Skip to main content
Figure 13 | EURASIP Journal on Advances in Signal Processing

Figure 13

From: Multi-pose lipreading and audio-visual speech recognition

Figure 13

Lipreading with frontal, lateral and posed normalized sequences. The first column corresponds to the frontal sequences tested on the frontal classifier. The second column to the lateral sequences with the corresponding lateral classifier and the rest of columns to the pose-normalized lateral sequences tested on the frontal classifier. Third and forth columns corresponds to the projective transforms, while the last three correspond to the best LR technique applied to each feature space, i.e., LLR for the image space ( with 32 × 32 pixel patches, 75% overlapping, β = 15) and GLR on the selected DCT (β = 5) and LDA (β = 0) feature spaces.

Back to article page