CrazyEngineers
CrazyEngineers
  • Hello

    I have to develop a project regarding Punjabi character Recognition.User will draw a any character of Punjabi language and my software will detect and find out which character it is.Please suggest the way to start.I am thinking to develop this Project in .NET technology .I think using MATLAB functions,calling them will be easy .But actually I do'nt have any idea to start my project.


    There are some methods suggested for recognition Project like using Neural Networks and Artificial Intelligence and fuzzy logic etc.But I want to use a simple approach for this Project.Can there be any statistical approach like just comapring height:width ratio and recognising pixels of characters . Please guide me on this.


    suggest the simplest approach for this.other than using neural and all that.
    Replies
Howdy guest!
Dear guest, you must be logged-in to participate on CrazyEngineers. We would love to have you as a member of our community. Consider creating an account or login.
Replies

  • Predictor

    Member • Mar 12, 2010

    Neha Kochhar
    I have to develop a project regarding Punjabi character Recognition.User will draw a any character of Punjabi language and my software will detect and find out which character it is.Please suggest the way to start.I am thinking to develop this Project in .NET technology .I think using MATLAB functions,calling them will be easy .But actually I do'nt have any idea to start my project.


    There are some methods suggested for recognition Project like using Neural Networks and Artificial Intelligence and fuzzy logic etc.But I want to use a simple approach for this Project.Can there be any statistical approach like just comapring height:width ratio and recognising pixels of characters .
    The usual approach to such problems follows this rough outline:

    1. Gather raster images of characters to be recognized.
    2. Perform any needed pre-processing (adjusting contrast, cleaning noise, stray pixels, etc.)
    3. Extract informative features from each example character (vertical and horizontal projections, etc.)
    4. Train a learning system (could be may things: discriminant, neural network, etc.)
    5. Test the system.
    6. Celebrate because you're done!

    A very simple classifier to implement in software is the k-nearest neighbors (k-NN) model. k-NN simply checks the distance between the new character and all historical characters. The k closest matches vote, and the most popular example wins.
    Are you sure? This action cannot be undone.
    Cancel
Home Channels Search Login Register