Table 6.
EEG-based DDD Accuracies obtained by using Binary Classification Models and Various Features and Ground Truths.
| Ref. No. | Features | Models | Acc (%) |
Sens (%) |
Spec (%) |
GND Truth No. |
|---|---|---|---|---|---|---|
| [72] | SHBP (1~27 Hz) and BPE: #5 | RBF-SVM | 97.48 | - | - | 9 |
| [76] | RPB: α band | Linear-SVM | 95.22 | 100 | 93.8 | 15 |
| [85] | Wavelet: WPT features that are selected by CSP method | SVM | 94.2 | - | - | 8 |
| [44] | BPE: #1, 3, 4, 7 and PBP: δ~β selected by PCA and fish score | SVM | 92.2 | - | - | 12 |
| [42] | Wavelet+: NZC and IEEG extracted from θ~β bands | ANN | - | 90.91 | 79.1 | 2 |
| [77] | FFT+: IEEG, SE and STD extracted from δ~γ bands | SVM | 92.5 | 85 | 100 | 9 |
| [69] | FFT+: DF, APDP, CGF, FV and MPF extracted from δ~β bands | RBF-SVM | 75 | 86 | 64 | 5 |
| [61] | FFT+: RBP-based MCT values | FI | - | 84.6 | 82.1 | 19 |
| [80] | Hybrid: three features from time-domain (Max, Min, STD); ten features from FFT-based methods (CenF, PF, RH/L, Q1F, Q3F, spectral STD, IR, MF, AC and KC); Wavelet-based methods (IEEG and NZC from θ~β bands) | LDA-ANN | - | 83.6 | 87.4 | 9 |
| [35] | PBP: δ~β bands | ANN | 81.49 | 80.53 | 82.44 | 13 |
| [36] | FFT+: SE extracted from SSVEP-based power spectrum | Single-layer feed-forward ANN | 72.5 | - | - | 11 |