P-band burst signal modulation mode identification based on fusion network of preamble images and multi-semantic features
In this paper,a P-band burst signal modulation mode identification algorithm based on signal preamble image and multi-semantic feature fusion network is proposed,which makes full use of the preamble features of P-band burst signals and solves the problem that the signal information segment features are difficult to distinguish in low signal-to-noise ratio environment,and at the same time,it avoids the dependence of traditional signal preamble autocorrelation recognition algorithm on signal preamble a priori information and the influence of frequency bias.Firstly,the proposed algorithm uses the signal preamble frequency spectrum to complete the image construction,and uses the spectral features formed by different burst signal leading regular code words to distinguish;secondly,the proposed multi-semantic feature fusion network makes full use of the texture features such as edge contours of the signal preamble spectrum extracted by the low-level residual network,and also incorporates the abstract and complex semantic features of the signal spectrum extracted by the high-level residual network.The proposed multi-semantic feature fusion network can solve the problem that the residual network only uses the high-level abstract semantic features but ignores the low-level features,and improve the recognition performance of burst signal modulation mode.Experimental results show that,compared with the signal preamble autocorrelation algorithm,the recognition algorithms based on the information segment frequency spectrum image and the ResNet50 network,and the recognition algorithms based on the information segment time-frequency map and the ResNet50 network,the proposed algorithm improves recognition performance by 20.84%,30.83%,and 60.39%respectively in a-15 dB signal-to-noise ratio environment.
P-bandburst signal preamblepictorial constructionresidual networksfeature fusion
NETL
NSTL
万方数据
