A radio interference source automatic positioning and troubleshooting system and method thereof

Abstract

The application provides a radio interference source automatic positioning and troubleshooting system and method thereof, the radio interference source automatic positioning and troubleshooting system comprises a front-end hardware enhancement module, a core algorithm optimization module and a system-level cooperative positioning module, the radio interference source automatic positioning and troubleshooting method comprises hardware deployment and preprocessing, spatial domain beamforming and subarray combination, digital domain multipath elimination and signal recovery and the like, the radio interference source automatic positioning and troubleshooting system and method thereof provided by the application effectively solve the problem that the traditional equipment cannot distinguish the direct wave from the reflected wave, so that the direction finding result is seriously deviated or even points to the wrong direction, and greatly improve the accuracy and reliability of the direct wave identification of the signal.

Description

Technical Field

[0001] This invention relates to the field of radio monitoring and positioning, and in particular to an automatic radio interference source location and investigation system and method. Background Technology

[0002] With the acceleration of urbanization, the urban electromagnetic environment is becoming increasingly complex. In modern urban areas filled with high-rise buildings, radio signals are easily affected by reflection, refraction, and diffraction during propagation, resulting in severe multipath effects. At the same time, high-density building clusters create significant signal shadow areas, making it difficult for ground-based and low-altitude monitoring equipment to effectively capture direct wave signals. This complex propagation environment poses a severe challenge to the location and investigation of radio interference sources. Technicians often have to rely on experience to conduct manual investigations from high vantage points or at specific times (such as midday off-peak hours), which is not only inefficient but also makes it difficult to guarantee the accuracy of the location.

[0003] Analysis reveals the following main problems with existing technologies for locating interference sources in complex urban environments: First, existing conventional monitoring stations mostly use a single omnidirectional antenna or a simple array antenna, with fixed beam pointing or omnidirectional reception, lacking adaptive spatial filtering capabilities. At the receiving end, the equipment cannot effectively distinguish between the direct wave of the signal and the strong reflected wave from the wall of a tall building at the physical level. Since the reflected wave usually has high energy, it is very easy to submerge or deflect the angle of arrival (DOA) estimation of the direct wave in the spatial domain, resulting in serious deviations in the direction finding results, or even pointing in a completely wrong direction. Secondly, traditional signal demodulation and parameter estimation algorithms perform poorly in urban multipath channels. On the one hand, multipath delay spread is prone to causing severe inter-symbol interference, which destroys the subcarrier orthogonality of broadband signals such as OFDM, leading to demodulation failure. On the other hand, in urban background noise, such as under non-Gaussian noise interference, conventional cross-correlation algorithms cannot effectively suppress the noise floor. Especially in scenarios where the signal is submerged in noise and the signal-to-noise ratio is low, existing algorithms are unable to accurately extract the weak time difference of arrival (TDOA) and frequency difference of arrival (FDOA) from the mixed signal, resulting in severe jitter in the time difference measurement value, which cannot meet the requirements of high-precision positioning. Finally, traditional monitoring stations often use a single detection station, which has a limited coverage area and is easily affected by building obstruction, resulting in monitoring blind spots. For non-line-of-sight (NLOS) propagation paths, the traditional TDOA positioning algorithm assumes that the signal propagates in a straight line, ignoring the additional path delay caused by diffraction and reflection. This leads to the positioning solution deviating from the actual location by hundreds of meters or even more, i.e., geometric position deviation. In addition, fixed monitoring stations cannot flexibly move to obstructed areas for close reconnaissance and lack the ability to use mobile platforms, such as drones or monitoring vehicles, for synthetic aperture imaging or multi-dimensional perspective blind spot filling. It is difficult to achieve full coverage and accurate locking of interference sources in three-dimensional space.

[0004] Therefore, it is necessary to provide a new automatic radio interference source location and investigation system and method to solve the above-mentioned technical problems. Summary of the Invention

[0005] To address the technical problems of traditional monitoring stations using single antennas or simple omnidirectional antennas, which cannot distinguish between direct and reflected waves, are easily biased by multipath signals reflected from tall buildings, are prone to inter-symbol interference (ISI) under strong multipath conditions, cannot accurately calculate time difference of arrival (TDOA) under low signal-to-noise ratio (SNR) conditions (i.e., when submerged in noise), and are greatly affected by shadow areas in single-station direction finding; TDOA positioning has huge errors in non-line-of-sight (NLOS) environments, even with deviations of hundreds of meters; and fixed stations cannot penetrate the blind spots of tall buildings, this invention provides an automatic radio interference source location and investigation system and method.

