Call For Paper Volume:5 Issue:10 Oct'2018 |

A navel approach of gradient and non-gradient search algorithms for segmentation of hand written data

Publication Date : 12/03/2018


DOI : 10.21884/IJMTER.2018.5059.8RW9E

Author(s) :

Dr.S.Karunakar Reddy , Chilupuri Anusha.


Volume/Issue :
Volume 5
,
Issue 3
(03 - 2018)



Abstract :

The paper shows an attempt of numerical simulation of a navel approach of both gradient (ANN) and non-gradient algorithms (PSO) on the cursive hand written data for its segmentation. The text of the cursive hand written data belongs to English language. The implemented method in the present paper shows a fast converging nature and found to be free from the mathematical difficulty of local minima. The data is processed thrice through the different algorithmic approaches in the order of neural networks, Particle swarm optimization and simulated annealing technique. Then each characters of scanned text image have been converted to the column vectors of 700 values which are then fed to neural networks linked to PSO and SA blocks. The efficiency of segmentation is with the help of PSO and SA is enhanced to 80% when compared to previously attempted research studies. Temperature term in simulated annealing is a significant parameter which being controlled carefully will make the solution towards global minimum. Two significant observations are found in the present numerical study one is the process is free from local minima and other is fast convergence.


No. of Downloads :

6


Indexing

Web Design MymensinghPremium WordPress ThemesWeb Development

A navel approach of gradient and non-gradient search algorithms for segmentation of hand written data

March 3, 2018