Table 11.
The classification performance of different methods for the BCI competition IV dataset 2b.
| Method | Researcher | Classifier | Mean Kappa Value |
|---|---|---|---|
| CSCNN * | [63] | CNN | 0.663 |
| CNN-VAE | [64] | VAE | 0.603 |
| NCA * + DTCWT * | [65] | SVM | 0.615 |
| FBCSP * | [62] | NBPW * | 0.599 |
| RES * + FBCSP | [66] | SVM | 0.643 |
| HCRF * | [67] | HCRF | 0.622 |
| CNN-DCGAN | Our method | CNN | 0.671 |
* Some acronyms are defined in the following: Continuous Small Convolutional Neural Network: CSCNN; Neighbourhood Component Analysis: NCA; Dual-Tree Complex Wavelet Transform: DTCWT; Restricted Exhaustive Search: RES; Filter Bank Common Spatial Pattern: FBCSP; Naïve Bayesian Parzen Window: NBPW; Hidden Conditional Random Fields: HCRF.