Digital Signal Processing Reference
In-Depth Information
Table 3.8
Emotion classification performance using local prosodic features computed over entire
sentences
Emotions
Ang.
Dis.
Fear
Hap.
Neu.
Sad
Sar.
Sur.
Duration, average emotion recognition: 48.75
Anger
30
20
7
3
23
3
7
7
Disgust
7
67
3
0
10
3
10
0
Fear
7
7
53
0
10
10
7
6
Happiness
17
7
10
30
3
13
7
13
Neutral
7
3
3
3
57
21
3
3
Sadness
3
7
23
7
20
30
10
0
Sarcasm
0
13
4
10
0
0
73
0
Surprise
7
3
14
10
3
10
3
50
Pitch, average emotion recognition: 53.75
Anger
27
43
7
0
3
3
17
0
Disgust
10
60
10
0
0
3
7
0
Fear
3
13
43
7
0
10
7
17
Happiness
4
7
13
40
3
13
13
7
Neutral
3
7
0
3
80
7
0
0
Sadness
3
10
7
7
10
57
6
0
Sarcasm
0
0
7
3
0
10
63
17
Surprise
0
0
7
10
0
3
20
60
Energy, average emotion recognition: 48
Anger
43
37
7
0
7
0
6
0
Disgust
27
37
0
0
13
0
20
3
Fear
0
0
57
7
10
13
0
13
Happiness
7
7
10
43
17
7
7
2
Neutral
20
3
3
10
47
10
3
4
Sadness
0
3
13
17
17
40
10
0
Sarcasm
0
10
3
0
0
0
80
7
Surprise
0
0
37
13
3
0
10
37
Duration
+
Pitch
+
Energy, Average: 64.38
Anger
40
4
23
27
3
0
3
0
Disgust
13
73
0
0
4
0
10
0
Fear
3
0
63
10
0
7
0
17
Happiness
7
0
10
57
3
13
3
7
Neutral
0
7
7
3
73
7
0
3
Sadness
0
7
13
7
0
63
10
0
Sarcasm
0
10
0
0
0
7
83
0
Surprise
3
0
17
10
0
0
7
63
Ang.
Anger,
Dis.
Disgust,
Hap.
Happiness,
Neu.
Neutral,
Sar.
Sarcasm, and
Sur.
Surprise
crimination of about 54%. Energy and duration dynamic features have also achieved
a recognition performance around 48%. From the results, it is observed that local
prosodic features play a major role in discriminating the emotions. Score level com-
bination of energy, pitch and duration features further improved the emotion recog-
nition performance up to around 65%. Measures of emotion recognition models
developed using global and local prosodic features are combined for improving the
Search WWH ::
Custom Search