Studies of Fingerprint Template Selection and Update - Advances in Computational Science and Engineering - page 156

Table 2. Greedy Maximum Match Scores (GMMS)

S

1 Initialize N, K, NN , Choose[K],

×

i ← 1

2

for to K do

3 begin

N

∑

s
=≠

j = 1, N

K

4 comput sum(j)=

jm ,

mmj 1,

j * j * , ∈ N

K

5 find ( )

j * ≥ sum j () j = 1, N

K

6 such that sum( )

Choose i [] ← j *

7

s ←←= 0, s 0, m 1, N

K

8 * *

jm , mj ,

9 end

5 Score Based Template Update

5.1 Introduction to Template Update

During enrollment, N samples are captured at the same time. However, to make system
user-friendly and save time, N is small usually. Also, it may be not a good strategy to
capture too many samples during one time because the samples will not vary of wide
scope. Biometric trait varies along with age and suddenness. For example, older people
will have more wrinkles than young people and fingerprints from older people may have
more cut than young people. Therefore, template update is very important in biometric
recognition system [3,12,13]. Umut Uludag [3] proposed two strategies BATCH-
UPDATE and AUGMENT-UPDATE for template update. And in his paper [3], the
experimental results show that AUGMENT-UPDATE gains better performance than
BATCH-UPDATE. This is because biometric samples are varying gradually not
abruptly. Also feature level fusion [2,5,7] and image level fusion [6,8] are used in tem-
plate update and these two strategies are not very different with template selection. In our
paper, template update means to update the K templates with the successively accept
samples a i during authentication. In order to conduct template update in authentication,
biometric system need to store the match scores among the K templates. As we assume
that the biometric database just store the feature exacted from raw data, we cannot get the
match score between template raw data i k and the successively accept sample a i during
authentication. Therefore, we don't distinguish between s m,n and s n, m in template update.

5.2 Two Strategies of Template Update

The strategies of template update can be divided into two classes. One is ONLINE
update which activates the update procedure whenever a new biometric sample is
accepted. The other is OFFLINE update which activates the update procedure during
a fixed period. And we just call the two methods: ONLINE and OFFLINE. On the
one hand, ONLINE strategy need no more space for update procedure and can update