Research Article
Improved Deep Feature Learning by Synchronization Measurements for Multi-Channel EEG Emotion Recognition
Table 4
Details from comparison studies.
| Studies | Features | Classifiers | Evaluation methods | Accuracy | Arousal | Valence |
| Koelstra et al. [42] | CNS feature-based single modality | Gaussian Naive Bayes | Leave-one-trial-out validation | 0.6200 | 0.5760 | Chen et al. [46] | Time and frequency-domain features | Decision trees | 10-fold cross-validation | 0.6909 | 0.6789 | Xu and Plataniotis [47] | Narrow-band spectral features | DBN | Leave-one-subject-out validation | 0.6988 | 0.6688 | Zhuang et al. [48] | The first difference in time series, etc. | SVM | Leave-one-trial-out validation for each subject | 0.7199 | 0.6910 | Huang et al. [49] | Autoregressive coefficients features | SVM | 10-fold cross-validation | 0.6730 | 0.6430 | Li et al. [50] | Linear and nonlinear features | SVM | Leave-one-subject-out validation | - | 0.5906 | Chen at al [51] | Raw EEG features | H-ATT-BGRU | Leave-one-subject-out validation | 0.6650 | 0.6790 | The proposed method | MIC-based features | SVM | 10-fold cross-validation | 0.7185 | 0.7021 |
|
|