Computing Reviews

Online recognition of handwritten music symbols
Oh J., Son S., Lee S., Kwon J., Kwak N. International Journal on Document Analysis and Recognition20(2):79-89,2017.Type:Article
Date Reviewed: 07/14/17

There has been some progress with the automatic recognition of handwriting in recent years, and the digital analysis of handwritten music poses similar problems. In this case, the online recognition of music symbols means analyzing individual symbols as they are written as opposed to optical music recognition (OMR), which analyzes musical images from sheet music or printed scores. The authors of this paper have analyzed discrete musical symbols as each is created using a stylus pen and device. This analysis uses the location of the start and finish of the strokes that are deemed to make up a symbol. To classify the set of strokes for each symbol, three features are deployed: size and the histograms of directional and unidirectional movement angles. Staves, bar lines, and time signatures are not included in the analysis. A novel algorithm based on feature allocation matches each set of strokes against two datasets of handwritten symbols. The first dataset, called HOMUS, was created by previous researchers. For this study, the authors identified 23 different strokes, which made up the chosen subset of 24 HOMUS symbols. The other dataset, SNU, was created during the project by 23 subjects, some of whom were music students. The SNU symbols were classified against the 24 identified in the HOMUS dataset, producing a subset of 16 in comparison.

One advantage of the method described here is that the writer does not have to use special methods such as a set of ordered discrete strokes to create the symbols. Another is that the rate of matching is better than a previous method, which used a different algorithm. However, several handwritten music editing systems for phones and tablets are currently available and the paper makes no reference to any of them. In addition, the history of the real-time digital analysis of all forms of music is much older than the 20 years proffered in this paper--for example, the Computer Music Journal was founded in 1977, and optical character recognition (OCR) for music notation was first attempted in 1966 [1].


1)

Bainbridge, D.; Bell, T. The challenge of optical music recognition. Computers and the Humanities 35, 2(2001), 95–121.

Reviewer:  Rosa Michaelson Review #: CR145426 (1711-0754)

Reproduction in whole or in part without permission is prohibited.   Copyright 2024 ComputingReviews.com™
Terms of Use
| Privacy Policy