Method for recognizing multi-language printed documents using strokes and non-strokes of characters

A method for extracting character features for recognizing characters, the method comprising the steps of:

• i) inputting the characters into an input means;
• ii) printing the input characters and scanning the printed characters to make character pictures;
• iii) constructing a standard input character set using the character pictures;
• iv) normalizing the character pictures to a fixed size;
• v) converting each of the size-fixed character pictures into mesh-type character composed of N×N mesh-blocks, wherein the N is a positive integer number and each of the N×N mesh-blocks has a mesh value of corresponding part of the size-fixed character;
• vi) extracting a stroke feature of the mesh-type character based on the mesh values of mesh-blocks related to a stroke part of the size-fixed character;
• vii) extracting a non-stroke feature of each of the mesh-type characters based on the mesh values of mesh-blocks related to a non-stroke part of the size-fixed character;
• and
• viii) extracting the character feature of the size-fixed character using the stroke feature and the non-stroke feature, wherein the step vi) includes the steps of:
• vi-1) accumulating position values of mesh-blocks related to the stroke of the size-fixed character;
• and
• vi-2) extracting the stroke features from the accumulated position values, wherein the step vi-2) has the steps of:
• vi-2-1) calculating accumulation frequency and an accumulation value of the Hamming distance with respect to each mesh-block related to the stroke of the size-fixed character in a mesh block set and storing the calculated accumulation frequency and the accumulation value of the Hamming distance of each mesh-block related to the stroke of the size-fixed character in the mesh block set, wherein the mesh block set has a predetermined number of the mesh blocks;
• vi-2-2) selecting the highest value among the stored calculated accumulation frequency and the accumulation values in the step xi);
• vi-2-3) normalizing a sum of weight of the mesh block set to a fixed integer;
• and
• vi-2-4) selecting the stroke feature from the highest values among the mesh block set of the size-fixed character based on the normalized integers.