A Method for Pre-Sorting FH Signals in the Case of Multiple Hopping Speeds

Abstract

The present invention belongs to the field of frequency-hopping communication reconnaissance, and particularly relates to a method for pre-sorting frequency-hopping signals in the case of multiple hopping speeds. In this scenario, there are multiple frequency-hopping network station signals with different hopping speeds and background radiation source signals at the same time. By using the regular characteristics of the pulse duration of the frequency-hopping network station, a noise threshold is set, and the duration of all signals exceeding the noise threshold in the sampled time-frequency band is detected, and a statistical histogram of the repetition times of the duration of the signals exceeding the noise threshold is formed. Taking the peak prominence as the repetition times threshold, the range value exceeding the threshold is the estimated value of the pulse duration range of the frequency-hopping network station at different hopping speeds. Through this estimated time range value, the frequency-hopping signals with different hopping speeds can be pre-sorted. The present invention can not only effectively remove the influence of noise and background signals like traditional pre-sorting methods, but also remove the mutual influence of frequency-hopping signals at different hopping speeds, which helps to reduce the complexity of frequency-hopping signal parameter estimation and signal sorting. The method is simple and has good effects.

Description

Technical Field

[0001] The present invention belongs to the field of frequency hopping communication reconnaissance, and particularly relates to a method for pre-sorting frequency hopping signals in the case of multiple hopping speeds. Background Art

[0002] Frequency hopping communication is one of the most commonly used spread spectrum communication methods. Its working principle refers to a communication method in which the carrier frequencies of the transmitting and receiving parties transmit signals discretely change according to a predetermined rule, that is, the carrier frequencies used in communication are randomly hopped under the control of a pseudo-random change code. From the perspective of the implementation method of communication technology, frequency hopping communication is a communication method using a code sequence for multi-frequency frequency shift keying, and is also a communication system with code-controlled carrier frequency hopping. Due to its strong anti-interference ability, low interception probability, and good suppression effect on frequency selective fading, frequency hopping communication is widely used in military communication, such as short-wave and ultra-short-wave radio stations using frequency hopping technology, and at the same time, it is rapidly penetrating into civilian communication, such as mobile communication, data transmission, computer wireless data transmission, wireless local area network, etc.

[0003] Frequency hopping communication reconnaissance usually includes three major tasks: detection of frequency hopping signals, parameter estimation, and signal sorting. Detection of frequency hopping signals refers to intercepting unknown frequency hopping signals of the enemy, including detecting frequency hopping signals mixed in noise; parameter estimation is to estimate parameters such as the hopping rate, hopping time, hopping pattern, and direction of arrival of the detected unknown frequency hopping signals; sorting is to use the estimated parameters to perform network sorting on the intercepted frequency hopping signals and sorting of network stations within the same network, so as to facilitate subsequent demodulation, decryption, or tracking interference, etc.

[0004] Currently, the mainstream pre-sorting technologies for frequency hopping network stations mainly include various methods to obtain the time-frequency channel noise threshold and denoise, and filter out the influence of non-frequency hopping signals such as fixed frequency, intermittent, and burst signals by setting the hopping speed range. Its essence is to convert the hopping speed range into a frequency hopping pulse duration range, and the signals outside this range are excluded as non-frequency hopping signals. However, there is no clear method for how to set the hopping speed range, and this method cannot effectively distinguish frequency hopping signals with different hopping speeds in the case of multiple hopping speeds.

[0005] The pre-sorting means adopted by the present invention is to use the frequency hopping pulse duration parameters at different hopping speeds for signal pre-sorting, which can decompose the sorting of frequency hopping signals at multiple hopping speeds into the sorting of frequency hopping signals at the same hopping speed, and helps to reduce the complexity of frequency hopping signal parameter estimation and signal sorting. Summary of the Invention

[0006] Aiming at the problem of pre-sorting frequency hopping signals, the present invention proposes a method for pre-sorting frequency hopping signals based on the estimation of the duration length of frequency hopping signal pulses.

