Table 2 Comparison of different schemes of linear classifier
| Schemes | Preprocessing | Feature extraction | Feature smoothing | Classification | Emotion states | Accuracy |
|---|---|---|---|---|---|---|
| Method by Li et al. | FT | CSP | liner-SVM | Happiness and sadness | 93.5% | |
| Method by Murugappan et al. | Surface Laplacian filtering Zero mean unit variance | Wavelet transform | KNN and LDA | Disgust, happy, surprise, fear and neutral | KNN: 77.68% LDA: 73.5% | |
| Method by Wang et al. | Wavelet transform, PCA, LDA, CFS | LDS | liner-SVM | Negative and positive | 87.53% | |
| Method by Petrantonakis et al. | Statistical values, wavelet transform and HOC | QDA, KNN, MD and SVMs | Happiness, surprise, anger, fear, disgust and sadness | QDA: 62.3% SVMs: 83.33% MD: 44.90% KNN: 34.60% | ||
| Method by Duan et al. | DE, DASM, RASM and ES | LDS PCA and MRMR | liner-SVM and kNN | Negative and positive | liner-SVM: 74.10% kNN: 69.24% |