以音樂內容之分析做為音樂資料之分類

Content-base Analysis for Music Classification

林奕昌(Yi-chang Lin)

指導教授 ： 羅有隆

朝陽科技大學/資訊學院/資訊管理系碩士在職專班/碩士(2010年)

摘要

已被發展出的音樂分類技術可概分為兩個方向，其一為以學習機器為訓練基礎的分類方式，另一則為以音樂內容為基礎的分析分類方式，各有其優缺點。而其中，現有以音樂內容為基礎的分類方式，多以音樂之單特徵進行，在少數類型的音樂辨識可達70%左右的分類精確度，而多數類型的音樂，可正確被分類的精確度則低了許多。本研究報告在於分析與研究音樂資料內容，並利用音樂資料的多特徵性質，找出是否有每一類型音樂所特有的內容特徵，並以這些特徵設計出具更高精確度的音樂資料分類公式。在我們的實驗中，我們發現利用音樂內容之多重特徵為分類基礎，可讓鄉村音樂、古典音樂、流行音樂與爵士音樂，分別達到85%、82%、80%與73%的分類精確度。

關鍵字

內容擷取 ； 音樂分類 ； 數位音樂 ； 音樂資料庫 ； 多媒體資料庫

並列摘要

The music classification techniques can be discriminated into two categories — based by music feature classification and training by learning machine classification. Both have their advantages and disadvantages. For music feature classifications, most of the approaches are based on single music feature, such as melody or chord, and the accuracy is about 70% in few genres of music. However, the accuracy for classification of most music genres is lower. In this research, we study the music contents and use the multi-features of music to design equation for more accuracy music classification. Our performance study shown that more than 85%, 82%, 80%, and 73% of folk, classic, pop, and jazz music can be classified correctly, respectively, by using multi-feature of music content for classification.

並列關鍵字

content-based retrieval ； music classification ； digital music ； music database ； multimedia database

參考文獻

[4] S. Brecheisen, H.-P. Kriegel, P. Kunath, and A. Pryakhin, “Hierarchical Genre Classification for Large Music Collections” IEEE Seventh international conference on multimedia and Expo, pp. 1385-1388, 2006.

[6] Y. Chen, J. Bi, and J. Z. Wang, “MILES: Multiple-instance learning via embedded instance selection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 28(12):1931-1947, 2006.

[8] G. Davenport, T. A. Smith, and N. Pincever, “Cinematic Primitives for Multimedia,” IEEE Computer Graphics & Applications, Vol. 11, No. 4, pp. 67-74, 1991.

[10] R. B. Dannenberg, B. Thom, and D. Watson, “A Machine Learning Approach to Musical Style Recognition,” Proceedings of International Computer Music Conference, 1997.

[11] P. Dhanalakshmi, S. Palanivel, and V. Ramalingam, “Classification of Audio Signals Using SVM and RBFNN” An International Journal Expert Systems with Application, Vol. 36, No. 3, pp. 6069-6075, 2009.