Research Article
EEG Signal Classification Using Manifold Learning and Matrix-Variate Gaussian Model
Table 2
Cross-validation performance results for different algorithms in Exp.1.
| Feature extraction | Subj.1 (%) | Subj.2 (%) | Subj.3 (%) | Subj.4 (%) | Subj.5 (%) | Subj.6 (%) | Subj.7 (%) | Subj.8 (%) | Subj.9 (%) | Average (%) |
| None | 84.72 m = 3 | 57.70 m = 1 | 87.23 m = 2 | 54.93 m = 4 | 63.81 m = 2 | 50.71 m = 4 | 88.61 m = 1 | 87.84 m = 4 | 81.05 m = 1 | 72.88 | LDA | 79.22 m = 1 | 51.78 m = 1 | 79.57 m = 1 | 50.38 m = 1 | 63.94 m = 1 | 44.81 m = 1 | 85.91 m = 1 | 75.04 m = 1 | 74.34 m = 1 | 67.22 | | | | | | | | | | 2DLDA | 83.27 m = 1 | 55.19 m = 2 | 78.44 m = 2 | 50.33 m = 1 | 67.32 m = 2 | 44.56 m = 1 | 86.10 m = 1 | 77.43 m = 2 | 74.97 m = 1 | 68.62 | | | | | | | | | | DLPP | 80.55 m = 2 | 55.56 m = 1 | 84.38 m = 1 | 54.17 m = 2 | 63.20 m = 1 | 50.69 m = 2 | 88.19 m = 1 | 80.55 m = 2 | 80.21 m = 2 | 70.83 | | | | | | | | | | 2DDLPP | 84.03 m = 1 | 56.40 m = 1 | 86.60 m = 1 | 52.71 m = 3 | 65.89 m = 2 | 49.66 m = 2 | 88.97 m = 1 | 81.79 m = 2 | 80.98 m = 1 | 71.89 | | | | | | | | | | B2DDLPP | 86.80 m = 2 | 61.11 m = 1 | 89.61 m = 1 | 58.64 m = 2 | 71.91 m = 2 | 53.48 m = 1 | 90.61 m = 1 | 85.38 m = 4 | 83.68 m = 1 | 75.69 | | | | | | | | | |
|
|
For each method and each subject, optimal related to FBCSP’s output and the optimal dimension ( ) are presented. |