Table 8.
Mean performances of the FCBF-filtered predictive models – N170 amplitude.
| Face | Dataset | EEG channels | Accuracy % | Sensitivity | Specificity | AUC |
|---|---|---|---|---|---|---|
| Fear | Albeniz vs. Mozart | – | – | – | – | – |
| Albeniz vs. Chopin | P3–CPz | 71.65 | 0.70 | 0.74 | 0.77 | |
| Happy | Albeniz vs. Mozart | T8–F4 | 69.27 | 0.64 | 0.74 | 0.72 |
| Albeniz vs. Chopin | – | – | – | – | – | |
| Neuter | Albeniz vs. Mozart | CP1 | 79.13 | 0.69 | 0.8925 | 0.81 |
| Albeniz vs. Chopin | FCz–CPz | 66.38 | 0.73 | 0.60 | 0.67 |
In this amplitude dataset, we can see electrodes more sensible to predictive model. In Fear face, EEG channels are P3 and CPz, In Happy face T8 and F4 and in Neuter Face CP, FCz, and CPz.