Table 4. Summary of other machine learning feature extraction and feature classification algorithms.
| A | M | P | D | S | DE | E |
|---|---|---|---|---|---|---|
| Idrees & Farooq (2016a) | TDC | /a/, /u/, other | Private data | 3 (2 m and 1 f) | BioSemi ActiveTwo | Combination average: 85–100 |
| LC | ||||||
| Idrees & Farooq (2016b) | WD | /a/, /u/, “no” | Private data | 3 (2 m and 1 f) | BioSemi ActiveTwo | Combination average: 81.25–98.75 |
| LC | ||||||
| Moattari, Parnianpour & Moradi (2017) | HON-ICA | /a/, /u:/ | Rostami & Moradi (2015) | 5 (3 m and 2 f/age from 23–30) | N/A | 66.67–93.33 |
| Nguyen, Karavas & Artemiadis (2017) | CM | “/a/,/i/,/u/” “in, out, up” |
Private data | 15 (11 m and 4 f/age from 22–32) | BrainProducts ActiCHamp amplifier | Highest: 95 (Binary classification), 70 (Three categories) |
| RVM | “cooperate, independent” | |||||
| Nguyen, Karavas & Artemiadis (2019) | SCM | Long word: “concentrate” | Private data | 8 (6 m and 2 f/age from 22–32) | BrainProducts ActiCHamp amplifier | Average: 52.5 |
| RVM | Short word: “split” | |||||
| Kim, Lee & Lee (2020) | ERP | “Ah”, Specific nouns | Private data | 2 (2 f/age from 22–27) | ActiCap EEG amplifier | Highest combination: 88.1 |
| RLDA | ||||||
| Wang et al. (2020) | PSD, SampEn ELM |
Chinese character “移(move)” | Private data | 12 (8 m and 4 f/age from 20–26) | SynAmps 2 | Average: 83 |
| Pan et al. (2023) | WPD | Chinese character “左(Left), 壹(One)” | Wang et al. (2021) | 8 (6 m and 2 f/age from 22–27) | SynAmps 2 | Average: 90 |
| LightGBM |
Note:
A, authors; M, methods; P, pronunciation materials; D, datasets; S, subjects (number); DE, device; E, the evaluation indicators (accuracy: %); m, males; f, females; TDC, time domain characteristics; LC, linear classifier; WD, wavelet decomposition; HON-ICA, higher orders of Non-Gaussianity independent component analysis; CM, covariance matrix; SCM, spatial covariance matrix; RVM, relevance vector machines; ERP, event-related potential; RLDA, regularized linear discriminant analysis; PSD, power spectral density; SampEn, sample entropy; ELM, Extreme Learning Machine; WPD, wavelet packet decomposition; LightGBM, light gradient boosting machine.