Blind signal modulation type identification method based on convolutional neural network

Abstract

The invention provides a blind signal modulation type recognition method based on a convolutional neural network, and the method comprises the steps: employing three phase difference histograms calculated at different sampling rates and an amplitude histogram as signal feature parameters, carrying out the classification of the signal feature parameters through a convolutional neural network classifier, and obtaining a signal modulation type. Because the texture of the phase difference histogram is insensitive to frequency offset and inaccurate bandwidth, the method has strong immunity to frequency offset and inaccurate bandwidth, and is suitable for blind signal modulation type identification in a complex electromagnetic environment.

Description

technical field [0001] The invention belongs to the field of radio communication, and in particular relates to a blind signal modulation type identification method based on a convolutional neural network. Background technique [0002] In the field of radio signal monitoring, modulation recognition technology is more faced with blind signal recognition, that is, the signal carrier frequency is unknown, the signal bandwidth is unknown, and the signal symbol rate is unknown. In a complex electromagnetic environment, it is difficult to accurately measure the carrier frequency and bandwidth of blind signals. From the viewpoint of engineering practicability, modulation recognition technology is required to be immune to frequency offset and inaccurate bandwidth. When using the modulation identification products provided by some well-known manufacturers, facing a blind signal, the operator often has to use several groups of carrier frequencies and bandwidths for identification to c...

AI Technical Summary

Problems solved by technology

[0003] In the traditional modulation type identification method based on instantaneous parameters and high-order cumulants, its characteristic parameters are sensitive to receiver frequency offset and filter bandwidth, and its performance in actual engineering cannot meet the application requirements

In the modulation type identification method based on the constellation diagram, it is necessary to obtain the accurate signal symbol rate and perform sampling rate conversion on the sampled data. In addition to the high computational overhead, the accurate symbol rate is often not obtained under the condition of low carrier-to-noise ratio, making it App is restricted

In the modulation type recognition algorithm using convolutional neural network, training data sets are often generated by means of simulation. The innate characteristics of these data sets are no frequency offset, and the trained model is difficult to adapt to the actual engineering application environment; some directly use IQ data As a modulation recognition method for neural network input data, it is sensitive to frequency offset, interference and carrier-to-noise ratio

Images