Table 6.
Imagined speech classification methods summary (continuation).
| References | Task | Methods | Brain area and waves | Performance |
|---|---|---|---|---|
| Cooney et al. (2020) | - Classification of /a/, /e/, /i/, /o/, and /u/ | - CNN | - All areas | - Vowels: 35% best accuracy |
| - Classification of “left,” “right,” “up,” “down,” “forward,” “backward” (in Spanish) | - ICA/LDA | −2–40 Hz bandpass | - Words: 30% best accuracy | |
| Tamm et al. (2020) | Classification of five vowels and six words | - CNN | - F3, F4, C3, C4, P3, P4 | −24% Average accuracy |
| - Transfer learning | ||||
| Chengaiyan et al. (2020) | Vowel classification | - RNN | - All areas | - RNN: 72% Average accuracy |
| - DBN | −2–40 Hz bandpass | - DBN: 80% Average accuracy | ||
| Jiménez-Guarneros and Gómez-Gil (2021) | Short and long words classification | - SRDA | - All areas | - Short: 61% Average accuracy |
| - Long: 63% Average accuracy |