Radio Signal Feature Extraction Method Based on Attention Deep Network

Abstract

The invention discloses a radio signal feature extraction method based on attention depth network. The implementation steps are: generating a training sample set and a test sample set on the basis of constructing a coded and modulated joint radio signal; constructing an attention mechanism layer for extracting The attention depth network of radio signal features; the training sample set is input into the attention depth network for training, and the trained attention depth network is obtained; the recognition accuracy is obtained by using the test sample set and the trained attention depth network. The invention has the advantages of strong universality, no need for manual feature extraction, redundant information removal, low complexity, accurate and stable classification results, and can be used for cognitive recognition and other processing of subsequent radio signals.

Description

Technical field [0001] The present invention belongs to the field of communication technology, and further relates to a radio signal feature extraction method based on the attention depth network in the field of signal processing technology. The present invention simulates the visual selective attention mechanism in the information processing mechanism of the human brain, and can automatically extract radio signal characteristics and quickly screen out high-value characteristics in a complex electromagnetic environment for subsequent cognitive recognition of radio signals. Compared with the existing deep learning model, the present invention not only has the characteristics of low calculation complexity, small parameter scale, and easy hardware implementation, but also can obtain accurate radio signal modulation and coding mode identification under a lower signal-to-noise ratio. Background technique [0002] Radio signal feature extraction is the basic step of radio signal proces...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that although this method proposes a method for identifying the modulation mode of cognitive radio signals under the condition of low signal-to-noise ratio, this method requires frequency domain conversion and complex mathematical operations when extracting signal features , requires a lot of prior knowledge, highly relies on manual feature extraction, and the feature extraction method is complex

The disadvantage of this method is that although this method proposes a radio feature extraction method based on neural network to realize signal classification, it cannot perform effective feature extraction on radio signals, so that it cannot realize the identification of radio signal modulation and coding methods , and this method requires wavelet analysis of one-dimensional signals to use neural networks for feature extraction and classification recognition

Images