Table 2.
Selected feature extraction methods, giving year, the feature extraction method, how speech classification was performed (if used), the database used for classification, and the task carried out, as well as the reported recognition rate
| Year | Reference | Feat. Extract. | Classif. | Database | Task | Performance |
|---|---|---|---|---|---|---|
| 1988 | Kass et al. [32] | ACM | ||||
| 1998 | Cootes et al. [8] | AAM | ||||
| 2008 | Shao and Barker [33] | DCT | HMM | GRID | Phrases | 58.40 |
| 2008 | Seymour et al. [34] | DCT | HMM | XM2VTS | Digits | 87.89 |
| PCA | HMM | XM2VTS | Digits | 86.57 | ||
| LDA | HMM | XM2VTS | Digits | 86.35 | ||
| 2009 | Zhao et al. [35] | LBP-TOP | SVM | AVLetters | Alphabet | 62.80 |
| LBP-TOP | SVM | OuluVS | Phrases | 62.40 | ||
| 2009 | Lan et al. [36] | AAM | HMM | GRID | Phrases | 65.00 |
| 2009 | Dakin et al. [37] | GWT | ||||
| 2011 | Hursig et al. [38] | GWT | ||||
| 2011 | Cappelletta and Harte [39] | Optical flow | HMM | VIDTIMIT | Sentences | 57.00 |
| PCA | HMM | VIDTIMIT | Sentences | 60.10 | ||
| 2016 | Lee et al. [40] | DCT + PCA | HMM | OuluVS2 | Phrases | 63.00 |