TRANSLITERASI CITRA AKSARA HIRAGANA MEMPERGUNAKAN JARINGAN BACKPROPAGATION

Abstrak

The purpose of this research is to understand the ability and to find out how much percentage of accuracy of Back Propagation algorithm in Japanese characters handwriting Hiragana’s pattern recognition. This study used the characterization of calculating a black pixel (intensity of character) and calculating the direction by applying masking diagonal left, right diagonal, vertical, and horizontal (mark direction). There were 797 letters that can be read correctly by backpropagation test equipment after applying feature combination 6 on the 7th test. Combination 6 had 5 charaterization feature 1, feature 2, feature 3, feature 4, and feature 5. Each feature was a colective feature of 9 segment which had explanation like this : feature 1 was Intensity of Character (Black), feature 2 was Mark Direction Diagonal 1 (Diag1), feature 3 was Mark Direction Diagonal 2 (Diag2), feature 4 was Mark Direction Horisontal (Horz) , and feature 5 was Mark Direction Vertical (Vert). This study revealed succeeded in proving that thebackpropagation algorithm was able to recognize Hiragana letters after achieving a success rate of accuracy above 85% which was 86.63%.S