Table 5.
LOSOCV accuracy in comparison with MV-CNN on five databases.
| ML model | Type of ML model | Domain adaptation method | Database | Num. of gestures for classification | LOSOCV accuracy (achieved with 200 ms window) |
|
| |||||
| MV-CNN [31] | Multiview CNN | MS-AdaBN | NinaProDB1 | 52 | 84.3% |
| HVPN-2-view | Multiview CNN | MS-AdaBN | NinaProDB1 | 52 | 84.5% |
| HVPN | Multiview CNN | MS-AdaBN | NinaProDB1 | 52 | 84.9% |
|
| |||||
| MV-CNN [31] | Multiview CNN | MS-AdaBN | NinaProDB2 | 50 | 80.1% |
| HVPN-2-view | Multiview CNN | MS-AdaBN | NinaProDB2 | 50 | 81.8% |
| HVPN | Multiview CNN | MS-AdaBN | NinaProDB2 | 50 | 82.0% |
|
| |||||
| MV-CNN [31] | Multiview CNN | MS-AdaBN | NinaProDB3 | 50 | 55.5% |
| HVPN-2-view | Multiview CNN | MS-AdaBN | NinaProDB3 | 50 | 65.4% |
| HVPN | Multiview CNN | MS-AdaBN | NinaProDB3 | 50 | 65.6% |
|
| |||||
| MV-CNN [31] | Multiview CNN | MS-AdaBN | NinaProDB4 | 53 | 52.6% |
| HVPN-2-view | Multiview CNN | MS-AdaBN | NinaProDB4 | 53 | 69.9% |
| HVPN | Multiview CNN | MS-AdaBN | NinaProDB4 | 53 | 70.2% |
|
| |||||
| MV-CNN [31] | Multiview CNN | MS-AdaBN | NinaProDB5 | 41 | 87.2% |
| HVPN-2-view | Multiview CNN | MS-AdaBN | NinaProDB5 | 41 | 88.8% |
| HVPN | Multiview CNN | MS-AdaBN | NinaProDB5 | 41 | 88.9% |
N.A. denotes not applicable, and bold entries indicate our proposed method. HVPN-2-view refers to the proposed HVPN framework with the “two-view” configuration (i.e., using v1 and v2 as its input).