Skip to Main Content
In this paper a novel approach for lexicon reduction of Farsi words is proposed. For this purpose we extract upper and lower profiles, vertical projection profile and black/white transition from word images. Using DTW similarity between words in the database is measured. The Isoclus algorithm is used to cluster handwritten word images of training dataset. The initial center of clusters is determined from agglomerative hierarchical clustering algorithm. Experimental results on IRANSHAHR dataset show a promising result. It yields a lexicon reduction of 77% with accuracy of 94%. We also evaluate the proposed system when combination of statistical features and different type of distance measures are used.