Table 1 Feature-based state-of-the-art methods for action recognition.
Method | Data type | Dataset | Performance | References |
|---|---|---|---|---|
Fast Fourier transform | RGB | UCF101, Kinetics | Accuracy: 99.21 | |
QSVM | RGB | UCF11, HMDB51 | Accuracy: 94.43 | |
SVM | RGB | UCSDped-1, UCSDped-2, UMN | Accuracy: 97.14 | |
SVM | RGB | UCF11, UCF50 | Accuracy: 78.6 | |
SVM | RGB | MSRAction3D, UTKinectAction | Accuracy: 94.3 | |
SVM | RGB | Weizmann, KTH, Hollywood2 | Accuracy: 86.3 | |
SVM | RGB | KTH, Weizmann, i3Dpost, Ballet, IXMAS | Accuracy: 95.5 | |
SVM | RGB | KTH, UCFSports, Hollywood2 | Accuracy: 91.8 | |
SVM with ASAGA | RGB | UCSDped 1 | Accuracy: 87.2 | |
SVM with PSO | Skeleton | MSRAction3D, UT Kinect, Florence3D action | Accuracy: 93.75 | |
SVM with GA | RGB | KTH, HMDB51, UCF YouTube, Hollywood2 | Accuracy: 95.0 | |
SVM-neural network | RGB | KTH, Weizmann | Average Accuracy: 96.4 | |
RF | Skeleton | UT Kinect | Accuracy: 92 | |
NBNN | 3D joints skeleton | MSRAction3D-Test1, MSRAction3D-Test2, MSRAction3D-cross-subject | Accuracy: 95.8 | |
HMM-Kernel Discriminant analysis | Silhouette | Elder care data | Accuracy: 95.8 | |
HMM | Skeleton | Im-DailyDepthActivity, MSRAction3D (CS), MSRDailyActivity3D (CS) | Accuracy: 74.23 |
