利用 Kinect 裝置自動產生舞蹈系統

隨著數位多媒體的發展與網路傳播速度的增加，愈來愈多研究有關於動作與音樂的合成。在多媒體以電腦視覺領域下，如何產生自然與流暢的跳舞動作已經變成一個很重要的議題。傳統上，人們在跳舞的時候是隨著音樂而自然地擺動身體，根據旋律、節奏自然的產生相對應的動作。然而，在設計一個舞蹈時，往往需要大量時間且不段練習以及專業編舞知識才能設計出一個好的舞蹈作品。在這邊論文當中，我們介紹了一個使用機器學習的方法應用在編舞技術來產生一個舞蹈作品，對於給一段音樂而產生相動應的舞蹈動作。在我們的系統當中，我們使用了低成本的動作擷取設備來幫助我們錄製舞蹈動作以解決在傳統上動作擷取設備昂貴的問題，在我們的實驗結果，顯示出我們的系統可以設計出一段自然且流暢的動作並且符合音樂節奏，旋律。

關鍵字

特徵擷取；動作合成；隱藏馬可夫模型；動作擷取； K平均演算法

並列摘要

As time goes by, there are more and more research and application about human motion and music with a rapidly increasing development of the digital animation and multimedia. How to generate a natural and smooth dance performance has been an important issue in computer graph and multimedia field. Traditionally, people design dance performance by listening to music, and they move their body with rhythm of music. For designing a dance performance, it took times to design a dance performance. In this thesis, we introduce a learning-based method for synthesizing dance movements of pieces of given music. We develop a system can automatic learn dance feature to design a new dance performance by given pieces of music without professional knowledge. We also use a cheap motion capturing device to solve the high cost problem of motion capturing device. In our work, the system can create dance performance as good as the dance performance designed by professional dancer. The experiment display that our approach and framework are useful for designing a dance performance without time cost and profession knowledge.

並列關鍵字

Feature Extraction ； KNN ； K-Mean ； HMM ； Motion Capture

參考文獻

[10] G.D. Tzeng, S.K. Chung. “Using Digital Music to Control 3D Character Motion “,

[2] Just Dance 4: Gangnam Style (Psy), http://youtu.be/Rjq2_Mg3pSA.

[3] Dawn of the Planet of the Apes, http://youtu.be/BFrWoDw-k54.

[4] R. Fan, S. Xu, W. Geng, “Example-Based Automatic Music-Driven Conventional Dance

no.3, pp.501,515, 2012

國際替代計量

利用 Kinect 裝置自動產生舞蹈系統

未授權

主題瀏覽