|
| Training stage with 80% of NIST’s data | |||||||
|
| Classic method | Probabilist method | Mean classes LCNN | 1-PPV LCNN | K-PPV LCNN (K = 8) | Proposed method | ||
| Training time | No | 46 min | 28 min | 46 min | 48 min | 58 min | ||
| Recognition time | 41 s | 5 s | 4 s | 5 s | 9 s | 11 s | ||
| AR (en %) | 91.5 | 94.3 | 49.5 | 88.0 | 92.4 | 94.4 | ||
| FRR (en %) | 8.0 | 5.1 | 17.6 | 11.6 | 7.0 | 5.1 | ||
| FAR (en %) | 0.5 | 0.6 | 32.9 | 0.4 | 0.6 | 0.5 | ||
|
| Training stage with 80% of BDAL’s data | |||||||
|
| Classic method | Probabilist method | Mean classes LCNN | 1-PPV LCNN | K-PPV LCNN (K = 8) | Proposed method | ||
| Training time | No | 15 min | 9 min | 15 min | 16 min | 19 min | ||
| Recognition time | 27 s | 3 s | 2 s | 3 s | 7 s | 9 s | ||
| AR (en %) | 91.1 | 93.9 | 49.1 | 87.6 | 92.0 | 94.0 | ||
| FRR (en %) | 8.3 | 5.4 | 17.9 | 11.9 | 7.3 | 5.4 | ||
| FAR (en %) | 0.6 | 0.7 | 33.0 | 0.5 | 0.7 | 0.6 | ||