基于贝叶斯理论的单音音乐节奏提取

The Extraction of Rhythm from Audio Monophonic Music Based on Bayesian Theory

音频音乐信号的特征分析具有广泛的应用前景,如音乐检索、乐器识别等。本文先引入贝叶斯节奏模型,然后用基于贝叶斯理论的序列蒙特卡罗方法,推断音乐片段的小节和节拍的位置。对不同乐器演奏的、不同音乐速度的、不同节奏模式的音乐,本方法均能有效地提取节奏特征。

The analysis on audio musical signals have extensive application prospects, such as music retrieval, instrument recognition. This paper introduces a Bayesian Rhythm model, then uses the Sequential Monte Carlo method to estimate the meter and the beat positions. The method can be used to extracting rhythm features effectively, in despite of different instrumental performance, tempo or even different rhythm pattern.