Research Article

A Chinese Lip-Reading System Based on Convolutional Block Attention Module

Table 2

Testing accuracy of each model.

Model/characterVgg16 + LSTM + Attention (%)Vgg16 + GRU + Attention (%)InceptionV3 + LSTM + Attention (%)InceptionV3 + GRU + Attention (%)ResNet50 + LSTM + Attention (%)ResNet50 + GRU + Attention (%)

Ling (Zero)63.0062.0099.7089.0099.5097.00
Yi (One)85.0057.0050.0099.0053.0098.00
Er (Two)99.0099.0099.4099.0099.0099.00
San (Three)67.0067.0099.0095.0087.5099.00
Si (Four)80.0087.0099.0081.0072.0064.00
Wu (Five)99.0099.0099.0099.0099.0099.00
Liu (Six)46.0098.0098.0098.0099.0099.00
Qi (Seven)67.0079.0051.0078.0097.0098.00
Ba (Eight)99.0042.0099.0099.0099.0099.00
Jiu (Nine)85.0025.0075.0081.0097.0096.00
Chi-Fan (Eat)76.0021.0016.0081.0077.0099.00
Dui-Bu-Qi (Sorry)24.0098.0099.0099.0060.0099.00
Ni-Hao (Hello)99.0099.0099.0099.0099.0099.00
Pao-Bu (Run)97.0099.0099.0097.0093.0098.00
Shui-Jiao (Sleep)80.0088.0099.0097.0031.0099.00
Wan-Shua (Play)96.0099.0099.0099.0099.0098.00
Xue-Xiao (School)97.0096.0086.0053.0097.0099.00
Zai-Jian (Good-Bye)99.0095.0028.0099.0089.0099.00
Zhong-Guo(China)97.0092.0099.0098.0099.0099.00
Zou-Lu (Walk)98.0094.0099.0099.0096.0099.00