[0007] The technical solution adopted by the present invention is as follows:

[0008] When there are multiple FH network station signals with different hopping speeds and background radiation source signals simultaneously, by using the pulse duration law characteristics of the FH network station, set the noise threshold, detect the duration of all signals exceeding the noise threshold in the sampling time-frequency band, and form a histogram of the repetition times of the duration of the signals exceeding the noise threshold. Taking the peak prominence as the repetition times threshold, the range value exceeding the threshold is the estimated value of the pulse duration range of the FH network station at different hopping speeds. Different hopping speed FH signals can be pre-sorted through this estimated time range. Specifically, it includes:

[0009] Define that m FH signals and n background signals are received in the set time-frequency band as:

[0010] y n = y TH1 + y TH2 + … + y THm + y bs1 + y bs2 + … + y bsn

[0011] Use the spectrogram to perform time-frequency representation on the received signal y n to obtain the time-frequency matrix S n , where the definition of the spectrogram is the square of the modulus of the short-time Fourier transform, and the time-frequency spectrogram is expressed as:

[0012]

[0013] Obtain the noise threshold ε1 through the denoising method, set the values in the time-frequency spectrogram that exceed the noise threshold to 1, and the values less than the noise threshold to 0:

[0014]

[0015] The obtained is an over-noise-threshold matrix that only contains elements 0 and 1. For each row of it, that is, each channel, the length of consecutive 1s is the length corresponding to the signal. Respectively count the length of consecutive 1s, that is, estimate the pulse lengths of all signals exceeding the threshold;

[0016] Perform histogram statistics on the repetition times of all pulse lengths;

[0017] Perform normalization processing on the obtained histogram. By using the characteristic that the pulse duration of the FH pulse signal is fixed, set the peak prominence as the repetition times threshold ε2. In the normalized statistical histogram, the repetition times exceeding the threshold are the estimated pulse duration range of the FH network station signal. Finally, obtain j pulse length ranges exceeding the threshold, and the values are recorded as:

[0018] T k = [T k_min , Tk_max , where \(k = 1:j\)

[0019] where \(T\) k_min is the minimum value of the \(k\)-th pulse length range, and \(T\) k _ max is the maximum value of the \(k\)-th pulse length range;

[0020] Use the obtained pulse length range \(T\) k to perform \(j\) screening processes respectively, and remove the parts that are not within the range of \(T\) k . Specifically:

[0021] ① Let \(k = 1\);

[0022] ② Remove the signals with a duration greater than \(T\) k_max or less than \(T\) k_mmin . The screened signal part is the pre-estimated frequency-hopping network station signal;

[0023] ③ \(k=k + 1\);

[0024] ④ When \(k\lt j\), go back to step ②, otherwise end the calculation;

[0025] Finally, \(j\) pre-estimated frequency-hopping signals are selected according to the pulse lengths of the frequency-hopping signals.

[0026] The present invention has the following beneficial effects:

[0027] After estimating the pulse duration of the frequency-hopping network station, the method of signal duration screening can be used to remove the influence of non-frequency-hopping signals such as fixed-frequency, intermittent, and burst signals;

[0028] Separate the network stations with different frequency-hopping pulse durations (i.e., different hopping speeds), which helps to reduce the complexity of frequency-hopping signal parameter estimation and signal sorting;

[0029] The present invention is simple and convenient and easy to implement. Description of the Drawings

[0030] Figure 1 is the flowchart of the pre-sorting method for frequency-hopping signals in the case of multiple hopping speeds;

[0031] Figure 2 is the time-frequency diagram of the detection of 4 frequency-hopping network stations, LFM signals, and fixed-frequency signals with 2 different hopping speeds;

[0032] Figure 3 is the statistical histogram of the received signal power obtained by the histogram denoising method;

[0033] Figure 4 is the statistical histogram of the number of signal durations exceeding the noise threshold;

[0034] Figure 5It is a diagram for the case where the repetition times of the signal duration exceed the peak prominence threshold;

[0035] Figure 6 It is the hopping network station signal at the first hopping speed screened out using the estimation range of the first hopping pulse duration;

[0036] Figure 7 It is the hopping network station signal at the second hopping speed screened out using the estimation range of the second hopping pulse duration; Detailed implementation method

