Skip to main content

Table 1 A breakdown of the 1428 marks into the categories of distortion (including no deliberate distortion), and the percentage of marks found to be EV by each of three experts, with decision on ground truth made by majority vote [3]

From: Identification performance of evidential value estimation for ridge-based biometrics

 

Number of

Prints of sufficient evidential value

 

Type of distortion

marks taken

Assessor 1 (%)

Assessor 2 (%)

Assessor 3 (%)

Ground truth (%)

EVA (%)

(i) Light placement

168

48.2

48.2

48.2

48.2

54.2

(ii) Smeared

168

3.6

4.2

3.6

3.6

14.9

(iii) Finger twisted lightly

168

4.2

4.8

4.8

4.8

11.3

(iv) Strong twist

168

0.0

0.0

0.0

0.0

6.0

(v) Heavy placement

168

69.6

65.5

65.5

65.5

64.9

(vi) Partial, heavy placement

168

45.8

48.2

48.2

48.2

50.6

(vii) Normal

420

47.4

49.0

50.0

49.0

50.7

Total

1428

34.1

34.5

34.7

34.5

38.66

  1. The EV distribution for EVA has been calculated for the mobile phone images scaled using CRE Global and the Fusion feature set at the decision threshold corresponding to the EER