Online Intelligent Character Recognition For Gujarati Script (Indian Language)

ONLINE INTELLIGENT
CHARACTER
RECOGNITION FOR
HANDWRITTEN
GUJARATI SCRIPT
Project Developed By: Prince Arora
www.linkedin.com/in/princearora01
Guided By - Mr. Jignesh Dholakia, Director OCR Project,
MS University Of Baroda, India
© Online Handwritten Character Recognition For
Guajarati Script - Developed By Prince Arora
© First Published On Linked In On - 03-Dec-2018 -
https://www.linkedin.com/in/princearora01/
1

AGENDA
 Background
 Need & The Requirements
 Approach
 Solution
 Demo
 Real World Application
 Conclusion
2

BACKGROUND
3

CHARACTER RECOGNITION
 Humans can read and identify characters
 The problem of making the machines do the
same is character recognition
© First Published On Linked In On - 03-Dec-2018
- https://www.linkedin.com/in/princearora01/
4

ONLINE CHARACTER
RECOGNITION
 The recognition begins when the user inputs data into the device.
 Recognize the characters as they are being written on a device
 Data input – electronic pen
 Recognition engine
 Recognizes the handwritten character
 Converts it into typed form.
 Example of the devices :
 Tablet PCs
 PDAs
 Smart Phones
5

The Need & The Requirements
6

THE NEED
 Keying in data in hand held device is a
tedious job.
 Gujarati script has a large character set. Not
feasible and convenient to provide a key for
each character
 This system will ensure the portability of the
device is maintained
7

THE REQUIREMENTS
 Develop an online character recognition system
for handwritten Gujarati script (Indian Language)
 At the end system must have two components
 Drawing pad
 Display pad
 User will draw character on the drawing pad
 System will recognize the handwritten character
& display in computer typed form
8

APPROACH
9

APPROACH USED TO DEVELOP
THE SOFTWARE
 Prototyping process model
 Technology to be used to tackle with the
problem was unknown.
10

THE SOLUTION
11

THE SOLUTION
INPUT HANDWRITTEN CHARACTER
OUTPUT RECOGNIZED TYPED CHARACTER
PREPROCESSOR
DRAWING PAD
FEATURE EXTRACTOR
RECOGNIZER
DISPLAY PAD
12

PREPROCESSOR
 Handwritten character exhibit variations in
size, shape & position
 Preprocessing steps tackle these problems
 Position normalization – achieved by
cropping
 Size normalization – achieved by down
sampling
13

CROPPING
14

CROPPING
15

HIGH RESOLUTION IMAGE
16

5x7 DOWNSAMPLED IMAGE
17

FEATURE EXTRACTOR
 How do you classify a person as male or
female , even if distance between you and
the person is too large?
 How do u recognize an old school friend even
if u meet him after so many years in a coffee
shop?
 How do u recognize a person’s voice on a
telephone conversation?
18

ANSWERS
 Invariant characteristics or features for
classification and recognition.
 Invariant ⇒ don’t change over a period of a time.
 Extracts the feature for each character
 Stores the features
 Input features to recognition engine
19

RECOGNIZER
 Approaches
 Hamming code
 Neural networks
 Stroke based recognizer
 The selection of recognizer depends on the
type of features chosen.
20

NEURAL NETWORKS BASED
RECOGNIZER
 Simulate human brain
 Train the network until it can discriminate between
the classes of inputs.
 Neural networks resemble the human brain in the
following two ways:
 Acquires knowledge through learning.
 Store the knowledge within inter-neuron connection
strengths known as synaptic weights
21

NEURON – THE
COMPUTATION UNIT
 BIOLOGICAL  ARTIFICIAL
22

RECOGNIZER
 The features obtained during feature extraction
process are input to neural networks algorithm
 Each of the neuron in the output layer is
specialized for recognition of only one character.
 During the recognition process one of the output
neurons wins and is fired.
 This results in identification of the character with
which the neuron is associated.
23

DISPLAY PAD
 Converts the processed image obtained from
the recognition engine into the glyph
representing the recognized character
 Displays the recognized character in typed
form.
24

DEMO
25

Working Demo Of The Solution
 https://www.youtube.com/watch?v=B_a6cQmD-z4
26

Real World Application
 This problem holds lots of importance
 The industry is moving towards reducing the size
of the handheld devices.
 One mode of keying in information would be
direct writing. For this, the system needs to
interpret and identify what is being written.
 The solution has the ability to be extended for
other languages
27

CONCLUSION
 A successful attempt has been made to
create a base for this technology by
developing a prototype.
 This helps us in making further attempts in
order to improve the system.
 The simulator proves our concept developed
for recognition of Gujarati characters
28

THANK YOU
29

Online Intelligent Character Recognition For Gujarati Script (Indian Language)

Recommended

Recommended

More Related Content

Similar to Online Intelligent Character Recognition For Gujarati Script (Indian Language)

Similar to Online Intelligent Character Recognition For Gujarati Script (Indian Language) (20)

Recently uploaded

Recently uploaded (20)

Online Intelligent Character Recognition For Gujarati Script (Indian Language)