Write This Down: Handwriting Biometrics for Personal Identification

NEW DELHI, India, Jan 6 -- A team of computer scientists at the Indian Institute of Technology (IIT-K) is working hard on biometrics systems for ascertaining the foolproof identity of a person.

Pleading anonymity, one of the senior professor scientists told Hindustan Times that automatic handwriting recognition, which dealt with computer measurement of individual handwriting features to identify and verify the writer, was of immense importance and could prove infallible in its results. He said verification work was mostly done by human experts, but their results were often challenged in courts for not being sufficiently scientific. However, during the past one decade some new algorithms were proposed for automatic processing and quantification of handwriting.

He felt for successful identification one had to look for some features of handwriting styles which included arrangement of lay out, class of allographs used, connection of characters, design of allographs, size of characters, spacing, abbreviations alignment, commencement and termination, diacritics and punctuation, pen control and many other such minor things.

The new image processing techniques developed so far were based on simultaneously matching the multiple and distinctive 2-D image patterns on finger from nail side of the hand for biometric verifications. He said a prototype system to test the technique had already been developed and it consisted of an imaging unit for guided placement of the hand.

However, he further said the Robust Face Recognition systems had potential applications in the area of surveillance and physical security and a new technique for Robust extraction of biometric features from frontal view of the test face was attracting scientists all over the globe.

He said two major techniques of face-based personal identification developed at the Bhaba Atomic Research Centre could help future scientists to create definite identification systems through this technique. The first one employed Partitioned Iterated Function System paradigm by treating human face as the fractal object. The other technique carried out personal identification by template matching in terms of a set of correlation scores corresponding to different areas of interest. The major problem of facial size normalization to counter the variation of distance of the subject from the video camera occurring during image acquisition was yet to be solved, he said.

The Hindustan Times is provided through HT Syndication, New Delhi.