A new database of confusing characters for testing character recognition algorithms

In this paper a new technique and software tool for the generation of systematically organised confusing characters is presented; by means of this tool it is possible to transfer human knowledge on character recognition into an open database in a systematic way. This base is useful to test automatic character recognisers by comparing their performance to human unchallenged ability in recognising handwritten characters. This base can be used also to accomplish experiments in the field of confusing character recognition, where the comparison with human knowledge is necessary. In fact the need to increase knowledge on human learning and classification mechanisms has been underlined, in order to design new and more efficient classification algorithms consistent with human recognition processes, also by studying the properties of confusing characters.