The sensitivity of some recognition algorithms for printed Arabic characters

The sensitivities of four proposed algorithms (A1 – A4), used for the recognition of printed Arabic characters, have been investigated for different square normalization matrices with a dimension of "m" binary dots. The performed analysis shows that, for all algorithms, the sensitivity increase with the value of "m", however, m=4 to 8 may be adopted as a satisfactory values for recognition. The algorithm A1 has the best recognition sensitivity and A4 is useful for detecting the fine structure of some Arabic characters. It is concluded that a combined feature vector based on (A1 and A4) or (A2 and A4) will produce a new powerful algorithm. This algorithm is expected to be much more sensitive, as an Arabic recognition tool, than any of the four investigated algorithms.