作者: João Lobato Oliveira , Luís Paulo Reis , Fabien Gouyon , Luis Gustavo Martins
DOI:
关键词:
摘要: This paper describes a tempo induction and beat tracking system based on the efficient strategy (initially introduced in BeatRoot [Dixon S., “Automatic extraction of from expressive performances.” Journal New Music Research, 30(1):39-58, 2001]) competing agents processing musical input sequentially considering parallel hypotheses regarding beats. In this paper, we propose to extend causal continuous data. The main reasons for are threefold: providing more robustness potentially noisy data, permitting consideration number low-level frame-based features as input, opening way real-time uses (as e.g. mobile robotic platform). is implemented C++, faster than audio It integrated MARSYAS framework, therefore available under GPL users and/or researchers. Detailed evaluation non-causal versions common benchmark datasets show performances reaching those state-of-the-art trackers. We series lines future work careful analysis results.