A method for estimating signal noise floor based on deep learning

Abstract

The invention discloses a signal noise floor estimation method based on deep learning. set and validation set; design a one-dimensional deep convolutional neural network structure based on a deep convolutional autoencoder; normalize the amplitude of the training samples generated by the simulation, and send the training set and validation set to the convolutional neural network Perform iterative training, synthesize the loss size and evaluation results, adjust the network structure and parameters, and save the trained network model; normalize the collected real broadband power spectrum amplitude, and record the minimum amplitude and amplitude span of the power spectrum , the normalized data is sent to the trained network model for calculation, and the result is combined with the minimum amplitude and amplitude span of the real power spectrum recorded in the normalization process, and the estimated size of the noise floor of the final real power spectrum is calculated. .

Description

technical field [0001] The invention belongs to the fields of deep learning application and signal processing, and in particular relates to an intelligent signal spectrum noise floor estimation method applied in signal spectrum monitoring. Background technique [0002] With the continuous improvement of information technology, society is full of various complex radio signals all the time. In the process of transmission of these information, it is inevitable to be interfered by a variety of noises. In addition to Gaussian white noise, there are also a large number of non-stationary noise and non-Gaussian noise, which make the signal receiver face a huge challenge to correctly receive the information. Especially in the military field, on the one hand, it is the noise interference in the natural environment, and on the other hand, it is more to face the targeted communication interference of the enemy. In the background of these noise interference, the signal spectrum will show...

AI Technical Summary

Problems solved by technology

However, the improved algorithm still needs to artificially set some initial values, and in the face of further degradation of the measured signal environment, such as uneven noise distribution, excessive noise power, etc., the estimation algorithm will be seriously degraded

In addition, the nonlinear autoregressive filtering algorithm needs to traverse the entire power spectrum. When the resolution of the power spectrum is small, it will take a lot of time to estimate the noise floor, which is not conducive to integration into the actual signal detection link.

Images