In this paper, a new fuzzy-logic approach is presented for character preclassif ication which gives a precise calculation method for the baseline detection algo rithm with tolerance analysis through analyzing the typographical structure of t extual blocks. Other virtual reference lines are extracted with clustering techn ique. In order to ensure correct character preclassification, a fuzzy-logic app roach is used to assign a membership to each typographical category for ambiguou s classes. The results prove that the proposed fuzzy typographical analysis for character preclassification is able to process to more than 10000 characters per second on a SUN 4/490 workstation and the method has been tested for different font sizes and different types with satisfaction.