Isolated handwritten Arabic letters recognition using online structural approach and Freeman Chain Code

الملخص EN

In this paper, a structural pattern recognition approach is proposed to identify online handwritten Arabic letters.

The proposed technique depends on extracting directions among drawn pixels.

The sequence of drawing directions is processed to extract a set of ordered structural components.

Letters extraction uses a set of predefined models of sorted components that represent all Arabic letters.

Experimental results show that the recognition rate is 95.21% for the drawn scripts.

When the rejected cases are excluded from the calculation, the rate can be increased to 97.98%.