Penmanship Counts

Back when we launched Dr. Dobb's Handwriting Recognition Contest, the term "biometric recognition" was rarely--if ever--used, and "security" had little to do with handwriting recognition.

These days, biometric recognition and security go, well, hand-in-hand and research into ways to make handwriting recognition more accurate, faster, and secure continues to move forward.

For instance, Juan Jose Igarza Ugaldea of the University of the Basque Country has proposed two approaches to handwritten signature biometric identification--on-line and off-line signature recognition.

At the core of Ugaldea's system is, of course, a database of signatures from anonymous signers and skilled forgerers. (The database is multimodal in that it also includes fingerprints and voices from all users.)

Ugaldea's proposal for on-line signature recognition is based on processing local features (coordinates, velocities, accelerations, pressure, and pen tilt) by using global features (center of mass of ink and principal inertia axes) as a reference system. He then uses a simple scaling algorithm for signature time normalization that is applicable to every local feature. He has limited the amount of local features to feed the Hidden Markov Models (HMM) to nine to balance the required security level and the processing and storing capacities. Using six stated Left-to-Right (LR)-HMM, Ugaldea claims to have obtained error rates equal to those currently considered state of the art.

His proposal for off-line signature recognition consists of two systems based on the LR-HMM technique. The models are fed on spatially ordered point sequences and geometrical "false-dynamic" derivative features. The goal of those proposals is to extend LR-HMMs to the field of static or off-line signature processing using results provided by image connectivity analysis, which separates images in connected components known as "blobs", each one made up of a cluster of adjacent pixels of the same nature.

Ugaldea has proposed two different methods of generating models, depending on how the blobs provided by the connectivity analysis are ordered. In the first proposed method, blobs are ordered according to their perimeter length. In the second proposal, the sorting criterion is based in the natural reading order. The models based on the second criterion are more than adequate for the Latin writing in which signatures have been written.

Your's truly,

Jonathan Erickson

Posted by Jon Erickson at 10:29 AM  Permalink