Multi-scale convolutional neural network based radio signal modulation identification method

Abstract

The invention discloses a multi-scale convolutional neural network based radio signal modulation identification method. The multi-scale convolutional neural network based radio signal modulation identification method comprises the steps of (1) generating a processed radio modulation signal; (2) generating a two-dimensional time-frequency diagram and performing Fourier transform on an instantaneouscorrelation function of the signal to obtain a Wigner-Ville time-frequency distribution diagram of the signals; (3) performing pre-processing on the time-frequency distribution diagram to generate atraining sample set and a test sample set; (4) building a multi-scale convolutional neural network module and training the model; and (5) testing the test set by utilizing the trained network model, calculating the correction rate, obtaining an identification accuracy rate and assessing the network performance. The multi-scale convolutional neural network based radio signal modulation identification method has the advantages of strong universality, no need for manual characteristic extraction and a plenty of priori knowledge, low complexity and accurate and stable classification results, and can be used in the field of signal classification identification technologies.

Description

Technical field [0001] The present invention belongs to the technical field of signal processing, and further relates to a radio signal automatic modulation recognition method based on a multi-scale convolutional neural network. The invention can be applied to a complex electromagnetic environment to realize automatic feature extraction and modulation mode classification of radio signals, thereby making the radio signal modulation mode classification more flexible and efficient. Background technique [0002] Radio signal modulation recognition plays an important role in military electronic countermeasures, hostile reconnaissance, and signal acquisition analysis. In the case of extreme lack of known information, signal modulation recognition is the first step in the signal processing process, and the final recognition of information Played a decisive role. Due to the lack of prior information, major research institutions and universities at home and abroad have done a lot of work...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that although this method proposes a communication signal modulation identification method, it can only be used to identify continuous phase frequency shift keying signals, and a large amount of prior knowledge is required for signal feature extraction

Although this method proposes a robust communication signal modulation recognition method, the shortcomings of this method are: the model is complex, and the signal feature extraction is highly dependent on manual feature extraction

Images