TABLE II.
Classification results (AR) in % and number of selected features (in parentheses) of individual datasets using leave-one-out cross-validation
| Modality/Dataset | BlackDog | AVEC | Pitt | |||
|---|---|---|---|---|---|---|
| Feature Selection | Variable set of features | |||||
|
| ||||||
| Eye Activity | 80.0 | (13) | 81.3 | (31) | 92.1 | (20) |
| Head Pose | 73.3 | (7) | 68.8 | (29) | 86.8 | (31) |
| Feature fusion | 85.0 | (20) | 75.0 | (60) | 92.1 | (51) |
|
| ||||||
| Hybrid fusion | 85.0 | 75.0 | 92.1 | |||
|
| ||||||
| Feature Selection | Fixed set of features | |||||
|
| ||||||
| Eye Activity | 73.3 | (7) | 71.9 | (7) | 89.5 | (7) |
| Head Pose | 61.7 | (5) | 71.9 | (5) | 84.2 | (5) |
| Feature fusion | 78.3 | (12) | 65.6 | (12) | 92.1 | (12) |
|
| ||||||
| Hybrid fusion | 76.7 | 68.8 | 94.7 | |||