Table 5.
Average classification rate for subgroups of features and different fusion rules
| subgroup | Features | Fusion method | Level 1 accuracy | Level 2 accuracy | Level 3 accuracy | Level 4 accuracy | Average accuracy in all levels |
|---|---|---|---|---|---|---|---|
| 1 | f0, f1, f2, f1/f2, f3, RI, nasality, approximate entropy | Stacked fusion | not classified |
54.0% | 75.0% | 80.0% | 65.7% |
| MVR | 93.8% | 68.8% | 100.0% | 87.9% | 87.4% | ||
| Linear combination | 93.8% | 87.5% | 100.0% | 93.9%% | 93.8% | ||
| 2 | f0, f1, f2, f1/f2, f3, nasality, approximate entropy, fractal dimension | Stacked fusion | 63.0% | 65.2% | 76.3% | 82.0%% | 71.2% |
| MVR | 100.0% | 81.2% | 100.0% | 91.0% | 93.1% | ||
| Linear combination | 87.5% | 81.2% | 100.0% | 100.0% | 92.2% | ||
| 3 | f0, f1, f2, f1/f2, f3, RI, nasality approxiamate entropy, lyapanov exponent | Stacked fusion | not classified | not classified | 71.2% | 79.4% | 62.5% |
| MVR | 100.0% | 75.0% | 100.0% | 91.0% | 91.5% | ||
| Linear combination | 87.5% | 81.2% | 100.0% | 94.0% | 90.7% | ||
| 4 | f0, f1, f2, f1/f2, f3, RI, nasality approxiamate entropy, fractal dimension, lyapanov exponent, wavelet coefficients in 3 scales | Stacked fusion | 60.0% | 62.0% | 80.0% | 87.5% | 73.8% |
| MVR | 100.0% | 93.8% | 100.0% | 94.0% | 96.9% | ||
| Linear combination | 100.0% | 68.8% | 100.0% | 100.0% | 92.2% | ||