考量水平與垂直適搭度的自動音樂混搭製作系統

在這篇論文當中，我們提供了一個有效創作音樂混搭 (或稱之為音樂串燒) 的系統。音樂混搭是一種將複數曲子藉由重疊或銜接等技巧而創造出的重製音樂。鑑於前人的研究只有考慮將不同的曲子以重疊的方式疊入單一一首基底伴奏，我們提供的系統可以製作多個不同基底伴奏以及多個不同主弦律的音樂混搭成品。為此，當我們在搜尋適當的曲子來完成音樂混搭成品時，我們除了要衡量兩首歌一上一下重疊在一起 (垂直方向) 是否和諧外，還要衡量兩首歌一前一後銜接在一起(水平方向) 是否順暢。如此一來，我們的系統在結構上較前人的系統具有更高的彈性。在垂直方向上，我們提出了兩個新的衡量因子：“結構平衡” 與 “音量權重”，即為 “垂直適搭度”。而在水平方向上，我們引入了前人用來尋找適當銜接不同曲子的技術，即為 “水平適搭度”。將上述兩個方向的適搭度結合，我們定義了可能會遭遇的四種情境，以及針對這四種情境各自的處理方式。主觀實驗結果顯示，我們新提出的垂直適搭度相較於前人的研究，對使用者滿意度有統計上顯著的改善。此外，實驗結果也證明，我們所提出的四種情境也確實會影響使用者滿意的程度。

關鍵字

音樂混搭；音樂串燒；適搭度；垂直適搭度；水平適搭度；結構平衡；音量權重；創作性音樂資訊檢索

並列摘要

In this paper, we proposed a system to effectively create music mashups – a kind of re-created music that is made by mixing parts of multiple existing music pieces. Unlike previous studies which merely generate mashups by overlaying music segments on one single base track, the proposed system creates mashups with multiple background (e.g. instrumental) and lead (e.g. vocal) track segments. So, besides the suitability between the vertically overlaid tracks (i.e. vertical mashability) used in previous studies, we proposed to further consider the suitability between the horizontally connected consecutive music segments (i.e. horizontal mashability) when searching for proper music segments to be combined. On the vertical side, two new factors:“texture balance” and “volume weight” have been considered. On the horizontal side, the methods used to find proper segments for concatenation in the studies of medley creation are incorporated. Combining vertical and horizontal mashabilities together, we defined four levelsof mashability that may be encountered and found the proper solution to each one of them. Subjective evaluations showed that the proposed four levels of mashability can appropriately reflect the degrees of listening enjoyment. Besides, by taking the newly proposed vertical mashability measurement into account, the improvement in user satisfaction is statistically significant.

並列關鍵字

music mashup ； mashability ； vertical mashability ； horizontal mashability ； texture balance ； volume weight ； creative MIR

參考文獻

[10] Matthew Davies, Adam Stark, Fabien Gouyon, and Masataka Goto. Improvasher:

[15] Yin-Tzu Lin, I-Ting Liu, Jyh-Shing Roger Jang, and Ja-Ling Wu. Audio Musical

[5] Matthew E P Davies, Philippe Hamel, Kazuyoshi Yoshii, and Masataka Goto. Au-

[13] Hiromi Ishizaki, Keiichiro Hoashi, and Yasuhiro Takishima. Full-Automatic DJ

[20] D. Robinson. Perceptual Model for Assessment of Coded Audio. PhD thesis, Uni-

國際替代計量

考量水平與垂直適搭度的自動音樂混搭製作系統

全文下載

主題瀏覽