Reconnaissance en-ligne de lettres manuscrites cursives par chaînes de Markov cachées

Eric Anquetil 1 Guy Lorette 2
1 IntuiDoc - intuitive user interaction for document
IRISA-D6 - MEDIA ET INTERACTIONS
2 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 : In this paper, we present an on-line handwritten character recognition system which is based on structured and logical modeling of handwriting using Hidden Markov Models. After some specific preprocessing, we extract two differen t classes of primitives which represent the two main aspects of handwriting : th e dynamic aspect for the notion of trajectory of the pen tip and the static aspec t for the notion of global geometry of the letter. We make an initial training to adjust the probabilities of each Hidden Markov Model. Then, the recognition system computes the probabilities of generation by each model of the letter to b e interpreted. This performs a clustering process based on similarity.
Document type :
Journal articles
Complete list of metadatas

https://hal.inria.fr/hal-01347197
Contributor : Eric Anquetil <>
Submitted on : Wednesday, July 20, 2016 - 3:34:32 PM
Last modification on : Thursday, February 7, 2019 - 3:05:42 PM

Identifiers

  • HAL Id : hal-01347197, version 1

Citation

Eric Anquetil, Guy Lorette. Reconnaissance en-ligne de lettres manuscrites cursives par chaînes de Markov cachées. Traitement du Signal, Lavoisier, 1995, 12 (6), pp.8. ⟨hal-01347197⟩

Share

Metrics

Record views

362