[0006] The automatic radio interference source location and investigation system provided by this invention includes: a front-end hardware enhancement module for suppressing multipath effects and noise at the signal receiving end, the module comprising: The anti-multipath antenna array unit consists of multiple antenna elements deployed according to a preset geometric structure, used to collect space signals; An adaptive beamforming processing unit, connected to the anti-multipath antenna array unit, is used to run the minimum variance distortionless response (MVDR) or linear constrained minimum variance (LCMV) algorithm, calculates the weighting coefficients in real time based on the signal angle of arrival (DOA), forms a main beam pointing towards the direction of the interference source, and forms nulls in the multipath reflection direction. The spatial diversity receiving unit, through multiple independent receiving antennas deployed in different spatial locations and the maximum ratio combining (MRC) circuit, is used to combat fast signal fading. The high-performance filtering unit uses a bandpass filter and a choke coil structure to resist multipath antennas, thereby filtering out out-of-band interference and suppressing low-angle reflected waves. The core algorithm optimization module is used to perform multipath separation and demodulation of the received signal in the digital domain. This module includes: The OFDM cyclic prefix processing unit is used to remove the cyclic prefix and perform FFT transformation to convert the time-domain multipath channel into a frequency-domain flat fading channel, and perform channel equalization based on the pilot signal. The Rake multipath merging unit is used to search for and identify multiple multipath components with energy exceeding a threshold on the time axis, and assign independent branches for phase alignment and weighted merging. The specific operation steps of the Rake multipath merging unit are as follows: Step 1: Multipath Component Search and Identification: First, the system performs energy detection on the received baseband signal and sets a dynamic threshold, typically 3-5 times the average noise power. Then, a correlation window is slid across the time axis to calculate the cross-correlation function between the received signal and the local pseudo-random code (PN code). When the correlation peak exceeds the set threshold, a valid multipath component is identified, and its arrival time and delay are recorded. and initial phase; Step 2, Branch Assignment: Assign an independent demodulation branch to each identified valid multipath component. Each branch contains an independent delay line, phase rotator, and weighted coefficient multiplier. Step 3, Phase Alignment: Using pilot symbols or known training sequences, estimate the phase offset of each branch. An opposite phase rotation factor is generated by a numerically controlled oscillator (NCO). Phase compensation is performed on the branch signal to ensure that the signal phases of all branches are consistent before merging, i.e., coherent alignment. Step 4: Weighted Combining: Calculate the weighting coefficients based on the instantaneous signal-to-noise ratio (SNR) of each branch. The signals from all branches, after phase alignment and weighting, are vector summed. This concentrates the multipath energy that was dispersed at different time points, thereby improving the demodulation signal-to-noise ratio.

