Table 5 Performance of different frontends with DML improvements on the v2 dataset.
From: Dynamic convolution models for cross-frontend keyword spotting
Network types | Independent | DML-enhanced | DML-improvement | ||||
|---|---|---|---|---|---|---|---|
Net1 | Net2 | Net1 | Net2 | Net1 | Net2 | Net1 | Net2 |
LEAF+ TempConvNet | MFCC+ TempConvNet | 97.08 | 97.04 | 97.40 | 97.21 | 0.32 | 0.17 |
LEAF+ DynTempConvNet | MFCC+ DynTempConvNet | 97.27 | 97.25 | 97.52 | 97.48 | 0.25 | 0.23 |
