| 摘要: | 本研究提出了一種基於模擬退火演算法的旋律骨架生成方法，旨在開發出一種能夠
隨機生成輸入音樂上連貫且有意義的旋律結構之方法，同時保持計算效率。旋律骨架是
音樂創作過程中的關鍵基礎，提供了主要的旋律框架和節奏結構，為更複雜的旋律細節
發展奠定了基礎。本研究方法利用模擬退火算法在探索廣闊搜索空間及避免局部最佳解
的優勢，通過理解旋律中相對重要的音樂元素，並設計評估生成旋律品質的標準來實現
旋律骨架的生成。
我們詳細介紹了模擬退火算法在本問題上的具體實現，包括解空間的表示、目標函
數的設計和退火過程。為驗證所提方法的有效性，我們進行了與原始作曲的相似性比較
實驗。結果顯示，基於模擬退火演算法的方法能夠隨機生成符合音樂先驗知識且具有多
樣性的高品質旋律骨架，可參考其附錄 1。此外，我們還探索了該方法在自動伴奏生成、
風格轉換、數據集創建、音樂理論教育工具和互動音樂生成系統等各個領域的潛在應用。
總結來說，本研究引入了一種生成音樂旋律骨架的穩健且創新的方法，利用模擬退
火演算法的隨機性和音樂理論的先驗知識，未來具有在自動伴奏生成、風格轉換、數據
集創建、音樂理論教育工具和互動音樂生成系統等方面的廣泛應用前景。
;This study proposes a melody skeleton generation method based on the simulated
annealing algorithm, aiming to develop a method that can randomly generate musically
coherent and meaningful melodic structures while maintaining computational efficiency.
Melody skeletons are a key foundation in the music composition process, providing the
primary melodic framework and rhythmic structure, thus laying the groundwork for the
development of more complex melodic details. This study′s method leverages the advantages
of the simulated annealing algorithm in exploring a vast search space and avoiding local
optima, by understanding relatively important musical elements in melodies and designing
criteria to evaluate the quality of generated melodies.
We provide a detailed explanation of the specific implementation of the simulated
annealing algorithm in this problem, including the representation of the solution space, the
design of the objective function, and the annealing process. To validate the effectiveness of
the proposed method, we conducted similarity comparison experiments with original
compositions, the results of which can be referenced in Appendix 1. The results show that the
simulated annealing algorithm-based method can randomly generate high-quality melodic
skeletons that are consistent with musical prior knowledge and exhibit diversity. Additionally,
we explored the potential applications of this method in various fields, such as automatic
accompaniment generation, style transformation, dataset creation, music theory education
tools, and interactive music generation systems.
In conclusion, this study introduces a robust and innovative method for generating
musical melody skeletons, utilizing the randomness of the simulated annealing algorithm and
iii
the prior knowledge of music theory. In the future, this method has broad application
prospects in automatic accompaniment generation, style transformation, dataset creation,
music theory education tools, and interactive music generation systems. |
