Machine-Learning-Based Classification of Frequency Hopping in Radio Networks for Communication Reconnaissance

Abstract

This article presents a customized approach for training a supervised learning neural network with the adaptive moment estimation algorithm, to classify the number of frequency hopping networks in an operational area. The algorithm was constructed based on data experimentally collected from a real-time spectrum analyzer for military very high frequency hopping networks. The impact of some training parameters on classification efficiency is briefly discussed while the obtained accuracy was above 97% for both test and validation data in all training variations. With this promising result, the proposed algorithm has the potential to be utilized in developing operational systems capable of real-time signal reconnaissance for military frequency hopping radio networks.