2. Introduction
Objectives
Scope and application
Features
System diagrams
Methodology
Technologies used
Future recommendation
Results
Overview
3. A Real time Devanagari handwritten character recognition
Converts to digital text
Recognize only Devanagari character
Developed for android mobiles
Based on pattern matching using signatures
Introduction
4. To build the system that recognize the Real time
handwritten Devanagari character
To build user friendly input gesture for providing any style
of Devanagari character
Objectives
5. Improve of human computer interface for computer illiterate
people by providing various computing services on inputs
Can be implemented on smart phones, tablets as an virtual
keyboard
The system can create paperless environment by digitizing
handwritten character
Scope & Applications
6. User can write the Devanagari character on their own styles
Output will be displayed as a digital text
User friendly GUI
Attractive input gesture
Features
9. Training
Pre-processing
Feature Extraction
Trained pattern
Recoginition
Pre-processing
Feature extraction
Character Recognition
Post processing
Methodology
10. The process of smoothing the data by removing the noise is
pre-processing
Gaussian low pass filter is used to remove noise and smooth
the data
The equation for the Gaussian low pass filter is
where, σ is the standard deviation of the distribution
Pre-processing
11. Extracting the features from the input samples
Features are number of strokes, length of stroke, width-
height ratio, number of intersection points, number of loops,
number of hooks, point density, initial point position and
direction, end point position and direction, initial to end
Sampling and geometric transformation is used
Geometric transformation includes scaling, translation and
rotation
Features extraction
12. The patter of the signature is created and stored
The pattern is determined using signature based algorithm
Trained Pattern
13. Step 1: create an empty signature
Step 2: Get the values of co-ordinates of two points i.e.
start and end points from the input
Step 3: Compute the slope of two points and hence angle of
deviation
Step 3: Compare the angle and append their symbol to
If angle If( degree>=0 and degree<45) and ( degree>=315
and degree<=360) assign “R” i.e. right to the signature
If( degree>=45 and degree<135) then append ‘U’ to string
signature.
Signature based algorithm
14. If( degree>=135 and degree<225) then append ‘L’ to string
signature
If( degree>=225 and degree<315) then append ‘D’ to
string signature.
Step 4: Go to 2 till end of input stroke
Step 5: Return signature
Signature based algorithm
15. Example
For the above input stroke the signature will be: 1-2:D,2-
3:U,3-4:L,4-5:D,5-6:R,6-7:D
Hence signature is: DULDRD
16. Recognition is the process of testing the input and identify
the desired output
In recognition first pattern of the signature is created using
the signature based algorithm and best match is identified
using the k-nearest neighbor’s (KNN) algorithm.
Recognition
17. k nearest neighbor is an non parametric method used for
classification and regression in pattern recognition.
An object is classified by a majority vote of its neighbors,
with the object being assigned to the class most common
among its k nearest neighbors
If k = 1, then the object is simply assigned to the class of
that single nearest neighbor, if k=2 then two nearest points
are assigned and so on.
k-NN algorithm
18. k-NN algorithm
Consider the following point distribution:
If the nearest neighbor of blue star is to be found than it can
be determined as:
20. To measure the distance between points A and B in a feature
space.
Euclidean distance function is the most widely used one to
calculate the nearest distance and given by the formula
Distance function
21. Cosine similarity measure is typically used to calculate
similarity values between documents in text retrieval by
Distance function
where, the numerator represents the dot product of the
vectors A⃗ and B⃗
denominator is the product of their Euclidean lengths.
22. Android Studio – IDE
Programming language as Java
MySQL database
Technology used
23. Characters with modifiers can be implemented
Recognition of words, lines, numbers and paragraph
Can be used as a keyboard with simple gesture input
Can be implemented in other different language
Future Recommendation