A direct joint detection and tracking method for frequency-hopping radiation sources

By employing a direct tracking method using fuzzy functions and Bernoulli filters in radiation source tracking, and utilizing all information from the received signal for channelization processing and likelihood function construction, the problem of error accumulation in traditional methods is solved, and accurate tracking of multiple radiation sources under low signal-to-noise ratio conditions is achieved.

CN118444306BActive Publication Date: 2026-11-17UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202410418599.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-09
Publication Date
2026-11-17
Estimated Expiration
2044-04-09

AI Technical Summary

Technical Problem

Traditional radiation source tracking methods suffer from error accumulation in complex scenarios, making it difficult to achieve real-time and effective target tracking. In particular, they are not performing well under low signal-to-noise ratio conditions, and it is difficult to accurately distinguish and correct the number and trajectory of targets when tracking multiple radiation sources.

Method used

A Bernoulli filter based on a fuzzy function is adopted to directly track the target using all the information of the received signal. By channelizing multiple frequency band signals and integrating the tracking results, a likelihood function is constructed to avoid intermediate errors in the two-step method. Soft decision is used instead of hard decision, and the particle approximation is used to realize the Bernoulli filter for target state estimation.

Benefits of technology

The robustness and adaptability of radiation source tracking are improved under low signal-to-noise ratio conditions, error accumulation is reduced, and accurate joint detection and tracking of multiple radiation sources are achieved, making it suitable for complex scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118444306B_ABST
    Figure CN118444306B_ABST
Patent Text Reader

Abstract

This invention proposes a direct joint detection and tracking method for frequency-hopping radiation sources to address the problem of tracking multiple unknown targets, including when they appear and under unknown signal-to-noise ratios. The invention directly uses the received signals from multiple sensors as measurements, performs channelization processing on the received signals, and implements a filtering process using Bernoulli filters in different channels to achieve multi-target tracking. It also employs a likelihood function based on a fuzzy function within a lumped framework, eliminating the need to estimate intermediate parameters of the tracked targets and effectively avoiding error accumulation in traditional tracking methods that involve moving from signal to parameter to measurement. Under low signal-to-noise ratio conditions, this likelihood function effectively reduces errors caused by unknown signals, thus exhibiting good environmental adaptability and robustness.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of electronic countermeasures technology, specifically relating to a direct joint detection and tracking method for frequency hopping radiation sources with unknown signal-to-noise ratio. Background Technology

[0002] With the development of network, communication, and computer technologies, radiation source tracking methods have gradually evolved from data-level to signal-level, among which direct tracking methods for moving radiation sources based on raw signal echoes have received increasing attention. Traditional radiation source tracking is a two-step method. First, parameters of the target to be tracked are estimated from the received signal. Then, the estimated intermediate parameters are used as input, and appropriate filters are selected to process these parameters, thereby estimating the number and state of targets within the area to be detected. This method leads to estimation errors that cannot be corrected in the next stage, resulting in error accumulation and making it difficult to achieve real-time and effective target tracking in complex scenarios. Compared to the traditional two-step tracking method, direct tracking algorithms utilize all the information from the received signal, thus reducing the error accumulation of the two-step method.

[0003] In the single-source tracking problem, inspired by the fuzzy function method for extracting the time delay and Doppler information of the radiation source, a fuzzy function-based likelihood function implemented in a lumped framework is designed. Under low signal-to-noise ratio (SNR) conditions, this likelihood function can effectively reduce the error caused by unknown signals. For a single radiation source, a Bernoulli filter implemented using the particle approximation is adopted, which has low computational complexity and is easy to implement with the proposed likelihood function. In low SNR environments, this lumped filter can effectively estimate the target position, velocity, and transmission power.

[0004] For the problem of multiple frequency-hopping radiation sources distinguished on multiple station tracking frequency bands, based on the work of single radiation source tracking methods, the signals of multiple frequency-hopping radiation sources are channelized into single-frequency band signals. Inspired by the interactive multiple model concept, multiple single-frequency band signals are tracked, and the tracking results are comprehensively corrected to obtain the correct tracking results (number of targets and trajectory). Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a direct detection and tracking algorithm for frequency-hopping radiation sources. In detection and tracking, a likelihood function is constructed using a fuzzy function, and a likelihood function based on a class of Bernoulli filters using intensity as the measurement is designed and implemented centrally, outperforming distributed likelihood functions. Regarding the received signal, the frequency-hopping signal is channelized to obtain multiple frequency band measurement signals, which are then tracked. The combined tracking results provide more reliable target information. This direct tracking method avoids the intermediate errors of two-step tracking and the hard decision in traditional detection and tracking algorithms. It also improves the algorithm's adaptability to complex scenarios and robustness under low signal-to-noise ratio conditions, thus enhancing target tracking performance. This invention starts directly from the received data to solve the joint detection and tracking problem of multiple frequency-hopping radiation source targets under conditions of unknown target appearance time, unknown signal-to-noise ratio, and unknown frequency band differentiation. It exhibits good performance, environmental adaptability, and robustness, meeting engineering design requirements.

