Information Technology Reference
In-Depth Information
discriminant distances are normalized, i.e. each QEPD divides by the maximum
of QEPD. Likewise the peak of palmprint outer-class approaches 0.24 and 0.34
by finger. The two peaks can be separated widely, hence scheme can effectively
discriminate these handmetric images.
3.3 CSID Computation
According to CSID discussed previous section, discriminant distance tables cor-
responding to figure 1 and 2 are shown in table 3 and 4.
Totally, about 160,000 (400
400) matching of palmprint and finger texture
have been performed respectively, where 120,000 matching are outer-class CSID
for each. Figure 3(c) and (d) show the distribution of inner-class and outer-class
CSID. From them, peak of inner-class CSID of palmprint and finger texture are
at about 0.09 and 0.14. And the peak of outer-class approach 0.21 and 0.32. The
two peaks can be separated widely, so the scheme discriminate them like QEPD.
×
3.4 Comparison of Result
Firstly, we make a comparison among four distribution graphes. From the fol-
lowing table, it is found that as to distance between inner-class and outer-class
peak, i.e. difference in the table 5, finger QEPD is widen than palmprint's, and
the same is true that CSID. This means QEPD may discriminant effectively than
CSID. Interestingly, regardless of QEPD or CSID, finger's is large than that of
palmprint. Here we provide an inference that this finger texture proposed is
an effective biometric. Then we will use ROC performance for testifying this
inference.
Table 3.
Palmprint CSID computation among figure 1
Fig 1(b) 1(c) 1(d) 1(e) 1(f)
1(a)0.523 0.445 0.541 0.592 0.479
1(b)
0.518 0.692 0.778 0.640
1(c)
0.564 0.504 0.534
1(d)
0.416 0.113
1(e)
0.446
Table 4.
Finger texture CSID computation among figure 2
Fig 2(b) 2(c) 2(d) 2(e) 2(f) 2(g) 2(h) 2(i)
2(a)0.224 0.147 0.386 0.196 0.258 0.408 0.227 0.393
2(b)
0.173 0.318 0.274 0.300 0.383 0.252 0.416
2(c)
0.289 0.259 0.349 0.409 0.298 0.375
2(d)
0.406 0.532 0.419 0.475 0.387
2(e)
0.186 0.394 0.151 0.399
2(f)
0.354 0.072 0.496
2(g)
0.317 0.451
2(h)
0.475
Search WWH ::




Custom Search