Abstract | ||
---|---|---|
Specific emitter identification (SEI) is a technology that distinguishes radio emitters by using the external feature carried by signals. This paper presents a novel deep-learning-based SEI approach that uses the nonlinear features of the steady-state signals. First, we perform Hilbert-Huang transform and bispectrum calculation on the signal to obtain the Hilbert spectrum and bispectrum. These two methods can effectively provide the nonlinear features contained in the signal. Then, we reduce the dimension of Hilbert spectrum by calculating the marginal spectrum (MS), and propose a broken-line contour integration method to reduce the dimension of bispectrum. Thereafter, we construct a multi-input convolutional neural network (CNN) with two network branches to identify different features in a targeted manner to achieve a better effect of identifying specific emitters. According to the analysis, the proposed approach retains complete feature information and reduces the number of network parameters. The experiments validate that the broken line contour integration method is effective, and also demonstrate that multi-input CNN outperforms single-input CNN. |
Year | DOI | Venue |
---|---|---|
2022 | 10.1109/ICCC55456.2022.9880738 | 2022 IEEE/CIC International Conference on Communications in China (ICCC) |
Keywords | DocType | ISSN |
Specific emitter identification,multi-input convolutional neural network,nonlinear signal,Hilbert spectrum,bispectrum,information integrity | Conference | 2377-8644 |
ISBN | Citations | PageRank |
978-1-6654-8481-7 | 0 | 0.34 |
References | Authors | |
6 | 4 |
Name | Order | Citations | PageRank |
---|---|---|---|
Xin Wang | 1 | 0 | 18.25 |
Jianing Zhao | 2 | 0 | 0.34 |
Tianwen Yang | 3 | 0 | 0.34 |
Fei Xu | 4 | 0 | 0.34 |