Research Article
Investigating Feature Ranking Methods for Sub-Band and Relative Power Features in Motor Imagery Task Classification
Table 2
Analysis of accuracy (in %) for impact of relative and varied power features.
| Subjects | Accuracy with all features | Accuracy without relative power features | Accuracy with best features from mutual information method | Accuracy with best features from chi-square method | Accuracy with dropping out correlated features |
| S01 | 94.3 | 94.9 | 93.8 | 94.2 | 92.4 | S02 | 92.8 | 93 | 93.7 | 93.1 | 92.5 | S03 | 94.2 | 93.3 | 93.2 | 94.2 | 90.4 | S04 | 93.6 | 91.9 | 92.1 | 93.6 | 91.7 | S05 | 93.1 | 92.5 | 92.6 | 93.1 | 91.8 | S06 | 93.1 | 92.7 | 92.3 | 93.1 | 90.1 | S07 | 91.6 | 91.4 | 91.7 | 91.6 | 89.7 | S08 | 92.9 | 92.6 | 92.2 | 93.01 | 90.6 | S09 | 95.2 | 95.4 | 95.4 | 95.2 | 91.7 | S10 | 94.3 | 94.9 | 93.8 | 94.3 | 92.1 |
|
|
Bold letters show the maximum classification accuracy of the classifier.
|