Structural Off-line Handwriting Character Recognition Using Approximate Subgraph Matching and Levenshtein Distance

Abstract

This research paper proposed alternatives method for off-line handwriting character recognition in structural approach. The task of character segmentation and normalisation is quite costly when processing large data, especially on off-line handwriting recognition with structural approach. The main idea of this paper is to model a handwritten character into string graph representation. The purpose of those model is to provide ability in improving recognition accuracy without relying in normalisation technique. The graph consists of several edges that indicate the connected vertices. The vertices are the curves that make the character. The curve is extracted by analyzing the character's chain code, and it's string feature is created using certain rules. In this paper, the similarity distance between graph is measured using approximate subgraph matching and string edit distance method. The recognition experiment conducted by comparing both methods on alphabet and number character images taken from ETL-1 AIST Database. We also did recognition accuracy comparison with another related works in number and alphabet handwritten character recognition. The recognition accuracy of levenshtein distance is better than approximate subgraph matching. It also has competitive performance with another method in same class.