Traditional character recognition methods construct a grid and represent a character as a combination of grid dots. By using a qualitative representation method [1,2], we introduce here a new approach to representing English letters. Using properties of the qualitative representation method, we can uniquely represent each English letter regardless of the size and position of each stroke on the board. Our success of implementation by neural networks shows the feasibility of the method, and our tolerance tests show its robustness.
Bibliographical noteFunding Information:
*This research was supported by AFOSR under Grant No. 890158.