Modulation signal identification method and system based on deep learning

Abstract

The invention provides a modulation signal identification method based on deep learning. The method comprises the following steps: generating different types of modulation signals containing noise; carrying out wiener filtering noise reduction on the noise-containing modulation signal; performing cyclic spectrum estimation on the modulated signal after noise reduction, and extracting a cyclic spectrum two-dimensional sectional view; constructing a deep neural network, inputting the cyclic spectrum two-dimensional sectional view as an input feature into the deep neural network, and training the deep neural network; and identifying a modulation mode of an unknown signal by using the trained deep neural network. The invention further provides a modulation signal recognition system based on deep learning, noise reduction processing is performed on the modulation signal through Wiener filtering, and the influence of noise on recognition precision can be effectively reduced; meanwhile, the cyclic spectrum two-dimensional sectional view is used as an input feature, on one hand, the cyclic spectrum two-dimensional sectional view is not sensitive to noise, so that the influence of the noise on an identification result can be effectively reduced, and on the other hand, the complexity of an algorithm can be greatly reduced, and the identification efficiency can be improved.

Description

technical field [0001] The present invention relates to the technical field of communication technology application, in particular to a modulation signal recognition method and system based on deep learning. Background technique [0002] The purpose of modulation recognition technologies is to identify modulation schemes from received or intercepted signals in a short period of time in civilian or military applications, thus, they can make radio transceivers aware of application scenarios such as surveillance and electronic warfare. Modulation recognition algorithms can be divided into two main types, one is based on maximum likelihood and the other is based on feature extraction. Compared with feature extraction methods, maximum likelihood methods can achieve better recognition accuracy, but require prior knowledge of channel state information (CSI) or statistical properties of received signals. In contrast, feature extraction systems do not require priority information, s...

AI Technical Summary

Problems solved by technology

Literature [2] B. Ramkumar, "Automatic modulation classification for cognitive radios using cyclic feature detection," in IEEE Circuits and Systems Magazine, vol.9, no.2, pp.27-45, Second Quarter 2009, doi:10.1109 / MCAS. 2008.931739. Using cyclic spectrum feature analysis and decision tree-based classifier to realize the modulation type identification of the signal, but this method needs to manually extract several features of the cyclic spectrum, and the computational complexity is relatively large

In order to overcome the problem of high computational complexity, literature [3] Z.Wu, E.Like and V.Chakravarthy, "Reliable Modulation Classification at Low SNR Using Spectral Correlation," 20074th IEEE Consumer Communications and Networking Conference, 2007, pp.1134-1138 , doi:10.1109 / CCNC.2007.228. analyzed the extraction of α-section and f-section in cyclic spectral features, and used deep learning neural network to classify section features, but this method is greatly affected by noise and reduces the Modulated signal recognition accuracy

Images