進階搜尋


 
系統識別號 U0026-0812200914281978
論文名稱(中文) 適用於音樂治療中之自動音樂情緒偵測的音樂特徵研究
論文名稱(英文) Discovering Musical Features for Automatic Emotion Classification in Music Therapy
校院名稱 成功大學
系所名稱(中) 電腦與通信工程研究所
系所名稱(英) Institute of Computer & Communication
學年度 96
學期 2
出版年 97
研究生(中文) 吳崟愷
研究生(英文) Yin-kai Wu
學號 q3694414
學位類別 碩士
語文別 英文
論文頁數 79頁
口試委員 口試委員-斯國峰
口試委員-詹寶珠
指導教授-焦惠津
中文關鍵字 情緒  音樂  情緒辨識  音樂特徵  信號特徵 
英文關鍵字 signal feature  emotion  music  emotion detection  music feature 
學科別分類
中文摘要 音樂情緒辨識在BMGIM音樂治療領域是重要的。現今音樂治療師需要以人工來辨別音樂中所表達的情緒。本實驗依照BMGIM之書中列出的古典音樂清單以及相對應的情緒標籤,設計了自動音樂情緒分類系統。為因應不同背景之病患的音樂偏好,此情緒自動分類系統將來可應用到非古典樂類型的音樂上。現今雖然有一些自動音樂情緒辨別系統之研究以大量的信號特徵與機器學習演算法來建立分類法則,然而,直接以信號特徵建立的分類法則很難被解釋其意義,此外,高維度的信號特徵也使得分類法則難以理解。
本研究以音樂的角度,重新定義了八個音樂特徵,並用來建立自動音樂情緒辨別系統的分類法則。這些音樂特徵並可以各自利用不同的信號特徵之組合而自動擷取。利用內部交叉驗證的結果,本實驗中利用音樂特徵建立的分類器達到了57% 的分類準確率。此準確率接近直接使用97個信號特徵學習出的分類器之分類準確率 (61%)。此外,應用主成分分析之方法,將97個特徵降維到18個主成分而訓練出的分類器之分類準確率 (46%),相較起本研究是較低的。
英文摘要 Emotion classification of music is important for the BMGIM music therapy domain. Currently, therapists would classify the emotion of music manually. The official classical music selections with their emotion labels for music therapy were suggested in the BMGIM book. Thus, an automatic emotion classification system in this study designed with these suggested classical music selections and their emotion labels in the BMGIM book. In the future, this system can be used to automatically classify different types of music outside of these suggested classical music.

Previous emotion classification researches built their classification models based on large set of signal level features and machine-learning algorithms. However, the classification model built with signal level features is hard to be interpreted. Besides, the high dimension of the signal level features also reduces the understandability of the classification model. These two issues cause the classification model built with large set of signal level features hard to be interpreted.

