Table 4 correct frames % for different languages (columns) recognised by different models (rows). The languages are: German (de), English (en), Flemish (fl) and Swedish (sv). Numbers in parentheses are the % of correct frames for perfect recognition, given the mismatch in phonetic inventory across languages.
From: SynFace—Speech-Driven Facial Animation for Virtual Speech-Reading Support
de | sv | fl | en | |
|---|---|---|---|---|
de | 61.0 (100) | 30.3 (82.6) | 27.1 (73.5) | 26.2 (71.6) |
sv | 31.5 (86.2) | 54.2 (100) | 26.3 (72.1) | 23.5 (74.2) |
fl | 34.2 (85.7) | 31.6 (77.9) | 51.0 (100) | 26.9 (69.8) |
en | 24.5 (74.6) | 23.7 (72.3) | 21.5 (66.8) | 46.1 (100) |