Table 6.
Comparison among existing methods in the Facial Emotion Recognition.
| References | Methods | Dataset | Accuracy |
|---|---|---|---|
| Zhou et al.49 | CNN + MVFE-LightNet | FER2013 | 68.4% |
| Ziyang Yu et al.50 | CNN + music algorithm | FER2013 | 62.1% |
| P. Ferandez et al.51 | FERAtt | CK+ | 82.11% |
| N. Christou and N. Kanojiya23 | CNN | FER2013 | 91% |
| F. Wang et al.29 | EFDMs + EEG + CNN | EEG data on SEED | 90.59% |
| DEAP | 82.84% | ||
| F. Nonis et al.31 | 3d approaches | BU-3DFE | 60% to 90% |
| Ben Niu et al.52 | SVM + LBP + ORB | JAFFE | 88.5% |
| CK+ | 93.2% | ||
| MMI | 79.8% | ||
| Ji-Hae Kim et al.53 | LBP + deep neural network | CK+ | 96.46% |
| JAFFE | 91.27% | ||
| Sawardekar and Naik54 | LBP + CNN | CK+ | 90% |
| Fei Wang et al.28 | EEE based EFDMs | Cross datasets | 82.84% |
| Hongli Zhang et al.9 | CNN + image edge computing | FER2013 + LFW | 88.56% |
| Ke Shan et al.55 | KNN + CNN | JAFFE | 76.74% |
| CK+ | 80.30% | ||
| Pham and Quang56 | CNN + FPGA | FER2013 | 66% |
| Guohang Zeng et al.57 | Deep learning + handcrafted feature | CK+ | 97.35% |
| Shan Li and Deng58 | Deep learning | All facial dataset | 45% to 95% |
| Zuheng Ming et al.59 | FaceLiveNet | FER2013 | 68.60% |
| CK + | 98% | ||
| Hussain and Balushi35 | Deep learning | KDEF | 88% |
| Smitha Rao et al.32 | CNN + LSTM | CREMA-D | 78.52% |
| RAVDEES | 63.35% | ||
| Khalid Bhatti et al.60 | Deep features + extreme learning | JAFFE | 92.4% |
| CK | 91.4% | ||
| FER2013 | 62.7% | ||
| Proposed method | Fusion features (CNN + LBP + ORB) + ConvNet | FER2013 | 91.01% |
| JAFFE | 92.05% | ||
| CK + | 98.13% |