Research Article
Field Deviation in Radiated Emission Measurement in Anechoic Chamber for Frequencies up to 60 GHz and Deep Learning-Based Correction
Table 2
Performance of the three deep learning models.
| Measurement condition | Model type | Minimum bias | Maximum bias | MAE | MRE (%) | Time-consuming (s) |
| 3 m distance in FAC | CNN-transformer | −3.36 | 4.63 | 2.12 | 4.83 | 2448 | CNN-GRU | −7.75 | 12.59 | 5.75 | 12.58 | 417 | CNN-LSTM | −2.39 | 10.38 | 5.39 | 11.50 | 886 |
| 3 m distance in SAC | CNN-transformer | −12.09 | 15.41 | 4.71 | 6.35 | 2464 | CNN-GRU | −10.66 | 12.19 | 4.82 | 6.46 | 413 | CNN-LSTM | −11.13 | 13.52 | 6.00 | 8.12 | 883 |
| 10 m distance in FAC | CNN-transformer | −0.86 | 0.21 | 0.26 | 2.61 | 2397 | CNN-GRU | −0.35 | 0.42 | 0.14 | 1.48 | 409 | CNN-LSTM | −0.72 | 0.88 | 0.39 | 4.33 | 875 |
| 10 m distance in SAC | CNN-transformer | −2.14 | 1.77 | 0.74 | 4.40 | 2400 | CNN-GRU | −2.31 | 1.63 | 0.89 | 5.74 | 409 | CNN-LSTM | −2.15 | 1.66 | 0.79 | 5.06 | 877 |
|
|