音樂節拍辨識與舞步檢測系統之研究

Abstract

如何準確對到音樂拍子對於舞者的表演來說是一件很重要的課題。本文提出一種採用音樂能量觀點來抓取拍子的音樂節拍辨識系統，並搭配檢測系統，來捕捉舞者的腳部動作。只需比對節拍與舞步數據後，可進而得知舞者是否有對到拍，可達到客觀與準確的判斷。本研究有助於舞者克服對拍上的問題。 本文針對華爾滋與探戈兩種腳步動作較明顯且大眾的音樂曲風進行測試，節拍辨識率可以達到87%以上，節拍速度辨識可達到90%以上。最後，分析實際測試數據，驗證本文中的音樂節拍辨識與舞步檢測系統工作良好且理論吻合。

Dancing exactly on the beat of a performance is one of the important issues for a dancer. A method is proposed in this thesis for music beat tracking analysis from an energy perspective and a dance step detection system is connected to capture the movement of the dancer. The dancer’s dance step information is compared with the music beats to check if the dancer is dancing on the beat. The paper can help dancers overcome difficulties on beat problems objectively and accurately. In this thesis, we choose Waltz and Tango for testing since the movements of the two dancing styles are more obvious and popular. The recognition rate of the beat tracking is over 87% , the recognition rate of tempo is over 90%. Finally, we analyze the experiment data and proved that our system works well and the theory matches the experiment data.