Online Pen-Based Recognition of Music Notation with Artificial Neural Networks

• Published in: Computer music journal Vol. 27; no. 2; pp. 70 - 79
• Main Author: George, Susan E.
• Format: Journal Article
• Language: English
• Published: 238 Main St., Suite 500, Cambridge, MA 02142-1046, USA MIT Press 01.06.2003; The MIT Press; MIT Press Journals, The
• Subjects: Artificial neural networks; Aural learning; Computer music; Computer pattern recognition; Data Collection; Editing; Experiments; Human-Machine Interfaces for Music; Music criticism; Music learning; Musical Notation; Neural Networks; Pattern Recognition; Training; Writing pens; Writing tablets
• ISSN: 0148-9267; 1531-5169
• DOI: 10.1162/014892603322022673

With online music input, the prevailing interface technology for music editing involves the conventional "point-and-click" mouse-based paradigm. A survey of online music input technology is presented. The approach presented proposes a neural network classifier for identifying complete symbols. The assumption is that a whole symbol primitive has been isolated, and the raw data points of this complete symbol are presented to the neural-based recognizer for training and subsequent generalization. Artificial neural networks (ANN) are a non-parametric pattern recognition approach that was born with the field of artificial intelligence in the 1950s. In general, ANNs provide an approach that is closer to human perception and recognition than traditional computing. Experimental work performed in music symbol recognition is traced.