Determining Cursive or Printed Nature of Handwritten Samples

Forensic document examiners tend to compare documents to one another only when they are both cursive or both hand printed documents. Thus, for an automated system to process writer identification or verification, it is desirable to determine whether a sample of handwritten text is of a hand printed or cursive nature. We identify a feature based upon the number of segmented words and number of connected components which are identified as single characters. We evaluate the performance of this feature by using all 4,701 documents present CEDAR letter dataset and find that using this feature identifies documents as hand printed correctly 95% of the time and identifies them as cursive 91% of the time.