Table 3 Comparison of model complexity and performance on RML2016.10a & RML2016.10b.
From: A novel automatic modulation recognition algorithm for OFDM signals based on FAFT
Model | Parameters (M) | FLOPs (M) | Max Acc (a) | Avg Acc (a) | Max Acc (b) | Avg Acc (b) |
|---|---|---|---|---|---|---|
FAFT | 0.13 | \(3.93\times 10^{1}\) | 91.8% | 60.88% | 92.6% | 61.99% |
MAMC | 1.58 | \(6.73\times 10^{1}\) | 91.4% | 59.89% | 92.7% | 61.79% |
ResNet | 0.56 | \(5.62\times 10^{1}\) | 89.2% | 57.11% | 90.8% | 59.52% |
Transformer | 3.9 | \(1.68\times 10^{2}\) | 88.9% | 57.05% | 90.4% | 59.85% |
CLDNN | 0.16 | \(4.08\times 10^{1}\) | 89.6% | 58.67% | 89.8% | 60.42% |
MCNET | 0.11 | \(3.78\times 10^{1}\) | 87.1% | 56.70% | 90.1% | 60.75% |
