Research Article
Online Signature Verification on MOBISIG Finger-Drawn Signature Corpus
Table 12
Verification performance in terms of EER and averaged individual EER
for the DooDB database using anomaly detectors (different feature sets).
| Detector | Skilled forgeries | Random forgeries | EER | aEER (std dev.) | EER | aEER (std dev.) |
| | feat56 | Eucl. | 23.48 | 20.98 (16.05) | 10.32 | 6.76 (10.28) | Manh. | 21.71 | 18.75 (16.66) | 7.82 | 4.19 (7.24) | kNN | 24.25 | 20.83 (16.19) | 10.10 | 7.07 (10.43) | SVM | 23.26 | 20.85 (16.07) | 10.40 | 6.97 (10.39) |
| | feat32 | Eucl. | 24.66 | 22.59 (17.09) | 11.20 | 7.79 (11.00) | Manh. | 22.60 | 19.63 (16.10) | 8.64 | 5.19 (8.20) | kNN | 25.43 | 22.68 (16.81) | 11.34 | 8.21 (10.74) | SVM | 24.60 | 22.43 (17.05) | 11.55 | 7.88 (11.05) |
|
|
Note. Training: 15 samples.
|