Table 9. Performance comparison of two classical classifiers (for region A).
Subject | LDA (% accuracy/computation time, sec) |
SVM (% accuracy/computation time, sec) |
||||
---|---|---|---|---|---|---|
| ||||||
Win size 0-5 sec |
Win size 0-10 sec |
Win size 0-15 sec |
Win size 0-5 sec |
Win size 0-10 sec |
Win size 0-15 sec |
|
1 | 78.8/ 0.06 | 80.4/ 0.03 | 82.1/ 0.02 | 80.3/ 1.2 | 83.3/ 0.7 | 84.3/ 0.4 |
2 | 91.5/ 0.04 | 91.8/ 0.04 | 94.5/ 0.03 | 92.2/ 0.5 | 93.0/ 0.4 | 94.2/ 0.3 |
3 | 86.8/ 0.05 | 88.7/ 0.03 | 89.5/ 0.03 | 88.6/ 0.7 | 90.1/ 0.4 | 91.3/ 0.3 |
4 | 85.9/ 0.06 | 87.0/ 0.03 | 89.8/ 0.02 | 85.9/ 1.4 | 87.0/ 0.8 | 89.8/ 0.3 |
5 | 74.6/ 0.06 | 76.1/ 0.04 | 79.4/ 0.02 | 75.5/ 0.9 | 77.7/ 0.6 | 83.6/ 0.5 |
6 | 78.0/ 0.05 | 79.2/ 0.03 | 80.7/ 0.02 | 78.8/ 0.9 | 80.1/ 0.5 | 81.9/ 0.4 |
7 | 85.6/ 0.06 | 85.9/ 0.03 | 86.4/ 0.02 | 88.4/ 1.6 | 89.2/ 0.9 | 90.8/ 0.6 |
8 | 91.3/ 0.04 | 90.1/0.03 | 90.4/ 0.02 | 93.1/ 0.7 | 93.8/ 0.4 | 94.7/ 0.3 |
9 | 78.8/ 0.06 | 77.9/ 0.04 | 79.0/ 0.02 | 79.0/ 0.8 | 79.8/ 0.4 | 80.1/ 0.2 |
10 | 82.2/ 0.04 | 82.8/ 0.03 | 83.6/ 0.02 | 83.4/ 0.8 | 84.6/ 0.5 | 85.6/ 0.3 |
11 | 73.6/ 0.06 | 74.5/ 0.04 | 73.2/ 0.03 | 75.8/ 1.0 | 77.4/ 0.6 | 78.9/ 0.4 |
12 | 85.0/ 0.05 | 86.4/ 0.03 | 85.8/ 0.02 | 87.4/ 0.8 | 88.3/ 0.5 | 89.4/ 0.3 |
13 | 88.6/ 0.05 | 88.4/ 0.03 | 89.4/ 0.02 | 89.1/ 0.9 | 90.3/ 0.5 | 91.0/ 0.3 |
| ||||||
Mean | 83.1/ 0.05 | 83.4/ 0.03 | 84.9/ 0.02 | 84.4/ 0.9 | 85.7/ 0.6 | 87.3/ 0.4 |
Note: The accuracies were acquired using only three features (mean ∆HbO, signal peak, and sum of peaks).