04-06-2012, 12:04 PM
Iris Recognition Using AdaBoost and Levenshtein Distances
Iris Recognition Using AdaBoost and Levenshtein Distances.pdf (Size: 517.8 KB / Downloads: 44)
Introduction
Biometric systems are becoming popular methods for personal identification. Each
biometric technology has its set of advantages, considering their usability and se-
curity. The human iris, located between the pupil and the sclera, has a complex
pattern determined by the chaotic morphogenetic processes during embryonic devel-
opment. The iris pattern is unique to each person and to each eye, and is essentially
stable during an entire lifespan. Furthermore, an iris image is typically captured
using a non-contact imaging device, of great importance in practical applications.
These reasons make iris recognition a robust technique for personal identification33.
The first automatic iris recognition system was developed by Daugman10.
Iris image encoding
In general, iris recognition systems are composed of five stages: acquisition, lo-
calisation, normalisation, encoding, and identification. Figure 1 shows the results
obtained after the four initial stages.
Prior to obtaining the IrisCode, the pupil must be located and the iris seg-
mented. We employ a standard technique to segment the iris34. The iris can be
located at the region between two concentric circles, one for the iris-sclera bound-
ary and another for the iris-pupil boundary, as shown in Figure 2.b.
The Edit Distance
A novel approach for comparing IrisCodes, which uses Levenshtein distance, is
presented in this paper. The Levenshtein distance5, also called edit distance, is
employed for measuring the difference between two strings. The distance is given as
the minimum number of operations needed to transform one string into the other,
where an ‘operation’ is an insertion, deletion, or substitution of a single character.
This metric is useful in a wide range of applications, and there is a large body of
work concerning string comparison using Levenshtein distance in recent literature
– see, for example, Ref. 25. The usual way to compute the Levenshtein distance is
with an (m+1)×(n+1) cost matrix L, where m and n are the lengths of the two
strings.
Iris classification using AdaBoost
Boosting is a meta-algorithm for automatic learning that builds a robust classifier
by a combination of a set of weak classifiers. A classifier is considered weak if it
has a correct classification ratio slightly better than chance.