[0006] The technical solution adopted in this invention is:

[0007] This invention utilizes a signal containing time delay and Doppler frequency shift. The received signal is discretized and used as the input to a Bernoulli filter. Compared to the traditional two-step tracking method, the direct tracking algorithm effectively reduces the intermediate errors. Furthermore, the entire system is Bernoulli-based, and its information transmission is probabilistic, thus the proposed scheme exhibits strong robustness and scalability. Because the likelihood function utilizes all relevant information from each receiving station, target tracking can be achieved at a lower signal-to-noise ratio.

[0008] Assume a typical passive radiation source tracking scenario with a multi-channel receiving station M. r Let d be the position vector formed by the x-axis and y-axis of the j-th receiving station. j All receiving stations are stationary. There are multiple targets, and the target positions are p = [p x ,p y ] Τ The speed is The target state is a concatenation of position and velocity vectors. Assuming the observation duration is T, α j Let z be the complex channel coefficient of the signal on the j-th channel, then the received signal z at the j-th receiving station j (t) is

[0009]

[0010]

[0011] Where, n j (t) represents noise, t j The time delay on the channel is much smaller than T, s(tt)j ) represents the time-delayed transmitted signal, n j f is the Doppler frequency along the target path. c Where c is the carrier frequency and c is the speed of electromagnetic wave propagation. The signal is channelized in frequency, and each channel is demodulated to baseband and fed at the same rate f. s The sampled material has a length of L = f. s T is a discrete signal, z j Let represent the complex baseband signal sampled on the j-th frequency channel, then the received signal at time k is...

[0012]

[0013] Among them, s k ∈C L×1 It is the transmitted signal, the L-shot complex baseband signal. If [·] m,n Let it be denoted as the (m,n)th element of a vector or matrix. For simplicity, we assume that... Omit the time k. It is circular Gaussian white noise, with the following distribution: Among them 0 L and I L Let represent the zero vector of L×1 and the identity matrix of L×L, respectively. This represents the noise variance, defining that the noise is independent of time and the receiving station, i.e. in(·) H It is the sign of conjugate transpose, δ x It is the Kronecker-Dirac function. Therefore, we can... and Approximately equal to the transmitted signal power and signal-to-noise ratio; It is the delay-Doppler operator, which reflects the effect of delay t and Doppler shift n on the sampled signal of length L when it propagates to the j-th receiving station.

[0014]

[0015] Where, D(u)=diag{[e ι2πu·0 ,…,e ι2πu·(L-1) ]} is C L×L A diagonal matrix, diag{[b1,…,b Q ]} represents b1,…,b Q Let W be a diagonal matrix with diagonal elements, and W∈C. L×L This represents the unit Discrete Fourier Transform (DFT) matrix, where the (m,n)th element is...

[0016]

[0017] Define the stitching of measurements from all receiving stations as the complete measurement information.

[0018]

[0019] A method for direct detection and tracking of passive radiation sources based on a Bernoulli filter mainly includes implementing the Bernoulli filter using a particle-like approach to approximate its spatial probability density function, constructing a likelihood function using the ambiguity function of the received signal, and using soft decision instead of hard decision to extract the state and detect the presence of the target. Specifically, it includes steps S1-S10:

[0020] S1, Bernoulli filter uses a batch of weighted particles To approximate the spatial probability density function p of the Bernoulli distribution k (x), the approximate rule is: Initialization: When k = 0, the Bernoulli distribution parameters are π0 = {r0, p0(x)}, where r0 is the probability of the target's existence, and p0(x) is the corresponding spatial probability density function. (Using particles...) To approximate p0(x), N is the number of predicted particles, and B is the number of newly generated particles. The weight corresponding to each particle;

[0021] Define the total observation duration as K. When K≥k≥1, repeat steps S2-S10.

