由於音樂內涵的多元和豐富性，音樂可分成很多種類型，因此產生許多音樂分類的系統。根據樂理上來說，可從旋律、和絃、節奏、和聲等。而每種音樂都有其特徵。比方說，佛朗明哥之於強烈的節奏，古典音樂之於繁複的和聲。也就是說特徵就可以代表每種特定的音樂方便人們去區分。所以說音樂分類中不可獲缺的重要因素就是選取特徵。

不同於一般的音樂分類，我們是針對特定的樂種—搖滾樂，希望在搖滾樂中分析出更細緻的樂派，如Folk, Heavy metal, Funk, Punk, Raggae等廣為人知的樂風。在這篇論文中，我們提供了一套特徵選取的方法，在樂理以及聽覺上針對節奏與和弦做詳盡的分析。為何選擇這節奏與和弦呢？第一，根據搖滾樂的起源來看，swing是搖滾樂重要元素之一，而肢體語言就跟節奏息息相關；第二，在作曲理論中，和弦的模進(chord progression)會影響曲風的走向，而在也提到曲風與和弦的關係。所以和弦與節奏在辨別搖滾樂曲風扮演著重要角色。接著，由向量支援機Support Vector Machine建立的分類器，進行分類結果的準確率評估。我們設計了三組特徵的實驗，和弦、節奏以及結合和弦與節奏，由實驗結果顯示，節奏的準確率比和弦高，而結合節奏與和弦的準確率又比節奏高。

Due to the diverse properties of music, music can be categorized into many types, generating huge amounts of music style classification systems to be proposed. According to musicology, melody, rhythm, harmony and other features can be employed in discriminating styles of the music. In other words, every music genre has its own specific features. In this thesis, we present a method of feature extraction based on the characteristic of the rock music style and musicology. Two features, i.e., chord and rhythm (bass and percussion), are extracted from the music objects for music classification. After feature extraction, the feature vectors of the training music objects are used as the input for the SVM classifier. The generated classification model is then adopted to discriminate one music style from another. We perform three series of experiments to show the accuracy of classification results.

[1] Charles T.Brown, The art of rock and roll, Prentice Hall, 1993.

[2] Robert Palmer, Rock & Roll: an unruly history, WGBH Educational Foundation and Robert Palmer, 1995.

[3] KG Johansson, “What Chord Was That? A Study of Strategies among Ear Players in Rock Music,” Number 23, 2004.

[4] Frith Simon, Straw Will, Street John, The Cambridge Companion to Pop and Rock, 2001.

[5] Tao Li, Mitsunori Ogihara, Qi Li, “A Comparative Study on Content-Based Music Genre Classification,” ACM SIGIR, 2003.

[6] Pedro J. Ponce de Leon, Carlos Perez-Sancho and Jose M. I˜nesta , “A Shallow Description Framework for Musical Style Recognition,” Lecture Notes in Computer Science, 2004

[7] Tao Li, Mitsunori Ogihara, “Music Artist Style Identification by semisupervised Learning from both Lyrics and Content,” ACM international conference on Multimedia, 2004.

[8] Cory McKay. Ichiro Fujinaga, “Automatic Genre Classification Using Large High-Level Musical Feature Sets,” ISMIR, 2004.

[9] MK Shan, FF Kuo, MF Chen, ”Music style mining and classification by melody,” Multimedia and Expo, 2002. ICME . Proceedings, 2002. IEEE International Conference, 2002.

[10] J.Miguel D□az-B□□ez…, ”El comp□s Flamenco: A Phylogenetic Analysis,” Proceedings of BRIDGES: Mathematical Connections in Art, 2004.

[11] D Van Steelant, K Tanghe, S Degroeve, B De Baets, “Classification of Percussive Sounds Using Support Vector Machines,” Proceedings of the annual machine learning conference, 2004.

[12] Xi Shao, Changsheng Xu, Mohan S Kankanhalli, “Unsupervised Classification of Music Genre Using Hidden Markov Model,” International Conference of Multimedia Expo ICME, 2004.

[13] Michael I. Mandel, Graham E. Poliner, Daniel P. W. Ellis, “Support vector machine active learning for music retrieval,” Multimedia Systems, Volume 12, Number 1, 3-13. 2006

[14] Juo-Han Chen, Content Base Music Emotion Analysis and Recognition 2006

[15] Krumhansl, C. L., & Kessler, E. J., Tracing the Dynamic Dhanges in Perceived Tonal Organization in a Spatial Representation of Musical Keys. Psychological Review, 89. 1982.
[16] http://www.musicrobot.com/

[17] http://www.csie.ntu.edu.tw/~cjlin/libsvm/