Handwriting recognition using local methods for global recognition - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Conference Papers Year : 2001

Handwriting recognition using local methods for global recognition

Christophe Choisy
  • Function : Author
  • PersonId : 836386
Abdel Belaïd
  • Function : Author

Abstract

These last years, many systems for handwriting recognition were developped. Because of the variability of writing the elastic approachs such Hidden Markov Models, DTW, Markov Random Fields, that can deal with local deformations become very popular. These approach are based on local observations which ones are given by features extraction, and more recently by global models as Neural Networks. SVM and NN are efficient to recognize patterns with small deformations. In the case of words recognition the major difficulty is to deal with the length variability of images. A Recurrent NN was used [cite SENIOR] to overcome this problem but deals with a local context, loosing a part of the power of NN, and gives some difficults to label the frames in the learning step.
No file

Dates and versions

inria-00100457 , version 1 (26-09-2006)

Identifiers

  • HAL Id : inria-00100457 , version 1

Cite

Christophe Choisy, Abdel Belaïd. Handwriting recognition using local methods for global recognition. International Conference on Document Analysis and Recognition, 2001, Seattle, USA, 5 p. ⟨inria-00100457⟩
70 View
0 Download

Share

Gmail Facebook X LinkedIn More