Table 3 Comparison of different schemes of nonlinear classifiers
Schemes | Preprocessing | Feature extraction | Feature smoothing | Classification | Emotion states | Accuracy |
|---|---|---|---|---|---|---|
Method by Liu et al. | Â | Â | Â | FD | Sad, frustrated, fear, satisfied, pleasant and happy | Â |
Method by Liu et al. | Â | ResNets, LFCC | Â | KNN, SVM, LR, RF, NB, DT and FC | Anger, joy, sadness and pleasure | KNN: 89.72% |
Method by Zheng et al. | Â | DE, DASM, RASM | Â | DBN, SVM, LR and KNN | Positive, neutral and negative | DBN: 86.08% SVM: 83.99% LR: 82.70% KNN: 72.60% |
Method by Dan Nie et al. | Â | FFT | LDS | SVM | Negative and positive | SVM: 87.53 |
Method by Zheng et al. | Â | PSD, DE, DASM, RASM, ASM and DCAU | MRMR | KNN, LR, SVM and GELM | Negative, positive and neutral | KNN: 70.43% LR: 84.08% SVM: 78.21 GELM: 91.07% |