Image Processing Reference
In-Depth Information
Table 3
FER Confusion Matrix Using Depth Faces with ICA
Expression Anger Happy Sad Surprise Fear Disgust
Anger
80%
0
10
0
0
10
Happy
0
82.50
10
7.25
0
0
Sad
0
0
85
15
0
0
Surprise
0
0
0
85
15
0
Fear
0
5
15
0
85
0
Disgust
0
0
15
2.50
0
82.50
Average
83.33
Bold values indicate correct expression recognition and others indicate incorrect recognition.
Table 4
FER Confusion Matrix Using Depth Faces with ICA-LDA
Expression Anger Happy Sad Surprise Fear Disgust
Anger
85%
0
5
0
0
10
Happy
0
85
10
5
0
0
Sad
0
0
85
15
0
0
Surprise
0
0
0
87.50
12.50 0
Fear
0
2.50
15
0
87.50
0
Disgust
0
0
15
0
0
85
Average
85.83
Bold values indicate correct expression recognition and others indicate incorrect recognition.
Then, LBP was tried on the same database that achieved the average recognition rate of
89.20% as shown in
Table 5
. Furthermore, LDP was employed and achieved the beter recog-
nition rate than LBP, that is, 90.83% as shown in
Table 6
. Finally, LDP-PCA-LDA was applied
with HMM that showed superiority over the other feature extraction methods achieving the
highest recognition rate (i.e., 98.33%) as shown in
Table 7
.
Figure 9
shows FER performances
using different approaches where LDP-PCA-LDA shows its superiority.
Search WWH ::
Custom Search