Frequency hopping signal detection and parameter estimation method based on deep neural network

Abstract

The invention discloses a frequency hopping signal detection and parameter estimation method based on a deep neural network, belongs to the field of image recognition, and can be used for frequency hopping signal detection and parameter estimation. The method comprises the following steps: 1) generating and acquiring frequency hopping signal data; 2) classifying and sorting the frequency hopping signal data according to parameters such as a modulation mode, a signal-to-noise ratio and a frequency modulation frequency; 3) generating a time-frequency distribution waterfall plot from the classified frequency hopping signals, and generating an image data set; 4) constructing an SSD model; 5) labeling the generated time-frequency waterfall plot as input of an SSD target detection framework model for training; 6) applying the trained model to a test set to complete frequency hopping signal detection and parameter estimation, wherein k _ means clustering-based time-frequency waterfall plot correction is continued aiming at the noise problem of the time-frequency waterfall plot, and 8) performing comparative analysis on the performance before and after the time-frequency waterfall plot correction. The method is high in frequency hopping signal detection rate and accurate in parameter estimation, and has important significance for frequency hopping signal processing.

Description

technical field [0001] The invention relates to a frequency hopping signal detection and parameter estimation method based on a deep neural network, which belongs to the field of image recognition and non-cooperative communication and can be used for frequency hopping signal detection and parameter estimation. Background technique [0002] In recent years, with the rise of deep learning, it has made great achievements in image processing and natural language processing. Especially in image processing, the functions of deep learning are mainly target detection and classification recognition. The core of target detection is to use deep neural network to train labeled target pictures, and then use the trained model to perform target detection on unlabeled pictures. The use of deep learning for target detection, dynamic real-time tracking and positioning of targets, and in intelligent transportation systems, intelligent monitoring systems, military target detection, and surgica...

AI Technical Summary

Problems solved by technology

At the same time, for the noise problem of the time-frequency distribution map, a time-frequency map correction method based on k_means clustering is proposed

Images