Frequent pattern mining for online handwriting recognition

Abstract

In this paper, we address the problem of defining and modeling the handwriting signal using its geometrical and spatio-temporal features, in order to improve the recognition task. We use the frequent pattern methods to enhance the quality of the signature vector extracted from the handwritten character. Two types of frequent patterns are employed to represent the handwritten characters pertinently: the maximal and closed frequent patterns. We created a new database that contains words of two different letters. The generated results are very promising, through which we have demonstrated that the “minimum threshold”, which is an essential parameter in the frequent patterns mining algorithms, represent a key feature in the characters description.