Feature Evaluation for Handwriting: a ranking-based approach

Feature selection is generally considered a very important step in any pattern recognition process. Its aim is that of reducing the computational cost of the classification task, trying to increase, or not to reduce, the performance of the whole recognition system. In the framework of handwriting recognition, the large variability exhibited by the samples produced by different writers makes even more complex the selection of appropriate feature sets, and has led to the development of a wide research activity, which has produced many interesting results published in the literature. Nonetheless, this problem is very far to be solved in the general case, since the above results present limitations at different levels. The main drawbacks include the complexity, the dependence on the adopted classifiers and the difficulty in evaluating the interactions among features. In this study, we have tried to overcome some of the above drawbacks by adopting a feature ranking based technique: we have considered different univariate measures to produce a feature ranking and we have proposed a greedy search approach for choosing the feature subset able to maximize the classification results. In our work, we have chosen one of the most widely used feature set in handwriting recognition, while the experimental results have been obtained by using three real word datasets of handwritten characters.