rami el hajjx

lovethreewayΤεχνίτη Νοημοσύνη και Ρομποτική

20 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

65 εμφανίσεις


Rami Al
-
Hajj


Co
-
authors :


Chafic Mokbel , Laurence Likforman


Title :



Handwriting Recognition Using Hybrid Systems and Automated Learning Techniques


Abstract :



The recognition of cursive handwriting is still an open problem due to the existence of
many difficulties such as the variability of the handwritten styles and shapes, writing skew
or slant and the size of the lexicon.

The recognition of writing has many

applications such as mail sorting, bank cheque
reading, automatic localisation and reading of Vehicle plates…, and more recently, the
recognition of historical manuscripts. Many learning machines techniques are used for
automatic handwriting recognition l
ike: Hidden Markov Models HMM, Artificial Neural
Networks ANN, and others.

Moreover HMMs systems stochastically model sequences of variable length and cope with
nonlinear distortions along one direction.

We present and describe our bi
-
stage system for the

off
-
line Arabic cursive handwritten
word recognition, based on an analytical approach. The first phase is for recognition by
using three parallel classifiers of type HMM. Each of these classifiers produces a list of
candidates as results. The second phase

is for combination of the produced results.
Different strategies of combination and voting are presented (voting by summation, using
and Artificial Neural Network,


…) and compared.

The first version of our system has won the first Competition organized f
or the recognition
of Arabic Handwriting in Seoul
-
Korea at ICDAR2005. Recently added enhancements as well
as obtained results are presented and compared to other systems.

References :

[1] Ramy El
-
Hajj, Laurence Likforman, Chafic Mokbel, “Combining slanted
classifiers for an
improved HMM
-
based Arabic handwriting recognition ", IEEE “Transactions on Pattern
Analysis and Machine Intelligence”
-

TPAMI
-

IEEE,(2008).

[2] Ramy El
-
Hajj, Chafic Mokbel, Laurence Likforman, “ Combinaison de Classifieurs HMMs
a fenêtre
s symétriques et Asymétriques pour la reconnaissance de mots manuscrits
Arabes’’.


CIFED’08, Rouen


France.

[3] Rami AL
-
HAJJ, Chafic MOKBEL, and Laurence LIKFORMAN
-
SULEM, “Recognition of
Arabic Handwritten Words using Contextual Character Models”. Documen
t Recognition and
Retreival XV, Part of the IS & T/ SPIE International Symposium on Electronic Imaging.
January 2008


San Jose, California, USA.

[4] Ramy El
-
Hajj, Chafic Mokbel, Laurence Likforman, “Combination of HMM
-
Based
Classifiers for the Recognition

of Arabic Handwritten Words”. The 9th International
Conference on Document Analysis and Recognition ICDAR 2007, Curitiba


Brazil.