Table 11 Comparison of the proposed approach against state-of-the-art studies evaluated on the Kumar imagined speech dataset across single categories.
From: EEG imagined speech neuro-signal preprocessing and deep learning classification
Research | Architecture | Char (%) | Digit (%) | Obj (%) | Year |
|---|---|---|---|---|---|
Kumar et al.26 | Random Forest | 66.90 | 68.50 | 65.70 | 2018 |
Tirupattur et al.38 | CNN | 71.20 | 72.90 | 73.00 | 2018 |
Ignazio et al.24 | CNN/transformers | 97.30 | 97.20 | 96.60 | 2024 |
Kumar et al.39 | CNN/LSTM | 87.30 | 85.90 | 87.50 | 2022 |
Proposed system | CNN-2-LSTM | 99.14 | 99.05 | 99.21 | 2025 |
CNN-2-Bi-LSTM | 99.40 | 99.17 | 99.29 | ||
CNN-3-LSTM | 99.20 | 99.07 | 99.20 | ||
3-LSTM | 99.06 | 98.80 | 99.28 |