In this research, musical level features are developed to build the classification rules of the automatic emotion classification system of music. Eight musical level features were redefined from the music viewpoint in this study. These musical level features can be automatically extracted with different combinations of related signal level features. The classification system built with these musical level features achieves 57% precision rate through leave-one out cross-validation. This precision rate is close to the precision rate (61%) of the classification model trained with 97 signal level features directly. Besides, the precision rate (46%) of the classification model trained with 18 factors mapped from 97 signal level features by principle component analysis for dimension reduction is lower than this study.
論文目次 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Background and Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Related Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1 BMGIM Music Therapy and The Suggested Classical Music List . . . . . . . . . . . . . . 5
2.2 Emotion Taxonomy: The Hevner’s Adjective Circle . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Related Work of Automatic Emotion Classification . . . . . . . . . . . . . . . . . . . . . . 8
2.3.1 Regression Approach to Music Emotion Recognition . . . . . . . . . . . . . . . . . 8
2.3.2 Automatic Mood Detection and Tracking of Music Audio Signals . . . . . . . . . . 8
2.3.3 Prediction of Musical Affect Using a Combination of Acoustic Structural Cues . . 9
2.3.4 Overall Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3 Material and Method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1 Data Collection and Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.1.1 Data Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1.2 Data Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.3 Discussion of the Segmentation Process . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Comparison between The Musical Level Features and The Signal Level Features . . . . . 16
3.3 System Description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.4 Redefined Musical Level features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.4.1 Musical Level Feature: Loudness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.4.2 Musical Level Feature: Loudness Variation . . . . . . . . . . . . . . . . . . . . . . 26
3.4.3 Musical Level Feature: Articulation . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.4.4 Musical Level Feature: Harmony . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.4.5 Musical Level Feature: Mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.4.6 Musical Level Feature: Pitch Level . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.4.7 Musical Level Feature: Pitch Range . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.8 Musical Level Feature: Motion of Rhythm . . . . . . . . . . . . . . . . . . . . . . . 33
3.4.9 Musical Level Feature: Tempo (onset density) . . . . . . . . . . . . . . . . . . . . . 34
3.5 Relation between Musical Level Features and Signal Level Features . . . . . . . . . . . . . 36
3.5.1 The Available Signal Level Features . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.5.2 Linking Musical Level Features with Potentially Correlated Signal Level Features . 38
4 Experiment and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.1 Manually Observing The Classification Rule . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.2 Training and Evaluating The Musical Level Feature Extraction Model . . . . . . . . . . . 48
4.2.1 Discussion of The Musical Level Feature Extraction Model . . . . . . . . . . . . . 51
4.3 Building and Evaluating Emotion Classification System with Musical Level Feature Extraction Model
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.1 Discussion of The Classification System . . . . . . . . . . . . . . . . . . . . . . . . 54
4.4 Comparison with Classification Model Trained with Signal Level Feature Directly . . . . . 55
4.5 A Suggested Chinese Music List: Applying The Classification on The Chinese Music
Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
6 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A.1 Information of Music Excerpts in The Dataset . . . . . . . . . . . . . . . . . . . . . . . . 66
A.2 Signal Level Features List . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
參考文獻 [1] H. L. Bonny and L. M. Savary, Eds., Music and your mind: Listening with a new consciousness.
New York: Harper and Row, 1977.
[2] K. E. Bruscia and D. E. Grocke, Eds., Guided Imagery and Music: The Bonny Method and Beyond.
Gilsum: Barcelona Publishers, 2002.
[3] Y.-H. Yang, Y.-C. Lin, Y.-F. Su, and H. H. Chen, “A regression approach to music emotion recognition,”
IEEE Transactions on Audio, Speech, and Language Processing, vol. 16, no. 2, pp. 448–457,
February 2008.
[4] L. Lu, D. Liu, and H.-J. Zhang, “Automatic mood detection and tracking of music audio signals,”
IEEE Transaction on Audii, Speech, and Language Processing, vol. 14, no. 1, pp. 5–18, January
2006.
[5] M. D. Korhonen, D. A. Clausi, and M. E. Jernigan, “Modeling emotional content of music using
system identification,” IEEE Transactions on Systems, Man, and Cybernetics, vol. 36, no. 3, pp.
588–599, June 2006.
[6] M. Leman, V. Vermeulen, L. D. Voogdt, D. Moelants, and M. Lesaffre, “Prediction of musical affect
using a combination of acoustic structural cues,” Journal of New Music Research, vol. 34, no. 1,
pp. 39–67, 2005.
[7] R. H. Gundlach, “Factors determining the characterization of musical phrases,” The American
Journal of Psychology, vol. 47, no. 4, pp. 624–643, October 1935.
[8] K. Hevner, “The affective character of the major and minor modes in music,” The American Journal
of Psychology, vol. 47, no. 1, pp. 103–118, January 1935.
[9] K. Hevner, “Experimental studies of the elements of expression in music,” The American Journal
of Psychology, vol. 48, no. 2, pp. 246–268, April 1936.
[10] K. Hevner, “The affective value of pitch and tempo in music,” The American Journal of Psychology,
vol. 49, no. 4, pp. 621–630, October 1937.
[11] M. G. Rigg, “The mood effects of music: A comparison of data from four investigators,” Journal of
Psychology, vol. 58, pp. 472–438, 1964.
[12] P. N. Juslin and J. A. Sloboda, Music and emotion: theory and research. NewYork: Oxford
University Press, 2001.
[13] D. E. Grocke, “A phenomenological study of pivotal moments in guided imagery and music (gim)
therapy,” Ph.D. dissertation, Faculty of Music, University of Melbourne, October 1999.
[14] A. Storr, Music and the Mind. New York: The Free Press, 1992.
[15] C. L. Krumhansl, “An exploratory study of musical emotions and psychophysiology,” Canadian
Journal of Experimental Psychology, vol. 51, no. 4, pp. 336–352, November 1997.
[16] B. Y. Chua and G. Lu, “Perceptual rhythm determination of music signal for emotion-based classification,”
in Multi-Media Modelling Conference Proceedings, 2006 12th International, Beijing, China,
January 2006, pp. 4–11.
[17] Y. Zhu and M. Kankanhalli, “Precise pitch profile feature extraction from musical audio for key
detection,” IEEE Transactions on Multimedia, vol. 8, no. 3, pp. 575–584, June 2006.
[18] “Marsyas (music analysis, retrieval and synthesis for audio signals),” [Online],
http://marsyas.sness.net/.
[19] “Mirtoolbox, an integrated set of functions written in matlab,” [Online], available:
http://www.jyu.fi/hum/laitokset/musiikki/en/research/coe/materials/mirtoolbox.
[20] D. McEnnis, C. McKay, I. Fujinaga, and P. Depalle, “jaudio: A feature extraction library,” in Proceeding
of the 6th International Conference on Music Information Retrieval, London, UK, September
2005, pp. 600–603.
[21] “Clam: A tool for analysis, synthesis and processing of audio signals.” [Online], available:
http://clam.iua.upf.edu.
[22] S. L. Cessie and J. C. V. Houwelingen, “Ridge estimators in logistic regression,” Applied Statistics,
vol. 41, no. 1, pp. 191–201, 1992.
[23] I. H. Witten and E. Frank, Data Mining: Practical machine learning tools and techniques, 2nd ed.
San Francisco, USA: Morgan Kaufmann, 2005.
[24] D. Temperley, “A bayesian key-finding model,” in Music and Artificial Intelligence, C. Anagnostopoulou,
M. Ferrand, and A. Smaill, Eds. Berlin: Springer-Verlag, 2002, pp. 195–206.
[25] C. L. Krumhansl, Cognitive Foundations of Musical Pitch. Oxford, U.K.: Oxford University Press,1990.
論文全文使用權限
  • 同意授權校內瀏覽/列印電子全文服務,於2018-08-14起公開。
  • 同意授權校外瀏覽/列印電子全文服務,於2018-08-14起公開。


  • 如您有疑問,請聯絡圖書館
    聯絡電話:(06)2757575#65773
    聯絡E-mail:etds@email.ncku.edu.tw