The code corresponds to the paper https://arxiv.org/abs/2211.15379 or https://ieeexplore.ieee.org/document/10026879
pytorch 1.10.2 python 3.6.13
We use the dataset proposed in paper [55] and [56] to evaluate our proposed MAT-based SS-SEI method. The former is a large-scale real-world radio signal dataset based on a special aeronautical monitoring system, ADS-B, and the latter is WiFi dataset collected from USRP X310 radios that emit IEEE 802.11a standards compliant frames. The number of categories of ADS-B dataset and WiFi dataset is 10 and 16, respectively. The length of each sample of ADS-B dataset and WiFi dataset is 4,800 and 6,000, respectively. The number of training samples of ADS-B dataset and WiFi datsset is 3, 080. The number of testing samples of ADS-B dataset and WiFi dataset is 1,000 and 16,004, respectively. We construct five semi-supervised scenarios and one fully supervised scenario, where the number of labeled training samples to the number of all training samples ratio is {5%, 10%, 20%, 50%, 100%}, to evaluate the identification performance of the proposed SS-SEI method. In addition, 30% of the training samples is used as the validating samples during the training process.
[55] Y. Tu, Y. Lin, et al., “Large-scale real-world radio signal recognition with deep learning,” Chin. J. Aeronaut., vol. 35, no. 9, pp. 35–48, Sept. 2022.
[56] K. Sankhe, M. Belgiovine, F. Zhou, S. Riyaz, S. Ioannidis, and K. Chowdhury, “ORACLE: Optimized radio classification through convolutional neural networks,” in IEEE Conf. Comput. Commun., Apr.2019, pp. 370-378.
The dataset can be downloaded from the Link: https://pan.baidu.com/s/13qW5mnfgUHBvWRid2tY2MA Passwd：eogv
| Methods | ADS-B (5%) | ADS-B (10%) | WiFi (5%) | WiFi (10%) |
|---|---|---|---|---|
| CVNN | 60.50% | 74.50% | 20.47% | 28.64% |
| DRCN | 54.20% | 72.40% | 21.94% | 47.51% |
| SSRCNN | 49.30% | 79.30% | 19.33% | 38.09% |
| TripleGAN | 45.10% | 61.10% | 27.57% | 37.27% |
| SimMIM | 65.90% | 77.90% | 31.71% | 49.59% |
| MAT-CL | 70.06% | 83.80% | 27.26% | 80.70% |
| MAT-PA | 74.00% | 84.80% | 28.82% | 54.96% |
CVNN
DRCN
SSRCNN
TripleGAN
SimMIM
MAT
If you have any question, please feel free to contact us by e-mail (1020010415@njupt.edu.cn).