Fast Online Incremental Learning with Few Examples For Online Handwritten Character Recognition.

Abdullah Almousa Almaksour 1 Harold Mouchère 1 Eric Anquetil 1
1 IMADOC - Interprétation et Reconnaissance d’Images et de Documents
UR1 - Université de Rennes 1, INSA Rennes - Institut National des Sciences Appliquées - Rennes, CNRS - Centre National de la Recherche Scientifique : UMR6074
Abstract : An incremental learning strategy for handwritten character recognition is proposed in this paper. The strategy is online and fast, in the sense that any new character class can be instantly learned by the system. The proposed strategy aims at overcoming the problem of lack of training data when introducing a new character class. Synthetic handwritten characters generation is used for this purpose. Our approach uses a Fuzzy Inference System (FIS) as a classifier. Results have shown that a good recognition rate (about 90%) can be achieved using only 3 training examples. And such rate rapidly improves reaching 96% for 10 examples, and 97% for 30 ones.
Document type :
Conference papers
Complete list of metadatas

Cited literature [11 references]  Display  Hide  Download
https://hal.archives-ouvertes.fr/hal-00491338
Contributor : Abdullah Almousa Almaksour <>
Submitted on : Friday, June 11, 2010 - 11:48:44 AM
Last modification on : Thursday, February 7, 2019 - 2:48:49 PM
Long-term archiving on: Friday, September 17, 2010 - 1:25:09 PM

File

ICFHR08.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00491338, version 1

Collections

CNRS | IRISA | UNIV-RENNES1 | INSA-RENNES | IRISA-D6 | UR1-UFR-ISTIC | UNIV-RENNES | INSA-GROUPE

Citation

Abdullah Almousa Almaksour, Harold Mouchère, Eric Anquetil. Fast Online Incremental Learning with Few Examples For Online Handwritten Character Recognition.. Eleventh International Conference on Frontiers in Handwriting Recognition (ICFHR'08), Aug 2008, Canada. pp.623-628. ⟨hal-00491338⟩

Export

BibTeX TEI DC DCterms EndNote

Share

Metrics

Record views

207

Files downloads

115

Contact

Informations

CCSD