Deep learning approaches for beat and downbeat tracking have brought advancements. However, these approaches continue to rely on hand-crafted, subsampled spectral features as input, restricting the information available to the model. In this work, we propose WaveBeat, an end-to-end approach for joint beat and downbeat tracking operating directly on waveforms. This method forgoes engineered spectral features, and instead, produces beat and downbeat predictions directly from the waveform, the first of its kind for this task. Our model utilizes temporal convolutional networks (TCNs) operating on waveforms that achieve a very large receptive field (= 30 s) at audio sample rates in a memory efficient manner by employing rapidly growing dilation factors with fewer layers. With a straightforward data augmentation strategy, our method outperforms previous state-of-the-art methods on some datasets, while producing comparable results on others, demonstrating the potential for time domain approaches.
https://www.aes.org/e-lib/browse.cfm?elib=21578
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.
Learn more about the AES E-Library
The Engineering Briefs at this Convention were selected on the basis of a submitted synopsis, ensuring that they are of interest to AES members, and are not overly commercial. These briefs have been reproduced from the authors' advance manuscripts, without editing, corrections, or consideration by the Review Board. The AES takes no responsibility for their contents. Paper copies are not available, but any member can freely access these briefs. Members are encouraged to provide comments that enhance their usefulness.
Start a discussion about this paper!
