Many different onset detection methods have been proposed over time, but other than for non-pitched percussive sounds, such as drums, there does not exist a general method that correctly identifies onsets with high accuracy. In this thesis, the domain is limited to pitched percussive sounds, particularly piano pieces.
This thesis focuses on statistical-based onset detection models. When compared to probabilistic and learning-based models, they perform very well, are simpler, and run much faster since no training is involved, all of which are ideal for an online setting. These models have a common workflow of building a spectrogram, applying an onset detection function to create an activation curve, and using a peak picking algorithm to select candidate onsets.
Even though they do not directly improve performance, pilot studies on the main stages of onset detection are described. This includes using a variable framerate spectrogram, an onset detection function that makes use of harmonic information, and dynamic peak picking thresholding using Otsu's method. These approaches touch on areas where onset detection can be improved and should be further investigated.
Finally, a new method is proposed, combining many techniques shown to improve onset detection. This includes doubling the framerate of the spectrogram, preprocessing the spectrogram with a log-scale and adaptive magnitude, and widening the frame index difference. The preprocessed time-frequency representation, though more complex than previous algorithms, is effective, allowing the use of a simple onset detection function. This new method is able to outperform both original implementations and state-of-the-art statistical-based methods on a subset of the MIDI Aligned Piano Sounds dataset.
|Commitee:||Chen, Yixin, Wilkins, Dawn E.|
|School:||The University of Mississippi|
|School Location:||United States -- Mississippi|
|Source:||MAI 58/02M(E), Masters Abstracts International|
|Keywords:||Onset detection, Piano|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be