Introduction
Related work
Mechanism of audio attention
The structure and function of the human air
The character of bottom-up attention model
Computational method
The framework of calculation method for attention
The computational model of auditory periphery
The Gammatone filter bank to simulate basilar membrane
The Meddis inner hair cell model
The two channels of getting local information entropy
The audio channel processing
The image channel processing
The exponential moving average (EMA) correlation process
Experiment and result analysis
Artificial sinusoidal audio signal
Audio signal of THCHS-30 corpus
The accuracy evaluation of testing result
Talk show | Category | Manual/s | Automatic/s | Accuracy/% |
---|---|---|---|---|
1 | 1 | 12.8–14.2 | 7.7–38.5 |
100
|
1 | 2 | 25.1–32.7 | 7.7–38.5 |
100
|
1 | 1 | 49.2–50.8 | – | – |
1 | 2 | 61.8–71.7 | 56.0–71.7 |
100
|
1 | 2 | 111.3–117.1 | 111.7–120.3 |
69
|
1 | 1 | 129.5–131.0 | – | – |
1 | 2 | 138.8–144.7 | 139.9–147.5 |
100
|
1 | 1 | 149.8–150.9 | – | – |
2 | 2 | 7.4–17.7 | 8.1–13.6 |
63.1
|
2 | 2 | 26.4–30.6 | 27.3–36.3 |
78.5
|
2 | 2 | 63.4–72.2 | 63.7–75.6 |
96.5
|
2 | 2 | 85.9–93.2 | 122.0–147.5 |
89.3
|
2 | 2 | 120.8–132.1 | 122.0–147.5 |
94.3
|
2 | 1 | 163.2–165.1 | – | – |