Research On Signal Recognition Of HF Frequency-hopping Radios

Frequency-hopping communication countermeasure is the main research area in electronic countermeasure.How to classify and recognize multi-hopping signals in complex electromagnetic environment is a critical issue in frequency-hopping signal reconnaissance.This thesis focuses on frequency-hopping signal of HF frequency-hopping radio and mainly studies how to count the quantity of multi-hopping signals and how to classify and recognize multi-hopping signals.contents are as follows:(1)Time-frequency analysis technologyAccording to three performance indexes: calculation time,cross-term interference and anti-noise,the performance comparisons of short-time Fourier transform,spectral transform,Wigner-Ville distribution and Cohen time-frequency analysis method are performed.The simulation results show that the spectral transform method has many advantages such as small computation,short calculation time,no cross-term interference and strong anti-noise performance,and is more suitable for time-frequency analysis of frequency-hopping signals.(2)Count the quantity of multi-hopping signalsThis thesis designs a parameter extraction algorithm for short-wave frequency-hopping signals to extract characteristic parameters such as hopping period,hopping frequency and hopping amplitude of frequency-hopping signals.Firstly,spectral transform is applied on the short-wave frequency-hopping signal to obtain time-frequency transform matrix.Secondly,the significant value of each time-frequency transform matrix is processed to obtain an array of frequencies and amplitudes.Finally,the frequency,amplitude and period of each hop of the frequency-hopping signal are obtained via comparing the difference between two adjacent arrays.This thesis designs a statistic algorithm to count the quantity of multi-frequency-hopping signals.First,Regarding the number of joint arrays at each moment as the quantity of frequency-hopping signals appearing at that moment;then,the number of frequency-hopping signals in a specific time interval are counted;finally,the proportion of the statistic values at all times is calculated.The maximum valueis used as the number of frequency-hopping signals during this time period.(3)Classification and identification of multi-hopping signalsA multi-hopping signal classification and recognition method based on stack sparse self-encoding is designed.Firstly,the stack sparse self-encoding and Softmax classifier are combined to form a short-wave hybrid frequency-hopping signal classification recognition model.Then,the time domain long signal is divided into multiple short-time signals.Finally,multiple short-time signals are inputted into the frequency-hopping signal classification recognition model as the training sample to do classification and identification,and the classification result is obtained.The experimental results show that: 1)Feature extraction method which based on two-layer stacking sparse self-encoding is feasible for the classification and identification of multi-hopping signals;2)when the SNR is no lower than-5dB,the average classification accuracy of this method is better than other similar algorithms;3)When the SNR is no less than-5dB and recognition accuracy is higher than 85%,the method can be applicable to classify and identify up to six hybrid frequency hopping signals.