Music Symbol Recognition Based on Imbalanced Large Margin Distribution Machine

Music symbol classification is an significant step of Optical Music Recognition (OMR). Several remarkable algorithms are used to recognize music symbols. When the quantity of symbols from different classes is imbalanced, the classifying surface will incline towards the minority class. However, none...

Full description

Saved in:
Bibliographic Details
Published in2022 37th Youth Academic Annual Conference of Chinese Association of Automation (YAC) pp. 1208 - 1213
Main Authors Wen, Cuihong, Zhu, Longjiao
Format Conference Proceeding
LanguageEnglish
Published IEEE 19.11.2022
Subjects
Automation
Classification algorithms
Margin Distribution
Music Symbol Classification
Optical imaging
Optical Music Recognition(OMR)
Pattern Classification
Symbols
Online AccessGet full text

Cover

More Information
Summary:Music symbol classification is an significant step of Optical Music Recognition (OMR). Several remarkable algorithms are used to recognize music symbols. When the quantity of symbols from different classes is imbalanced, the classifying surface will incline towards the minority class. However, none of the studies considered the imbalance of the music symbols. In this paper, the Imbalanced Large margin Distribution Machine (ILDM) is proposed to increase the classification accuracy of the minority music symbols by introducing cost-sensitive penalty. We conduct tests on more than 10000 music symbols, which are handwritten and printed images. The tests show that the accuracy of the minority class is greatly improved with the increasing of the cost-sensitive penalty. The inefficiency of classification due to the over-widened gap in sample quantity of music symbols is solved by the ILDM.
DOI:10.1109/YAC57282.2022.10023625