Automatic Piano Sheet Music Transcription with Machine Learning
- 1 Bina Nusantara University, Indonesia
- 2 7 Square Gabriel Fauré, France
Abstract
Automatic Music Transcription (AMT) is becoming more and more popular throughout the day, it has piqued the interest of many in addition to academic research. A successful AMT system would be able to bridge multiple ranges of interactions between people and music, including music education. The goal of this research is to transcribe an audio input to music notation. Research methods were conducted by training multiple neural networks architectures in different kinds of cases. The evaluation used two approaches, those were objective evaluation and subjective evaluation. The result of this research was an achievement of 74.80% F1 score and 73.3% out of 30 respondents claimed that Bidirectional Long Short-Term Memory (BiLSTM) has the best result. It could be concluded that BiLSTM is the best architecture suited for automatic music transcription.
DOI: https://doi.org/10.3844/jcssp.2021.178.187
Copyright: © 2021 Fernandes Saputra, Un Greffin Namyu, Vincent, Derwin Suhartono and Aryo Pradipta Gema. This is an open access article distributed under the terms of the
Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
- 4,933 Views
- 2,801 Downloads
- 5 Citations
Download
Keywords
- Automatic Music Transcription
- Recurrent Neural Network
- Convolutional Neural Network