Pengenalan Karakter Tulisan Tangan Jawi Menggunakan Metode New Relative Context dan SVM

Abstract

Dot is an important attribute in character recognition. Similarly in Jawi characters, a dot becomes a special characteristic that distinguish different characters with the same basic shape. Most of feature extraction methods only recognize the characters based on their basic shape and ignore the dots, such as Relative Context (RC). RC classifies characters with the same basic shape into a group. Therefore, the result recognition of RC is not individual characters, but the name of group character. To identify individual character, a new method for RC enhancement is introduced. The method is called New Relative Context (NRC). NRC works by separating characters into some areas. The wider area is defined as the basic shape, while other areas are defined as dot attribute. In this paper Support Vector Machine (SVM) is used to classify eleven sets of isolated Jawi characters. Eight sets of character images are used in the training phase, while in the testing phase three sets of images are used. The recognition rate of this method achieves 80%.