Table 1 Comparisons of recognition accuracies on KTH dataset
Method | Brief description | Acc. % |
|---|---|---|
ICPR’04 [22] | HOGHOF + SVM | 71.7 |
CVPR’07 [36] | cuboid + WX-SVM | 91.6 |
BMVC’09 [35] | cuboid + BoW+ χ2 | 89.1 |
|  | HOGHOF + BoW + χ2 | 91.8 |
ECCV’10 [37] | convolutional nets | 90.0 |
In this paper | cuboid + BoW + χ2 (baseline) | 88.2 |
| Â | cuboid + AMK I U c S=2 | 91.7 |
| Â | cuboid + AMK I U b S=8 | 93.0 |
| Â | cuboid + AMK I B S=7 | 92.7 |
| Â | cuboid + AMK II | 93.2 |
|  | HOGHOF + BoW + χ2 (baseline) | 88.5 |
| Â | HOGHOF + AMK I U c S=3 | 91.1 |
| Â | HOGHOF + AMK I U b S=27 | 95.2 |
| Â | HOGHOF + AMK I B S=5 | 92.3 |
| Â | HOGHOF + AMK II | 93.5 |