Table 1. Comparative results for magnitude domain transformation-based data augmentation methods.
| Model | None | Jittering | Rotation | Scaling | Magnitude Warping | ||||
|---|---|---|---|---|---|---|---|---|---|
| Ave. (%) | Ave. (%) | t | Ave. (%) | t | Ave. (%) | t | Ave. (%) | t | |
| MLP | 70.23±21.91 | 70.52±21.67 | 1.55 | 69.13±21.70 | -1.95* | 70.24±21.97 | 0.05 | 69.43±22.50 | -2.91*** |
| VGG | 73.02±23.05 | 74.21±22.57 | 0.91 | 71.87±22.14 | -0.84 | 73.69±23.86 | 0.53 | 74.42±22.63 | 1.45 |
| ResNet | 81.39±17.19 | 80.87±18.08 | -0.87 | 78.01±19.93 | -.31*** | 81.92±16.81 | 1.05 | 81.33±17.15 | -0.08 |
| LSTM | 53.46±28.10 | 54.72±27.10 | 1.09 | 49.64±26.93 | -2.95*** | 53.69±26.94 | 0.21 | 54.35±27.29 | 0.93 |
| BLSTM | 62.51±24.74 | 60.23±24.87 | -2.01** | 57.76±24.76 | -3.51*** | 62.13±25.89 | -0.37 | 62.00±26.51 | -0.44 |
| LSTM-FCN | 81.54±17.53 | 79.18±19.87 | -4.89*** | 78.87±19.37 | -3.66*** | 80.73±18.37 | -1.74* | 79.46±19.68 | -3.15*** |
* p < 0.1,
** p < 0.05,
*** p < 0.01