Telugu Letter Classifier

 

Problem:

Handwriting varies tremendously, based on various factors. We are trying to find a method that will recognize the letters regardless of the exact handwriting, so documents can be scanned in and then edited using a normal word processor.

 

Method:

 

This is fundamentally a classification problem, and therefore, solving it will require the implementation of a classifier.

I could not find a paper specific to this problem.  However, these were the methods mentioned in reference to the identification of handwritten numbers.

 

Gaussian Classifier

Baseline Nearest Neighbor Classifer

K-nearest neighbor classifer

Pairwise Linear Classifer

Principal Component Analysis

Radial Basis Function Network

Large Fully Connected Multi-Layer Neural Network:

Tangent Distance Classier (TDC)

PCA+quadratic

LeNet 1

LeNet 4

LeNet 4 / Local

LeNet 4 / K−NN

LeNet 5

Boosted LeNet 4

K−NN Euclidean

Tangent Distance

Soft Margin

Optimal Margin Classier (OMC)

 

Data Set:

There are no known databases of the Telugu script.

Therefore, I will have to collect my own data

 

By the Milestone:

I hope to have a complete Data set and narrow down the best method to distinguish Telugu letters.  I also plan to get some part of the programming done.