Radar radiation source identification method based on DCNN and Transform

Abstract

The invention discloses a radar radiation source identification method based on DCNN and Transform, which introduces a Transform architecture into the field of radar radiation source identification, improves the architecture, overcomes the difficulty of learning remote dependence in the prior art, and breaks through the limitation of a previous convolutional neural network when calculating the correlation between two positions. An attention mechanism model is adopted, for data feature extraction of multiple channels, data channel features before global feature extraction are learned, channel dependency is learned, the importance of each channel after feature extraction is highlighted, and the characterization capacity of a radiation source time-frequency graph is improved. According to the method provided by the invention, local feature extraction can be firstly carried out on the time-frequency graph with the low signal-to-noise ratio, then the global feature is considered by considering the relationship between the associated positions, and the local feature and the global feature are integrated under the condition of the signal-to-noise ratio, so that a good recognition effect can be achieved.

Description

technical field [0001] The invention belongs to the field of radar countermeasure investigation, in particular to the field of radar radiation source signal recognition and relates to a radar radiation source recognition method based on DCNN and Transformer. Background technique [0002] In modern electronic warfare and radar technology confrontation warfare, the combat capability of radar reflects the military technical strength of a country. Its main function is to provide a strong guarantee for long-range detection and attacking enemy targets. Radar radiation source identification is an important aspect of wireless electronic warfare. The key core function, the process is to compare the measured radiation source signal parameters with the pre-existing parameters of our library samples to confirm the characteristics of the radiation source, and finally, analyze the observed and intercepted radar signals, Obtaining the opponent's tactical intelligence through identification...

AI Technical Summary

Problems solved by technology

[0006] In order to solve the above problems, we propose a radar emitter target recognition method based on DCNN and Transformer, which aims to solve the low signal-to-noise ratio recognition accuracy in the existing technology, and the modulation method with similar characteristics is difficult to recognize. Problem, this method first constructs the radar radiation source signal time-frequency map data set, preprocesses the time-frequency map to reduce the sensitivity of the sample; then builds the DCNN local feature extraction module, and the normalized data is used as the input of the DCNN, through The pooling layer captures the detailed features, and then uses the channel attention mechanism to model and analyze the features of each channel, adjust the dependence of each channel, and use the adjusted features as output, and then use the Transformer to process the features that have passed the channel attention , to make up for the weak global modeling ability of DCNN, build a classifier to classify radar radiation source targets, and use full connection to retain more effective features for the output of the Transformer model, and finally use softmax to classify the output of the network;

Images