Abstract
Out of the commonly used techniques, fingerprint authentication till date, remains the most reliable. Previously, a plethora of schemes for identification has been employed, however they failed to address a notable challenge- rotational issues, associated with the fingerprint scheme. This leads to incorrect orientation identification, ultimately leading to error in results. Our paper attempts to solve this issue, by proposing a fast pattern matching technique that caters for differences in orientation by firstly, implementing a pre-matching level called the orientation identification stage, and then match the correctly identified oriented fingerprint image to the stored image. To this end, the derived fingerprint image is intercepted with several scan circles to obtain the minutiae information. This information then, is translated into a string, having its staring point as the least lexicographical rotation value. Using approximate string matching techniques, this string information is matched against a database of stored images. The experiment was conducted on solving the rotation stage to prove the efficiency of this method, where the extracting and re-rotation is done in less than a second, with a linear time algorithm, yet practically sub linear in respect to the short extracted binary strings.