Modulated Signal Denoising Method Based on Autoencoder Neural Network

Abstract

The invention discloses a modulation signal denoising method based on self-encoding neural network, comprising the following steps: Step 1, using MATLAB simulation software to simulate a general communication link structure, generating noisy sample data sets and pure Sample data set; step 2, normalize [0, 1] for each sample set; step 3, build a modulation signal denoising autoencoder based on an autoencoder neural network and set hyperparameters; step 4, train a denoising model , using the backpropagation algorithm and the gradient descent method to optimize and update the values ​​of the parameters in the neural network to obtain the denoising model. The present invention uses a denoising network model based on the self-encoding neural network, avoids the complex preprocessing process of the signal in the traditional modulation signal denoising algorithm, has a simple overall structure process, a small amount of network calculation, and a fast denoising speed.

Description

technical field [0001] The invention belongs to the technical field of communication signal processing, in particular to a modulation signal denoising method based on an autoencoder neural network. Background technique [0002] The modulation recognition technology of communication signals is the key step and basic link to realize signal detection and demodulation, and has a wide range of applications in the fields of spectrum sensing, electronic countermeasures and defense. In recent years, feature-based automatic modulation recognition technology has made great progress, such as methods based on instantaneous features, high-order cumulants, cyclic spectrum techniques, time-frequency analysis, wavelet transform, etc. These methods usually use noisy modulation signals as samples set, it is necessary to design complex algorithms to extract feature quantities, and these feature quantities are often affected by noise information in the signal, especially in a low signal-to-nois...

AI Technical Summary

Problems solved by technology

Although this method proposes a denoising method for modulated signals, there are still shortcomings in this method: the training data only contains a single pulse waveform, and the pulse width only contains 2 to 3 ns, which does not cover the modulation in actual multiple modulation modes The various complex situations of the signal do not mean that a good effect has been achieved for the modulated signal in the actual complex and changeable situation.

Images