Table 1. Results of the model prediction.
| Subject | EEG2Code | Code2EEG | |||
|---|---|---|---|---|---|
| ACC [%] | ITR [bpm] | Pt | r | Pt | |
| S1 | 69.1 | 389.9 | 7.3e−15 | 0.509 | 2.8e−17 |
| S2 | 64.5 | 222.4 | 1.1e−05 | 0.338 | 2.7e−04 |
| S3 | 63.7 | 196.5 | 2.8e−03 | 0.364 | 8.0e−03 |
| S4 | 65.6 | 257.8 | 4.8e−11 | 0.396 | 1.1e−04 |
| S5 | 66.3 | 282.7 | 8.0e−11 | 0.464 | 8.9e−16 |
| S6 | 67.1 | 308.2 | 1.4e−16 | 0.409 | 8.5e−08 |
| S7 | 63.7 | 196.8 | 3.7e−07 | 0.270 | 4.4e−08 |
| S8 | 60.9 | 124.5 | 8.3e−04 | 0.344 | 9.1e−06 |
| S9 | 60.2 | 109.4 | 1.7e−03 | 0.257 | 3.2e−03 |
| mean | 64.6 | 232.0 | 6.0e−04 | 0.384 | 1.3e−03 |
Shown are the average results of all subjects, whereas best results are in bold font. For the EEG2Code model the accuracy (ACC) indicates how many bits of the stimulation pattern can be predicted correctly. The ITR is calculated using Eq 5 with N = 2 and T = 1/60s. Pt are the average p-values under the hypothesis that the correlation between the model prediction and the stimulation pattern are equal to zero. For the Code2EEG model the correlation (r) between the model prediction the measured EEG data is given, as well as the average Pt.