Table 8 Friedman test on the SI-SNRi for ablation experiment
From: Time-domain adaptive attention network for single-channel speech separation
i | Model | \(z=(R_{taanet} - R_i)/SE\) | p | \(\alpha /(k-1)\) |
|---|---|---|---|---|
1 | Base model | \((3.2587-1.6468)/0.03566=45.15\) | 0.0000 | 0.0167 |
2 | Local only | \((3.2587-2.2163)/0.03566=29.22\) | 0.0000 | 0.0250 |
3 | Global only | \((3.2587-2.8764)/0.03566=10.72\) | \(1.85\times {10^{-26}}\) | 0.0500 |