Title: Music note position recognition in optical music recognition using convolutional neural network
Authors: Andrea; Paoline; Amalia Zahra
Addresses: Computer Science Department, Bina Nusantara University, Jakarta, 11480, Indonesia ' Computer Science Department, Bina Nusantara University, Jakarta, 11480, Indonesia ' Computer Science Department, Bina Nusantara University, Jakarta, 11480, Indonesia
Abstract: Technology improvement is rapidly changing. This impacts many fields, including the music field. Technology has helped the music field to be recognised in machine understanding. This field is called optical music recognition (OMR), a computer vision enabler in music. With OMR, we can define the position and music notation in music note. We propose a deep learning and convolutional neural network (CNN) approach to recognise a music position in music note. Music note position in staff is one of the keys to achieve pitch recognition. While we have music clef, key signature, and note position in staff, we can give machines the understanding of a note pitch. This experiment can bring and broaden the experiments in recognising music pitch, which take music note image as an input and position as the output. We use our own dataset and use CNN in experiments. This note position recognition experiment achieved 80% accuracy.
Keywords: optical music recognition; OMR; music; pitch recognition; deep learning; convolutional neural network; CNN; music sheet.
International Journal of Arts and Technology, 2021 Vol.13 No.1, pp.45 - 60
Received: 22 Feb 2020
Accepted: 23 Nov 2020
Published online: 16 Jun 2021 *