Automatic estimation of the readability of handwritten text

In this paper, the problem of estimating the readability of handwritten text is addressed. The estimation problem is posed as a two class classification problem where a text is classified as either readable or unreadable. A classifier is trained on this two class classification problem. In the training phase, for each text a number of features are extracted. At the same time the recognition rate achieved on the text is determined. Based on the recognition rate, each feature vector is labelled, i.e., assigned to one of the two classes. The labelled data is then used to train a classifier. The k-Nearest Neighbour (k-NN) and the Support Vector Machine (SVM) classifier are evaluated in this work. Both classifiers show promising results on a test set of 715 text lines from 20 writers.