Research Article
Indonesian Lip-Reading Detection and Recognition Based on Lip Shape Using Face Mesh and Long-Term Recurrent Convolutional Network
Table 4
Results of the performance evaluation (precision, recall, and F1 score) on the word dataset.
| Words | Conv-LSTM | LRCN-2Conv | LRCN-3Conv | Precision (%) | Recall (%) | F1 score (%) | Precision (%) | Recall (%) | F1 score (%) | Precision (%) | Recall (%) | F1 score (%) |
| Maaf | 96 | 96 | 96 | 100 | 100 | 100 | 100 | 100 | 100 | Tolong | 85 | 96 | 90 | 88 | 96 | 92 | 96 | 96 | 96 | Permisi | 96 | 96 | 96 | 92 | 96 | 94 | 100 | 96 | 98 | Halo | 91 | 88 | 89 | 96 | 92 | 94 | 91 | 88 | 89 | Mulai | 89 | 71 | 79 | 100 | 83 | 91 | 96 | 92 | 94 | Berhenti | 89 | 100 | 94 | 96 | 96 | 96 | 96 | 100 | 98 | Lanjut | 95 | 75 | 84 | 90 | 79 | 84 | 88 | 92 | 90 | Sakit | 92 | 96 | 94 | 92 | 100 | 96 | 96 | 96 | 96 | Kembali | 88 | 92 | 90 | 96 | 92 | 94 | 96 | 96 | 96 | Awas | 85 | 96 | 90 | 82 | 96 | 88 | 96 | 100 | 98 |
| Accuracy | | | 90 | | | 93 | | | 95 | Macro avg | 91 | 90 | 90 | 93 | 93 | 93 | 95 | 95 | 95 | Weighted avg | 91 | 90 | 90 | 93 | 93 | 93 | 95 | 95 | 95 |
|
|