Research Article
Using a Multilearner to Fuse Multimodal Features for Human Action Recognition
Table 3
Comparison of various action recognition methods.
| Method | Descriptor | Recognition Methods | Accuracy | G3D (%) | CAD60 (%) |
| Dollar’s method [38] | Sparse spatiotemporal features | SVM | 78 | 83 | Liu’s method [42] | PMI spatiotemporal features | SVM | 82 | 86 | Laptev’s method [43] | Spatiotemporal corner | SVM | 87 | 84 | Rapantzikos’s method [44] | Dense saliency spatiotemporal features | KNN | 88 | 89 | Rodriguez’s method [45] | Spatiotemporal template | Template matching | 88 | 89 | Dropout-based CNN [46] | — | Random dropout-based CNN | 89 | 90 | Our Approach | Mixed features | SE-SVM | 93 | 94 |
|
|