[0022] S2, Input the prior target Bernoulli distribution parameters and the measurement matrix z at the current time k First, a prediction is made: the prediction parameter is π. k|k-1 ={r k|k-1 ,p k|k-1 (x)}, where the predicted existence probability is calculated as

[0023] r k|k-1 =p b (1-r k-1 )+p s r k-1

[0024] Where r k-1 , and Let p represent the existence probability, particle weight, and particle state at time k-1, respectively. b Let p be the probability of a newborn. s The probability of survival. This forms the spatial probability density p. k|k-1 The particle (x) consists of a predicted part and a newly generated part:

[0025]

[0026]

[0027] in, It is the predicted probability density of the state. Let Q be a Gaussian function with mean m and variance P. k-1 Let b be the process noise covariance matrix. k (x;z k The known density of newborn particles is denoted as . The predicted particle weights are denoted as . They also need to be calculated separately.

[0028]

[0029] S3, Abbreviation for It is a function of the target state. The sum of samples from L snapshots of the received signal forms a complex Wishart distribution, which is considered to be the form of the final likelihood function. The complex Wishart distribution is a positive definite matrix random distribution in statistics. If a d×d dimensional matrix Y conforms to a complex Wishart distribution with δ degrees of freedom, its probability density function can be expressed as follows:

[0030]

[0031] Where V is a d×d positive definite symmetric matrix, and tr(·) represents the trace of the matrix. The complex Gamma function representing the variable d is defined as follows:

[0032]

[0033] and It is the Gamma function.

[0034] S4. For each particle (l = 1, ..., N+B), repeat steps S41-S43 to calculate the likelihood-related parameters.

[0035] S41. First, calculate the expectation of the ambiguity function of the v-th snapshot with respect to the j-th receiving station when there is no target. This expectation is a quantity independent of the snapshot and time.

[0036]

[0037] Calculate the expectation of the fuzzy function when there is a target.

[0038]

[0039] S42, the unknown matrix variables are

[0040]

[0041] If the fuzzy function is the sum of samples from L snapshots, then the likelihood when there is no target is:

[0042]

[0043] The likelihood function when there is a target is:

[0044]

[0045] S43. Calculate the likelihood ratio

[0046]

[0047] S5. Calculate the integral of the approximate likelihood ratio by summing the particle likelihood ratios calculated using S43.

[0048]

[0049] S6. Update: Based on the predicted Bernoulli parameters, and combined with the measurement vector z obtained at the current time step. k We can obtain the updated Bernoulli parameter π at time k. k|k ={r k|k ,p k|k (x)}. The probability of its existence is...

[0050]

[0051] S7. Spatial probability density is: For each particle (l=1,…,N+B), first update the particle weights using the likelihood calculation in S42 when there is a target: Then the particle weights are normalized:

[0052]

[0053] S8. Resampling: For l = 1, ..., N and m (l) ∈{1,…,N+B}, with The probability of selecting particles is determined by their weight; particles with higher weights have a higher chance of being selected. The resampled particle state is...

[0054]

[0055] S9. Reset the particle weights. l = 1, ..., N.

[0056] S10, Output N particles: r k , Used for iteration in the next time step, and for state extraction: Include the estimated position, velocity, and signal power of the target at time k, k = k + 1, and return to step S2.

[0057] The beneficial effects of this invention are as follows:

[0058] 1) This invention uses a Bernoulli filter to solve the tracking problem with signals as the measurement, and introduces typical radiation source tracking scenario modeling and a likelihood function based on fuzzy function to solve the problem of calculating the likelihood function when the transmitted signal and the transmitted power are unknown, thus providing a solution for radiation source tracking in more complex scenarios.

[0059] 2) This invention can solve the problem of joint detection and direct tracking of radiation sources, avoiding the intermediate error caused by the traditional two-step tracking method. This method has strong robustness and good performance under low signal-to-noise ratio. Attached Figure Description

[0060] Figure 1 The receiving station location, the target's true trajectory, and a first Monte Carlo plot when SNR = 0dB;

[0061] Figure 2 The specific tracking curves on the x and y axes when SNR = 0dB;

[0062] Figure 3 The curve showing the change of average OSPA with signal-to-noise ratio at SNR = -5.0dB;

[0063] Figure 4 The curve showing the average number of target estimates as a function of signal-to-noise ratio at SNR = -5.0dB.

[0064] Figure 5 The average OSPA of the average number of target estimates as a function of the signal-to-noise ratio at SNR = -5.0 dB;

[0065] Figure 6 This is a schematic diagram illustrating the detection effect of the present invention. Detailed Implementation

[0066] The practicality and effectiveness of this invention are demonstrated below with reference to the accompanying drawings and simulation examples;

[0067] Example

[0068] This example uses MATLAB to verify the above-mentioned joint detection and tracking algorithm for radiation sources. For simplicity, the following assumptions are made about the algorithm model:

[0069] The effectiveness of the present invention will be illustrated below with reference to the accompanying drawings and simulation examples.

