內容為基礎的多媒體分析(content-base multimedia analysis)已經成為學術界研究的焦點。其中音樂調性是一種高階音樂內涵,常在許多音樂分類應用上扮演著重要的角色,如音樂類型[1]或音樂的情緒分析[2]等等。除此之外,音樂調性提供了一個非常方便的音樂資料庫搜尋線索,對資料庫使用者與相關領域研究人員而言是一個非常有用的檢索條件。
本文以統計的觀點,重新檢視Krumhansl[4]的「最大調性輪廓相關法」(maximum key-profile correlation),利用調性參數平均法(key parameter average)提出新的參數表,再結合樂理中音階結構的特性與最後一個音加權,來減少計算量與提高辨識率。
第一章介紹音樂調性分類的研究。調性分類的研究以Krumhansl的「最大調性輪廓相關法」為其始祖,在此之後其他研究者提出了以螺旋陣列CEG(Spiral Array Center of Effect Generator)、隱藏馬可夫模型(Hidden Markov Model)、和弦和音程等等不同判別調性的特徵與方法。
為了讀者閱讀上的方便,本文在第二章先簡介在往後章節皆有用到的樂理基本概念,接著詳述Krumhansl[4]獲得「最大調性輪廓相關法」參數實驗的實驗器材、方法、以及流程。在第二章的最後,我們補充Temperley[5]的看法:他以樂理的觀點,修正Krumhansl[4]參數不合理處,並提出改善方法。
由於Krumhansl[4]與Temperley[5]提出的參數皆以人為的方法決定「最大調性輪廓相關法」的參數,主觀性較重。第三章將排除人為的因素,以統計的觀點,利用調性參數平均法獲得一組「最大調性輪廓相關法」的參數。
本文第四章以樂理中音階結構的特性,應用於「最大調性輪廓相關法」中,實驗結果得知可有效的減少計算量。接著以樂理及經驗法則為基礎,利用最小均方誤差法(minimum mean square error, MMSE)的概念,對歌曲最後一個音的調性做加權,提升「最大調性輪廓相關法」的性能,並增加演算法在音樂調性分類的實用性。
最後是本研究的結論與展望。期望將來以本研究所提出的調性參數平均法,利用較大的訓練資料庫以獲得一組更一般性的參數,並且結合樂風分類,讓「最大調性輪廓相關法」在調性的判斷上有更準確的分類性能。

Abstract
In recent years, by the fast development of the Internet and the advancement in the compression techniques, there are large amounts of multimedia data have been rapidly spread on the Internet, the applications of digital multimedia data have been increasing and content-base multimedia analysis has become the focus of recent research. The music key in high level music content is very important. It plays a crucial role in many music classification applications, such as the music style classification [1] and the music mood analysis [2]. Besides, the music key provides a very convenient clue for digital multimedia data searching, and it is also a very useful searching condition for the user and related field researcher.
This study reconsiders the Krumhansl's maximum key-profile correlation algorithm via the viewpoint of statistics, then using key parameter average to train a new weighting table. Finally, this study combines music theory and weighting the last note to reduce the amount of calculation and classification accuracy.
Chapter one introduces the study of music key classification. The pioneer of music key classification is Krumhansl's maximum key-profile correlation algorithm. Afterwards other researchers using Spiral Array Center of Effect, Hidden Markov Model, the chord transient, and the interval of the music to classify music key.
For the reading convenience, chapter two introduces some concepts about music theory first, and then states the experiment in [4] about the experimental materials, methods and the procedures. At the end of this chapter, Temperley [5] noted that some unreasonable part of the experiment result and the improvement.
Due to the weighting table from Krumhansl [4] and Temperley [5] are artificial, there may exist some arbitrary decisions. In order to avoid such case, this study uses key parameter average to train a new weighting table in chapter three.
Chapter four uses the scale structure of music theory to reduce the amount of calculation, and weighting the last note by using minimum mean square error. It applies to the maximum key-profile correlation algorithm to enhance the performance and the practicality.
To conclude, this study shows a method to establish the weighting table of maximum key-profile correlation algorithm. It also provides some ways to improve the performance which includes reducing the amount of calculation and raising classification accuracy. The idea can be used to establish a normal weighting table by a large amount database for future research, and combining music style classification to enhance the music key classification performance.
URI: http://hdl.handle.net/11455/8759
其他識別: U0005-0208201019070000
Appears in Collections:電機工程學系所

Show full item record
 

Google ScholarTM

Check


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.