Intelligent Method for Real-Time Portable EEG Artifact Annotation in Semiconstrained Environment Based on Computer Vision
Table 1
Five groups of experimental result.
Blink left
Blink right
Slight head movement
Severe head movement
Detection
Wrong
Miss
TPR (%)
SPC (%)
Detection
Wrong
Miss
TPR (%)
SPC (%)
Detection
Wrong
Miss
TPR (%)
SPC (%)
Detection
Wrong
Miss
TPR (%)
SPC (%)
CV
108
24
2
97.67
97.05
104
21
3
96.51
97.42
35
4
2
93.94
99.54
9
0
0
100.00
100.00
Observer A
91
11
6
93.02
98.65
91
11
6
93.02
98.65
40
15
8
75.76
98.27
22
13
0
100.00
98.54
Observer B
89
9
6
93.02
98.89
89
9
6
93.02
98.89
61
36
8
75.75
95.85
15
6
0
100.00
99.33
Observer C
88
10
8
90.70
98.77
88
10
8
90.70
98.77
32
4
15
84.85
99.53
11
2
0
100.00
99.86
Reference value
86
—
—
—
—
86
—
—
—
—
33
—
—
—
—
9
—
—
—
—
TPR represents the sensitivity of the data: the rate of real blinks or head movements. SPC represents the specificity of the data: the rate of real nonblinking or non-head-movements. TPR = (detection-wrong)/(detection-wrong + miss) × 100%. SPC = (total-detection-miss)/(total-detection-miss + wrong) × 100%.