[0037] Refer to Figure 1 , this method specifically includes the following steps:

[0038] (1) Detect m hopping signals and n background signals within a certain time-frequency band as:

[0039] y n = y TH1 + y TH2 +…+ y THm + y bs1 + y bs2 +…+ y bsn (1)

[0040] (2) Based on the spectrogram effect of the short-time Fourier transform, which can clearly identify signals and has been widely used in engineering, the received signal y n is represented in the time-frequency domain to obtain the time-frequency matrix S n . The so-called spectrogram is defined as the square of the modulus of the short-time Fourier transform, and the time-frequency spectrogram is expressed as:

[0041]

[0042] (3) Obtain the noise threshold ε1 through denoising methods such as energy threshold denoising and histogram denoising. Set the values in the time-frequency spectrogram that exceed the noise threshold to 1 and those less than the noise threshold to 0:

[0043]

[0044] (4) What is obtained in step (3) is actually a matrix that only contains elements 0 and 1 and exceeds the noise threshold. For each row of it, that is, each channel, the length of consecutive 1s is the length corresponding to the signal. Therefore, the lengths of consecutive 1s are respectively counted to estimate the pulse lengths of all signals that exceed the threshold.

[0045] (5) Perform a histogram statistics on the repetition times of all pulse lengths. Considering the influence of detection errors, the statistical width of the statistical histogram can be appropriately adjusted.

[0046] (6) Normalize the histogram in (5). Taking advantage of the characteristic that the pulse duration of a general frequency-hopping pulse signal is fixed, set the peak prominence as the repetition count threshold ε2. The peak prominence is used to measure the prominence of a peak (the degree of prominence relative to the positions of other peaks), and it can be set to 0.33 according to empirical values. In the normalized statistical histogram, those with a repetition count exceeding this threshold are the estimated pulse duration ranges of the frequency-hopping network station signals.

[0047] Suppose finally j pulse length ranges exceeding the threshold are obtained, and their values are denoted as:

[0048] T k =[T k_min ,T k_max , k = 1∶j (4)

[0049] where T k_min is the minimum value of the k-th pulse length range, and T k_max is the maximum value of the k-th pulse length range.

[0050] (7) Perform j times of screening processing respectively with the pulse length ranges T k obtained in (6), and remove the parts that are not within the range of T k . Specifically:

[0051] ① Let k = 1;

[0052] ② Remove the signals with a duration greater than T k_max or less than, T k_max . The filtered signal part is considered as the pre-estimated frequency-hopping network station signal;

[0053] ③ k = k + 1;

[0054] ④ When k < j, go back to step ②, otherwise end the calculation.

[0055] Finally, j pre-estimated frequency-hopping signals will be selected according to the pulse lengths of the frequency-hopping signals. Considering the influence of complex signals, a certain false alarm rate is allowed, that is, j ≥ m. In the subsequent frequency-hopping signal detection, it is easy to eliminate the pre-screened signals that are not frequency-hopping network station signals by using the hop period parameter.

[0056] In a simulation example, 4 non-orthogonal frequency-hopping network stations are designed as the frequency-hopping reconnaissance target objects, with a frequency range of 31 - 46 MHz and hopping speeds of 1000 hops / s, 1000 hops / s, 2000 hops / s, and 2000 hops / s respectively; 1 LFM signal with a frequency range of 40 - 41 MHz and 1 fixed-frequency signal with a center frequency of 32 MHz are designed as background signals.

[0057] (1) Detecting 4 frequency-hopping signals and 2 background signals within the frequency band of 31 - 46 MHz are as follows:

[0058] y n = y TH1 + y TH2 + y TH3 + y TH4 + y bs1 + y bs2 (5)

[0059] (2) Represent the detected signal y n using a time-frequency spectrogram, and obtain the time-frequency diagram of the detected signal as shown in Figure 2 .

