An Architecture for Optical Music Recognition of Numbered Music Notation.
Journal
International Conference on Internet Multimedia Computing and Service, ICIMCS '14, Xiamen, China, July 10-12, 2014
Pages
241
Date Issued
2014
Author(s)
Wu, Fu-Hai Frank
Abstract
Music notations, for example staff notation and numbered music notation, are the most precise media for musical applications, since they contain complete information for both scores and signatures. The purpose of this study is to empirically define an optical music recognition (OMR) system architecture to transcript numbered music notation. The architecture includes several parts: preprocessing of document image, partition of semantic zones, class definition and model training, extraction and recognition of notations, music semantic model, media for audiovisual inspection, and performance metrics of processing steps. Moreover, the OMR prototype with the algorithms illustrates the framework of architecture and plays an important role in constructing the dataset from the songbook with the semi-automatic groundtruth and attributes for publicity and internet multimedia applications.
SDGs
Type
conference paper