ppt prakash

Upload: pprakasho

Post on 07-Apr-2018

243 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/3/2019 Ppt Prakash

    1/16

    SEMINAR REPORTON

    Recognition and Editing of Devnagari

    Handwriting Using Neural Network

    SUBMITTED BY

    Prakash A. Narkhede

    DEPT. OF ELECTRONICS AND TELECOMMUNICATIONANURADHA ENGINEERING COLLEGE

    SAKEGAON ROAD, CHIKHLI 443201

    AECC/ExTC/2009-10

    SEMINAR GUIDE

    14 DEC. 2009

    Prof. R. B. Mapari

  • 8/3/2019 Ppt Prakash

    2/16

    CONTENTS

    1. INTRODUCTION

    2. PROPERTIES OF DEVNAGARI SCRIPT

    3. STEPS INVOLVED

    3.1 CHARACTER SEPARATION

    3.1.1 LINE SEGMENTATION

    3.1.2 WORD SEGMENTATION

    3.1.3 CHARACTER SEGMENTATION

    3.2 PREPROCESSING.

    3.2.1 IMAGE BINARISATION

    . 3.2.2 THINNING OF BINARISED IMAGE

    3.2.3 WINDOWING

    3.3 CHARACTER RECOGNITION AND EDITING

    4. STEPS INVOLVED IN RECOGNITION OF CHARACTER

    4.1 MATRIX GENERATION

    4.2 NEURAL NETWORK

    4.3 ARCHITECTURE

    5. RESULTS

    6. CONCLUSION

    7. REFERENCES

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    3/16

    1. INTRODUCTION

    14 VOWELS AND 33 SIMPLE CONSONANTS

    COMPOUND CHARACTORS

    OCR ONE OF THE APPLICATION USED IN

    SCANNERS AND FAXES, EYE ,FACERECOGNITION ,IN BANKS, ROBOTICS FIELDetc.

    NN MEANS SIMPLY CREATION OF NETWORKTHAT WORKS LIKE HUMAN BRAIN

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    4/16

    2. PROPERTIES OF DEVNAGARI SCRIPT

    (a)

    (b)

    FIGURE 1: SAMPLES OF HANDWRITTEN DEVNAGARI BASIC

    CHARACTERS (a) VOWELS (b) CONSONANTS

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    5/16

    3. STEPS INVOLVED

    A). CHARACTER SEPARATION

    a). Line Segmentation

    b). Word Segmentation

    c). Character Segmentation

    B). PREPROCESSING

    a. Image Binarisation

    I(x, y) = 0 I(x, y) =t

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    6/16

    b. Thinning of Binarised Image

    c. Windowing

    FIGURE 2. THINNING OF BINARISED IMAGE.

    C). CREATING A CHARACTER RECOGNITION SYSTEM

    Character recognition by neural network

    Replacing the recognized characters by standard fonts.

    Assembling all the separated characters in the same order as they appeared

    in the input image to give final output.

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    7/16

    4. RECOGNITION OF CHARACTER

    A. Matrix generation

    B. Neural Network

    Network receives the 900 Boolean values as a 900- element inputvectorIt require 49-element output vector to identify the character

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    8/16

    C. Architecture

    The neural network needs 900 inputs and 49 neurons in itsoutput layer to identify the character

    The hidden (first) layer has 600 neurons

    Multilayer perceptrons trained by Error Back Propagation (EBP) algorithm.

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    9/16

    5. RESULTS

    FIGURE 5: SAMPLE OF IMAGE CONTAINING DEVNAGARI HAND WRITING

    FIGURE 6. HISTOGRAM OF IMAGE CONTAINING DEVNAGARI HANDWRITING. AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    10/16

    FIGURE 7. RESULT OF LINE SEPARATION

    FIGURE 8. RESULT OF WORD SEPARATION

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    11/16

    FIGURE 9. COMPLETE CHARACTER SEPARATION RESULTS

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    12/16

    FIGURE 10. COMPLETE PROCESS OF RECOGNITION BY NEURAL

    NETWORK AND EDITING

    AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    13/16

    FIG.11 INPUT IMAGE OF HANDWRITTEN DEVNAGARI AND FINAL OUTPUT

    OBTAINED FOR THE SAMPLE INPUT OF FIGUR4.AECC/ExTC/2009-10

    14 DEC. 2009

  • 8/3/2019 Ppt Prakash

    14/16

    6. CONCLUSION

    The method for recognition of devnagari characters usingneural network presented in this paper is able to successfullyrecognize most of the hand writings. However, the success ofthe method lies in the size of database, i.e. larger the size of

    database used for training the neural network higher isprobability of successful recognition. However the larger database places the limit on the speed of recognition, and hencethis method can be used for offline recognition.

    AECC/ExTC/2009-10

    14 DEC. 2009

    DEC

  • 8/3/2019 Ppt Prakash

    15/16

    7. REFERENCES [1] Krishnamachari Jayanthi ,Akihiro Suzuki,Hiroshi Kanai,Yoshiyuki

    Kawazoe, Masayuki Kimura and Keniti Kido, Devnagari characterrecognition using structure analysis, IEEE-1989.CH2766-4/89/0000- 0363.

    [2] Dileep Kumar, An AI approach to hand written Devnagari scriptrecognition, IIT Delhi.

    [3] Yi Li,Yefeng Zheng ,and David Doermann, Detecting text lines inhandwritten documents ,The 18th International Conference on PatternRecognition (ICPR'06).

    [4] K.H. Aparna, Vidhya Subramanian, M. Kasirajan, G. Vijay Prakash, V.S.Chakravarthy, Online Handwriting Recognition for Tamil , Proceedings ofthe 9th Intl Workshop on Frontiers in Handwriting Recognition (IWFHR-92004).

    [5] Fakhraddin Mamedov and Jamal Fathi Abu Hasna, Characterrecognition using neural networks Near East University, North Cyprus,Turkey via Mersin-10, KKTC

    [6] U. Bhattacharya and B. B. Chaudhuri, Databases for Research onRecognition of Handwritten Characters of Indian Scripts, Proceedings ofthe 2005 Eight International Conference on Document Analysis andRecognition (ICDAR05).

    AECC/ExTC/2009-10

    14 DEC. 2009

    14 DEC 2009

  • 8/3/2019 Ppt Prakash

    16/16

    THANK YOU

    14 DEC. 2009

    AECC/ExTC/2009-10