[0060] (3) Conduct a histogram statistics on the power magnitudes in the signal time-frequency diagram obtained in (2), and obtain the power distribution of the received signal as shown in Figure 3 . Using the histogram denoising method, obtain the noise threshold ε1 = -113 dB. The idea of histogram denoising is to regard the noise power as the background part and the signal power as the useful information part, and find the difference point between the two to remove the noise. Set the values in the time-frequency spectrogram that exceed the noise threshold to 1 and those less than the noise threshold to 0:

[0061]

[0062] (4) From the matrix exceeding the noise threshold obtained in step (3), for each row, that is, for each channel, statistically count the length of consecutive 1s, which is to estimate the pulse lengths of all signals exceeding the threshold.

[0063] (5) Conduct a histogram statistics on the repetition times of all pulse lengths, and obtain the pulse length statistical histogram as shown in Figure 4 .

[0064] (6) Normalize the histogram in (5) to obtain the normalized pulse duration distribution as shown in Figure 5 . Set the peak prominence threshold ε2 to 0.33, and there are two peaks exceeding the threshold ε2.

[0065] According to the two peaks, obtain two pulse length ranges exceeding the threshold, and their values are respectively: [T 1_min , T 1_max = [988, 1050] μs; [T 2_min , T 2_maax = [1988, 2050] μs;:

[0066] (7) Use the two pulse length ranges obtained in (6) to perform two screening processes respectively, and remove the parts outside the pulse length ranges. Specifically:

[0067] ① Let k = 1:2;

[0068] ② Remove signals with a duration greater than T k_max or less than T k_end The filtered signal part is considered as the pre-estimated FH network station signal;

[0069] ③ k = k + 1;

[0070] ④ When k < 2, go back to step ②, otherwise end the calculation.

[0071] Finally, two pre-estimated FH signals will be selected according to the pulse length of the FH signal. The time-frequency diagrams of the FH network station signals at two hopping speeds selected according to the pulse length ranges of the two different pre-estimated FH signals are respectively as Figure 6 、 Figure 7 shown.

[0072] Pre-sorting effect:

[0073] Compare the original simulation parameters of the FH signal, Figure 1 the time-frequency diagram of signal detection and reception shown as Figure 5 、 Figure 6 the filtered time-frequency diagrams shown as. This method can successfully sort out the FH network station signals at different hopping speeds, and effectively filter the influence of background signals such as noise, fixed frequency, and intermittent signals, which verifies the effectiveness of the proposed pre-sorting method for FH signals in the case of multiple hopping speeds.

Claims

1. A method for pre-sorting frequency-hopping signals in the case of multi-hop speeds, characterized in that Including: Define that m hopping signals and n background signals are detected within the set time-frequency band as: ， Use a spectrogram for the intercepted signal to perform time-frequency representation and obtain a time-frequency matrix , where the definition of the spectrogram is the square of the modulus of the short-time Fourier transform, and the time-frequency spectrogram is expressed as: ， Obtain the noise threshold through the denoising method , set the values in the time-frequency spectrogram that exceed the noise threshold to 1 and those less than the noise threshold to 0: ， Obtained is a matrix passing the noise threshold that only contains elements 0 and 1. For each row of it, that is, for each channel, the length of consecutive 1s is the length corresponding to the signal. The lengths of consecutive 1s are statistically counted respectively, that is, the pulse lengths of all signals passing the threshold are estimated; Perform a histogram statistics on the repetition times of all pulse lengths; Normalize the obtained histogram. Utilize the characteristic that the pulse duration of the frequency-hopping pulse signal is fixed, and set the peak prominence threshold to . In the normalized statistical histogram, those exceeding the threshold are the estimated pulse duration range of the frequency-hopping network station signal. Finally, pulse length ranges exceeding the threshold are obtained, and the values are recorded as: ， wherein is the minimum value of the k-th pulse length range, is the maximum value of the k-th pulse length range; Using the obtained pulse length range Perform respectively times of screening processing, and remove the parts that are not within the range. Specifically: ① Let k = 1; ② Remove signals with a duration greater than or less than . The filtered signal part is the pre-estimated frequency-hopping network station signal; ③ k = k + 1; ④ When k < j, go back to step ②, otherwise end the calculation; Finally, j hopping signals under the pre-estimation are screened out according to the pulse lengths of the hopping signals.

Images