In this paper, the automatic recognition of broken and blurred, multifont typewritten digits in forms will be addressed. The classification, which is based on the utilization of a global feature, is divided in two phases: first, a minimum distance method (1-NN) is applied to provide a global classification of the patterns in a form; second, the patterns in the form previously classified are used to validate, or reject and reclassify them, on the basis of the mean distance to the predefined classes. In this way, a classification accuracy rate of 99.42% has been achieved.