The basic goal of the Adaptive Optical Music Recognition system presented herein is to create an adaptive software for the recognition of musical notation. The focus of this research has been to create a robust framework upon which a practical optical music recognizer can be built. / The strength of this system is its ability to learn new music symbols and handwritten notations. It also continually improves its accuracy in recognizing these objects by adjusting internal parameters. Given the wide range of music notation styles, these are essential characteristics of a music recognizer. / The implementation of the adaptive system is based on exemplar-based incremental learning, analogous to the idea of "learning by example," that identifies unknown objects by their similarity to one or more of the known stored examples. The entire process is based on two simple, yet powerful algorithms: k-nearest neighbour classifier and genetic algorithm. Using these algorithms, the system is designed to increase its accuracy over time as more data are processed.
| Identifer | oai:union.ndltd.org:LACETR/oai:collectionscanada.gc.ca:QMM.42033 |
| Date | January 1996 |
| Creators | Fujinaga, Ichiro. |
| Contributors | Pennycook, Bruce (advisor) |
| Publisher | McGill University |
| Source Sets | Library and Archives Canada ETDs Repository / Centre d'archives des thèses électroniques de Bibliothèque et Archives Canada |
| Language | English |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Format | application/pdf |
| Coverage | Doctor of Philosophy (Faculty of Music.) |
| Rights | All items in eScholarship@McGill are protected by copyright with all rights reserved unless otherwise indicated. |
| Relation | alephsysno: 001555812, proquestno: NQ29937, Theses scanned by UMI/ProQuest. |
Page generated in 0.0097 seconds