[0070] Simulation conditions and parameters

[0071] Simulation Environment: For ease of explanation, consider a representative two-dimensional scenario, assuming 4 passive receiving stations d 1 …d 4Located at [190000,0], [230000,0], [270000,0], [310000,0] (m), respectively, to simplify the problem, we assume that the attenuation of each receiver is the same, i.e., the target state vector is... The initial position is determined, the velocity is unknown, and the total observation time is assumed to be K = 100 s. The linear Gaussian equation of motion for the target is x. k =Fx k-1 +ω k G among them

[0072]

[0073] G is a random number matrix, where ω k The process noise is Gaussian distributed. The survival probability p of each target. s =0.99. The measurement equation has been given above, and the transmitter's carrier frequency is f. c =10 7 Hz, sampling frequency f s =5×10 6 Hz, number of snapshots L = 5 × 10 5 c = 3 × 10 8 m / s. Targets are created at 0s and disappear at 100s. Targets originate from a fixed point; the target creation model is as follows:

[0074]

[0075] in,

[0076] m b =[x,0,y,0] T

[0077] P b =diag{[100,10,100,10]} 2

[0078] p b =0.01

[0079] Used to simulate m b Natural regeneration in the vicinity (x, y are the initial positions of the target). A total of 3000 particles were used, of which N = 2500 were surviving particles and B = 500 were newly generated particles. The number of Monte Carlo simulations was 100.

[0080] like Figures 1-3 As shown, even at a low signal-to-noise ratio (SNR = 0dB), the tracking trajectory and x and y axis tracking curves of the proposed algorithm are still quite close to the true values. Figures 4-5The average OSPA was plotted when SNR = 0dB. It can be seen that the algorithm's estimation error is within 400m, which is within a reasonable range for the monitored scenario. Figure 6 This demonstrates the algorithm's excellent detection performance; it consistently detects the target at both its appearance and disappearance points in Monte Carlo simulations. (Summary) Figure 1-5 The proposed algorithm is applicable to the joint detection and direct tracking of radiation sources, and it has strong robustness to low signal-to-noise ratios and strong adaptability to complex environments with unknown signal-to-noise ratios.

