A Text-Independent Persian Writer Identification System Using LCS Based Classifier

Writer identification and verification's been considered by researchers. Because of several writing styles in Persian handwriting (i.e. Naskh, Nastaligh,...), partitioning the characters is almost impossible. Therefore texture based Persian writer identification systems are more useful. In this paper we have proposed a texture based algorithm that uses a new feature extracting method. The proposed algorithm uses a new classifier that unlike other methods is not based on similarity of two feature vectors, but it is based on the usual sequences that the features appear in the document (LCS (Longest Common Subsequence)). The system appeared to have about 95% accuracy in 100 Persian writers.