臺灣博碩士論文加值系統

一首樂曲就和一篇文章一樣，由幾個句子組成一個段落，幾個段落組成一個完整的篇章。如果我們朗讀一篇沒有標點符號的文章，那將不知道何時該停頓，因而念得上氣不接下氣，語調既平淡又生硬，像個機器人。因此，音樂演奏時，演奏者如何正確的斷句也是件不可或缺的事。知名的作曲家蕭邦曾經對他的學生說:「一個演奏家斷句錯誤，就如同一個人不清楚自己所說的語言。」有了正確的斷句，才能在不同的句子分別加入像是漸強或漸弱、漸快或漸慢等不同的演奏技巧，讓一首曲子聽起來不會單調。因此本論文透過分析樂譜上的符號資訊，找出劃分樂句的方法，以用於合成出富有表情和抑揚頓挫的樂曲。
我們首先改良了現有音樂結構分析的演算法，該演算法是根據音樂中的旋律來找到一首歌中重複出現的結構，進而實現劃分樂句的功能。為了有更好的結果，除了旋律，我們也將音符時值和拍子納入我們劃分樂句的依據，並評估從旋律、音符時值和拍子這三樣資訊對於劃分樂句的成效。
我們使用了32首童謠作為資料庫；由於這些歌曲都是有歌詞的，因此就把歌詞中的標點符號，當作是樂句劃分的標準答案，並以此來驗證我們的演算法。
之後，我們便可以將劃分好的樂句應用於現有的音樂合成方法，賦予各個樂句不同的表情，使合成出的音樂更像真人在演奏。

A piece of music, like an article, consists of a few sentences that form a paragraph, and several paragraphs form a complete chapter. If we read an article without punctuation, we don’t know when to break, so we could be out of breath. Besides, the tone should be plain, like a robot. Therefore, when the music is played, it is also indispensable for the player to correctly phrase the sentence. The well-known composer Chopin once said to his students, “He who phrases incorrectly is like a man who does not understand the language he speaks.” With the correct phrasing, we can apply different musical expressions, such as crescendo, diminuendo, accelerando or ritardando, to make a song no longer monotonous. Therefore, this paper analyzes the symbolic representation information on scores to find out the method of musical phrasing, so that it can be used to synthesize the music with expression and intonation.
We first improved the existing music structure analysis algorithm, which is based on the melody of music to find the repeat structure of a song, and then realize the function of musical phrasing. In order to have better results, in addition to melody, we also included note value and beat into the basis of our musical phrasing, and evaluated the effect of the three kinds of information from the melody, note value and beat on the musical phrasing.
We used 32 nursery rhymes as a database; since these songs all have lyrics, we used the punctuation in the lyrics as the ground truth to the musical phrasing and used this to validate our algorithm.
After taht, we can apply the musical phrasing to the existing music synthesis method, giving different expressions to each phrase, making the synthesized music sounding like played by a human.

LIST OF TABLES VI
LIST OF FIGURES VII
Chapter 1 Introduction 1
1.1 Motivation 1
Chapter 2 Related Work 2
2.1 Phrasing 2
2.2 Synthesis 2
2.3 Musical Structure 3
Chapter 3 Materials 6
3.1 System Architecture 6
3.2 Possible Boundaries 6
3.3 Musical Phrasing 8
3.4 Synthesis by Phrases 9
Chapter 4 Evaluation 10
4.1 Dataset 10
4.2 Evaluation of Phrasing 10
Chapter 5 Discussion 12
Chapter 6 Conclusion and Future Work 13
References 15
Appendix I (PBMelody) 17
Appendix II (PBIOI) 19
Appendix III (PBBeat) 21
Appendix IV (PBMelody∩PBIOI) 23
Appendix V (PBMelody∩PBBeat) 25
Appendix VI (PBIOI∩PBBeat) 27
Appendix VII (PBMelody∩PBIOI∩PBBeat) 29

[1]Friberg, A., Bresin, R., Frydén, L. and Sunberg, J, Musical Punctuation on the Microlevel: Automatic Indentification and Performance of Small Melodic Units, 1998
[2]Cambouropoulos, E. The Local Boundary Detection Model (LBDM) and its application in the study of expressive timing. Proceedings of ICMC2001. pp.290–293. Havana, 2001
[3]Pei-Yu Chen, “Computer Music Form Analysis of Classical Piano Sonatas, National Cheng Kung University, 2018.
[4]Chih-Hong Yang, “Machine Learning Based Expressive Violin Synthesis and its Subjective Listening Evaluation, National Cheng Kung University, 2016.
[5]Chi-Ching Shih, “Analysis/Synthesis of the Violin Playing Style of J. Heifetz and D. Oistrakh with Application to Music Remix, National Cheng Kung University, 2017.
[6]Yen-Kuang Lu, “Analysis/Synthesis of the Violin Playing Style of J. Heifetz and D. Oistrakh, National Cheng Kung University, 2018.
[7]Masataka Goto, Hiroki Hashiguchi, Takuichi Nishimura, and Ryuichi Oka, “RWC music database: Database of copyrightcleared musical pieces and instrument sounds for research purposes, Transactions of Information Processing Society of Japan(IPSJ), vol. 45, no. 3, pp. 728–738, 2004.
[8]Chih-Hong Yang, Pei-Ching Li, Alvin W.Y. Su, Li Su, and Yi-Hsuan Yang, “Automatic violin synthesis using expressive musical term features, in Proc. of the 16th International Conference on Digital Audio Effects (DAFx), 2016.
[9]Sidorov, Kirill, Jones, Andrew Clifford and Marshall, Andrew David, “A smallest grammar approach to the symbolic analysis of music, Kostagiolas, Petros, Martzoukou, Konstantina and Lavranos, Charilaos eds. Trends in Music Information Seeking, Behavior, and Retrieval for Creativity, Advances in Multimedia and Interactive Technologies, Hershey, Pennsylvania, USA: IGI Global, pp. 228-257, 2016.
[10]C.G. Nevill-Manning and I.H. Witten, “Online and offline heuristics for inferring hierarchies of repetitions in sequences, Proc. IEEE, volume 88, pages 1745–1755, 2000.
[11]M. Dolson, “The phase vocoder: A tutorial, Computer Music Journal, vol. 10,
no. 4, pp. 14–27, 1986.
[12]Wickland, David D. Evaluating Melodic Similarity Using Pairwise Sequence Alignments and Suffix Trees. Diss. 2017.