Claims

1. A method for direct joint detection and tracking of frequency-hopping radiation sources, characterized in that, In the scenario of radiation source tracking, there is a multi-channel receiving station M. r Let d be the position vector formed by the x-axis and y-axis of the j-th receiving station. j All receiving stations are stationary, and there are multiple targets. Let the target positions be p = [p x ,p y ] Τ The speed is The target state is a concatenation of position and velocity vectors. Define the observation duration as T, α j Let z be the complex channel coefficient of the signal on the j-th channel, and z be the received signal at the j-th receiving station. j (t) is: Among them, s(tt) j () represents the delayed transmission signal. ν j The time delay on the channel is much smaller than T,n j Let nj(t) be the Doppler frequency on the target path, nj(t) be the noise, and f be the frequency on the target path. c Where c is the carrier frequency and c is the speed of electromagnetic wave propagation. Frequency-hopping signals are channelized in frequency, and each channel is demodulated to baseband and fed at the same rate f. s The sampled material has a length of L = f. s T is a discrete signal, z j Let represent the complex baseband signal sampled on the j-th frequency channel. Then the received signal at time k is: Among them, s k ∈C L×1 It is the transmitted signal, the L-shot complex baseband signal, [·] m,n Let the (m,n)th element of the vector be denoted as . Omit the time k. It is circular Gaussian white noise, with the following distribution: Among them 0 L and I L Let represent the zero vector of L×1 and the identity matrix of L×L, respectively. This represents the noise variance, defining that the noise is independent of time and the receiving station, i.e. in(·) H It is the sign of conjugate transpose, δ x It is the Kronecker-Dirac function; therefore, and Approximately equal to the transmitted signal power and signal-to-noise ratio; It is the delay-Doppler operator, which reflects the delay τ and the Doppler frequency shift. ν The impact on the sampled signal of length L when propagated to the j-th receiving station: Where, D(u)=diag{[e ι2πu·0 ,…,e ι2πu·(L-1) ]} is C L×L A diagonal matrix, diag{[b1,…,b Q ]} represents b1,…,b Q Let W be a diagonal matrix with diagonal elements, and W∈C. L×L Let the unit Discrete Fourier Transform (DFT) matrix have the following element as its (m,n)th element: Define the stitching of measurements from all receiving stations as the complete measurement information. Because there are multiple targets, target matching must be performed first to determine the current channel signal's affiliation. The specific method is as follows: a. Regarding the measurement matrix z at the current time k The position and velocity information of multiple targets are sequentially substituted into the joint detection and direct tracking method to obtain the corresponding target existence probability r. k , in r k When r > 0.5, select r k Larger targets are considered potential targets; otherwise, there are no targets. b. If the total observation time T is 100s, and the target is detected, and the three consecutive suspected targets are the same target, then lock the target and track it; otherwise, repeat a. c. With the target locked, regarding the measurement matrix z at the current time... k Substitute the target's position and velocity information into the joint detection and direct tracking method. After the current channel signal completes tracking, switch to the next channel and return to step a. If different channel signals determine that the same target is the same target, combine the target state information obtained from different channel signals to form the target's final state information. The joint detection and direct tracking method includes: S1. Implement a Bernoulli filter using particles to approximate its spatial probability density function: The Bernoulli filter uses a batch of weighted particles. To approximate the spatial probability density function p of the Bernoulli distribution k (x), the approximate rule is: When k=0, the Bernoulli distribution parameters are initialized as π0={r0,p0(x)}, where r0 is the existence probability of the target and p0(x) is the corresponding spatial probability density function. (The last part, "using particles...", appears to be an error and doesn't need a direct translation.) To approximate p0(x), N is the number of predicted particles, and B is the number of newly generated particles. Define the weight for each particle; define the total observation time as K; S2. Determine if K≥k≥1 holds true. If not, end the tracking; otherwise, execute: Input the prior target Bernoulli distribution parameters. and the measurement matrix z at the current time k ; First, a prediction is made: the prediction parameter is π. k|k-1 ={r k|k-1 ,p k|k-1 (x)}, where the predicted existence probability is calculated as r k|k-1 =p b (1-r k-1 )+p s r k-1 Where r k-1 , and Let p represent the existence probability, particle weight, and particle state at time k-1, respectively. b Let p be the probability of a newborn. s The probability of survival; forming the spatial probability density p k|k-1 The particle (x) consists of a predicted part and a newly generated part: in, It is the predicted probability density of the state. Let Q be a Gaussian function with mean m and variance P. k-1 Let b be the process noise covariance matrix. k (x;z k ) represents the newborn density of known states; the predicted particle weights. Calculations need to be done separately. S3, Definition for It is a function of the target state; the sum of samples from L snapshots of the received signal forms a complex Wishart distribution, which is considered to be the final form of the likelihood function; the complex Wishart distribution is a positive definite matrix random distribution in statistics. If a d×d dimensional matrix Y conforms to a complex Wishart distribution with δ degrees of freedom, its probability density function is expressed as follows: Where V is a d×d positive definite symmetric matrix, and tr(·) represents the trace of the matrix. The complex Gamma function representing the variable d is defined as follows: and It is the Gamma function; S4. Calculate the likelihood-related parameters for each particle l, l = 1, ..., N+B: S41. First, calculate the expectation of the ambiguity function of the v-th snapshot with respect to the j-th receiving station when there is no target. This expectation is a quantity independent of the snapshot and time. Calculate the expectation of the fuzzy function when there is a target. S42, the unknown matrix variables are If the fuzzy function is the sum of samples from L snapshots, then the likelihood when there is no target is: The likelihood function when there is a target is: S43. Calculate the likelihood ratio S5. Using particles to find the integral of the approximate likelihood ratio. S6. Update: Based on the predicted Bernoulli parameters, and combined with the measurement vector z obtained at the current time step. k We can obtain the updated Bernoulli parameter π at time k. k|k ={r k|k ,p k|k ∑(x)} exists with probability . S7, Spatial probability density: For each particle, the likelihood update particle weights when there is a target are calculated using S42: Then the particle weights are normalized: S8. Resampling: For l = 1, ..., N and m (l) ∈{1,…,N+B}, with The probability of selecting particles is determined by their weight; particles with higher weights have a higher probability of being selected. The resampled particle state is... S9. Reset the particle weights. S10, Output N particles: r k , Used for iteration in the next time step, and for state extraction: Include the estimated position, velocity, and signal power of the target at time k, k = k + 1, and return to step S2.

Citation Information

Patent Citations

  • Random finite set based multi-target tracking method in outer radiation source radar

    CN104101876A

  • External radiation source radar high repetition frequency hopping signal coherent accumulation method

    CN113820678A