Using Optical Music Recognition to Encode 17th-Century Music Prints
- 16 October 2020
- conference paper
- conference paper
- Published by Association for Computing Machinery (ACM) in 7th International Conference on Digital Libraries for Musicology
Abstract
There have been several attempts to improve the retrieval of symbolic music information by Optical Music Recognition (OMR) to increase the “searchability” of digital music libraries of early music prints and to facilitate the collection of data for musicological research. Their success has varied. This report describes a new online OMR system based upon industry-standard platforms to automate the encoding of early 17th-century music prints. Due to our research on composers of canons in early 17th-century Rome, we have used as a test case the early music prints of Paolo Agostini. Agostini was maestro di cappella at St Peter’s Basilica and the most active exponent of advanced contrapuntal techniques, especially canon, in Rome in the 1620s. We developed a digital tool to process images of Agostini’s printed music and to classify 7,092 automatically selected objects according to 38 music symbols using supervised learning with convolutional neural networks (CNN). The resulting system, IntelliOMR, exhibits up to an average of 99% accuracy for classifying unseen items after 50 training epochs. It has proven effective for rapidly encoding all of Agostini’s works in the Music Encoding Initiative’s XML format for a critical edition and computer-assisted musical analysis. The approach and design of this digital tool offer significant opportunities for enhancing digital library systems and for future research projects investigating digital corpora of early printed music.Keywords
Funding Information
- Australian Research Council (DP180100680)
This publication has 19 references indexed in Scilit:
- Handwritten Music Recognition for Mensural notation with convolutional recurrent neural networksPattern Recognition Letters, 2019
- State-of-the-Art Model for Music Object Recognition with Deep LearningApplied Sciences, 2019
- Hybrid hidden Markov models and artificial neural networks for handwritten music recognition in mensural notationPattern Analysis and Applications, 2019
- Avoiding staff removal stage in optical music recognition: application to scores written in white mensural notationPattern Analysis and Applications, 2014
- Introduction to SIMSSA (Single Interface for Music Score Searching and Analysis)Published by Association for Computing Machinery (ACM) ,2014
- The MNIST Database of Handwritten Digit Images for Machine Learning Research [Best of the Web]IEEE Signal Processing Magazine, 2012
- Template Matching Techniques in Computer VisionPublished by Wiley ,2009
- The Challenge of Optical Music RecognitionComputers and the Humanities, 2001
- Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in positionBiological Cybernetics, 1980
- Receptive fields, binocular interaction and functional architecture in the cat's visual cortexJournal Of Physiology-London, 1962