狠狠撸

狠狠撸Share a Scribd company logo
Touchless writer
飩� First
      of all we would like to
 pay glowing tribute to the
 Language Movement martyrs
 who sacrificed their lives for
 the mother tongue in 1952.
Touchless writer
Touchless writer
Touchless introduces a new way of interacting
with the computers by means of object
tracking through webcams for Bengali
character writing.
Here data is inserted for
writing purpose using webcam
without use of keyboard or on-
screen keyboard by mouse.
飩�   Webcam
飩�   Pen with a head of red color/any colored object
飩�   Windows Platform
飩�   Avro Keyboard installed
飩�   .Net Framework 3.5
飩�   Capturing video using aforge .net
飩�   From the tracking environment it first detect
    the red colored object and mark it with a
    rectangle using EuclideanColorFiltering.
飩�   Getting (x , y) coordinate of the rectangle and
    putting pixel at that point on a white panel.
飩�   So finally we get 33*30 pixel bitmap image
    which is ready for neural network input.
Touchless writer
飩� Training network
飩� Recognize character

飩� Speak out character
Touchless writer
Touchless writer
Touchless writer
Touchless writer
So how neural network
      works ?
Retrieve data


  Feature
 Extraction

  Training



  Testing
飩�   Total number of input = 33*30 = 990 pixels
In this method it scans the binary image
until it finds the boundary. The searching
follows according to the clockwise
direction.
For any foreground pixel p, the set of
all foreground pixels connected to it is
called connected component containing
p.
The pixel p and its 8-neighbors
are shown in Figure 4. Once a
white pixel is detected, it checks
another new white pixel and so
on.
p

   FIG: pixel p with its 8 adjacent
After feature extraction our input will
     approximately reduced to 67%
Touchless writer
In this phase we will test the
network by giving some
patterns. We match it with
every trained pattern and find
out the pattern that gives
highest match and lowest
match also.
飩�   Visual studio 2008
飩�   XML
飩�   Avro Keyboard installed
飩�   Aforge .Net
飩�   Due to brightness and contrast
    sometimes webcam can hardly
    detect the expected color.

飩�   Because of the similarity of tracking
    environment background color and
    object color the writing panel gets
    unexpected pixels.
飩�   As we draw character using object
    movement it is not properly drawn
    as like as original character,
    sometimes it becomes totally
    different from the original. For that
    reason neural network can鈥檛
    understand or recognize the original
    character and it outputs wrong
    character as input value or character.
飩�   Add facility for writing for both
    Bangla and English
飩�   Add facility to make the software
    capable of running without the help
    of keyboard and mouse.
飩�   Adding printing capabilities of
    written text.
飩�   Adding written text reading
    capabilities in Bangla.

飩�   Adding capabilities of tracking more
    than one object and take several
    decisions depending on object
    combinations
飩�   Microsoft Press Microsoft Visual C Sharp 2008
    Step by Step
飩�   Beginners C#.net 2005 Worx Publication
飩�   Professional C#.net 2005 Wrox Publication
飩�   MSDN Library
飩�   www.c-sharpcorner.com
飩�   www.codeproject.com
飩�   www.aforgenet.com
Touchless writer

More Related Content

Touchless writer

  • 2. 飩� First of all we would like to pay glowing tribute to the Language Movement martyrs who sacrificed their lives for the mother tongue in 1952.
  • 5. Touchless introduces a new way of interacting with the computers by means of object tracking through webcams for Bengali character writing.
  • 6. Here data is inserted for writing purpose using webcam without use of keyboard or on- screen keyboard by mouse.
  • 7. 飩� Webcam 飩� Pen with a head of red color/any colored object 飩� Windows Platform 飩� Avro Keyboard installed 飩� .Net Framework 3.5
  • 8. 飩� Capturing video using aforge .net 飩� From the tracking environment it first detect the red colored object and mark it with a rectangle using EuclideanColorFiltering. 飩� Getting (x , y) coordinate of the rectangle and putting pixel at that point on a white panel. 飩� So finally we get 33*30 pixel bitmap image which is ready for neural network input.
  • 10. 飩� Training network 飩� Recognize character 飩� Speak out character
  • 15. So how neural network works ?
  • 16. Retrieve data Feature Extraction Training Testing
  • 17. 飩� Total number of input = 33*30 = 990 pixels
  • 18. In this method it scans the binary image until it finds the boundary. The searching follows according to the clockwise direction.
  • 19. For any foreground pixel p, the set of all foreground pixels connected to it is called connected component containing p.
  • 20. The pixel p and its 8-neighbors are shown in Figure 4. Once a white pixel is detected, it checks another new white pixel and so on.
  • 21. p FIG: pixel p with its 8 adjacent After feature extraction our input will approximately reduced to 67%
  • 23. In this phase we will test the network by giving some patterns. We match it with every trained pattern and find out the pattern that gives highest match and lowest match also.
  • 24. 飩� Visual studio 2008 飩� XML 飩� Avro Keyboard installed 飩� Aforge .Net
  • 25. 飩� Due to brightness and contrast sometimes webcam can hardly detect the expected color. 飩� Because of the similarity of tracking environment background color and object color the writing panel gets unexpected pixels.
  • 26. 飩� As we draw character using object movement it is not properly drawn as like as original character, sometimes it becomes totally different from the original. For that reason neural network can鈥檛 understand or recognize the original character and it outputs wrong character as input value or character.
  • 27. 飩� Add facility for writing for both Bangla and English 飩� Add facility to make the software capable of running without the help of keyboard and mouse. 飩� Adding printing capabilities of written text.
  • 28. 飩� Adding written text reading capabilities in Bangla. 飩� Adding capabilities of tracking more than one object and take several decisions depending on object combinations
  • 29. 飩� Microsoft Press Microsoft Visual C Sharp 2008 Step by Step 飩� Beginners C#.net 2005 Worx Publication 飩� Professional C#.net 2005 Wrox Publication 飩� MSDN Library 飩� www.c-sharpcorner.com 飩� www.codeproject.com 飩� www.aforgenet.com