Table 9 Subject Performance Distribution of proposed CNN-2-Bi-LSTM utilizing proposed preprocessing method FD-F versus NetTraST for the 30-class CharDigObj classification under LOSO cross-validation strategy.
From: EEG imagined speech neuro-signal preprocessing and deep learning classification
Statistic | CNN-2-Bi-LSTM | NetTraST | Difference |
|---|---|---|---|
Mean | 78.86% | 28.41% | + 50.45% |
Median | 80.59% | 28.17% | + 52.42% |
Min | 56.40% (Subj 14) | 11.87% (Subj 6) | + 44.53% |
Max | 91.66% (Subj 8) | 43.64% (Subj 8) | + 48.02% |
Range | 35.26% | 31.77% | – |
Std dev | 9.84% | 10.18% | – |
CV | 13.05% | 33.56% | – |