[0007] Furthermore, usually It is proportional to SNR, i.e., it follows the maximum ratio merging MRC criterion; Furthermore, maximum ratio combining (MRC) is a commonly used receiver diversity technique widely applied in wireless communication systems to improve the signal-to-noise ratio (SNR) of the received signal and system reliability. In multi-antenna receiving scenarios, the same signal will reach the receiver through different paths, and the fading and noise of each path are different. The core idea of ​​MRC is to weight the signal of each receiving antenna so that the path with the highest SNR has the largest weight, and then add the weighted signals together to obtain the optimal combining effect. The spread spectrum despreading and channel coding / decoding unit is used to perform correlation despreading using pseudo-random codes to concentrate signal energy, and to perform error correction decoding and deinterleaving using Turbo codes or LDPC codes. The high-precision time-frequency estimation unit uses high-order statistical analysis to establish signal models for direct and reflected waves. It employs a two-dimensional mutual ambiguity function (CAF) algorithm based on fourth-order cumulants to accurately estimate the time difference of arrival (TDOA) and frequency difference (FDOA) in a noisy environment. The method for establishing the signal models for direct waves and reflected waves is as follows: Assume the signal received by the receiving antenna array has L paths, consisting of a linear superposition of 1 direct wave and L-1 reflected or diffracted waves, plus Gaussian white noise; The mathematical expression is: ) Where s(t) is the baseband signal of the transmitter, which is a known PN code or OFDM symbol; Let be the complex attenuation coefficient for the i-th path, including amplitude attenuation and phase rotation; Let i be the arrival time difference of the i-th path relative to the reference point; It is the Doppler frequency shift of the i-th path; It is additive white Gaussian noise; At this point, the continuous-time model is discretized and sampled to obtain a discrete-time series. Gaussian noise is eliminated by using higher-order statistics, specifically fourth-order cumulants, to calculate cumulant slices of the received signal. Due to the influence of Gaussian noise, since the fourth-order cumulant of Gaussian noise is zero, the set of parameters that minimizes the error between the signal model output of the direct wave and the reflected wave and the actual received signal is obtained by maximizing the likelihood function. ); A system-level collaborative positioning module is used to achieve multi-dimensional stereo coverage and precise positioning. The module includes: The multi-station collaborative positioning unit is used to control at least three monitoring stations to synchronize their time, collect the time difference of arrival (TDOA) or frequency difference (FDOA) data of each station, and construct a system of equations for hyperboloid and ellipsoid to solve for the coordinates of the interference source. The specific steps for constructing and simultaneously solving the equations of the hyperboloid and ellipsoid to obtain the coordinates of the interference source are as follows: Step 1: Time Synchronization and Data Acquisition: The master station sends synchronization pulses to the slave stations via GPS / BeiDou satellite time synchronization, and each station records the arrival time of interference signals. and Doppler shift ; Step 2: Constructing the system of equations: TDOA Equation (Hyperboloid): Taking the master station Station 0 as a reference, for any slave station... ,have: =

[0008] in, Coordinates of the interference source Let c be the coordinates of the monitoring station, and c be the speed of light. The time difference for measurement; FDOA equations (ellipsoidal / hyperboloid): Equations are constructed using frequency shifts, resulting in: =

[0009] in,( () represents the velocity vector of the interference source; if stationary, it is 0. Wavelength; Step 3: Solve the simultaneous equations: Linearize the nonlinear system of TDOA and FDOA equations, perform Taylor expansion, and construct the matrix form. ; At this point, the Chan algorithm can be used: it utilizes the weighted least squares (WLS) method twice to solve for the intermediate variables first, and then substitutes them to solve for the final coordinates, which has a fast calculation speed; Alternatively, Taylor series expansion can be used: set an initial guess value ( ), through iterative correction ( The solution is gradually approximated until the residual is less than a threshold. Results: Output the latitude and longitude (x, y), altitude (z) of the interference source, and the error ellipse parameters (representing the confidence interval of the positioning accuracy). A motorized synthetic aperture unit is used to control a drone or monitoring vehicle to move along a predetermined trajectory, collect coherent sampling data from multiple locations, and synthesize a large-aperture virtual array to improve angular resolution. The synthesis process of the large-aperture virtual array is as follows: Step 1: Trajectory Planning and Data Acquisition: Control the drone or monitoring vehicle to move at a constant speed along a predetermined trajectory (such as an S-shape or a straight line), and collect data at N discrete sampling points along the trajectory. At this location, the precise coordinates of each point are recorded using a high-precision POS (Position and Attitude System). Simultaneously, it receives and stores baseband IQ sampling data at each point, including attitude angles. ; Step 2, Phase Center Position Calibration: Since the phase center of the antenna does not coincide with the phase center of the GPS antenna, lever arm compensation is performed to unify the phase reference points of all sampled data to the same spatial point, usually the trajectory center or starting point. Step 3: Coherent integration and virtual array synthesis: Phase compensation: for the target point ( ), calculate the theoretical phase delay of the signal propagating from this point to the kth sampling point: ,in Slope distance; Coherent superposition: Multiplying the received signal at the k-th point by the compensated phase. Then sum the signals from all N points:

[0010] in, The synthesized baseband signal is equivalent to the signal received by a giant antenna array of length L at a certain moment; k is the sampling point index, representing the k-th discrete sampling point on the trajectory, k=1,2,...,N; N is the total number of sampling points, representing the number of samples collected by the mobile platform along the entire trajectory. The larger N is, the larger the synthesized array aperture and the higher the resolution. This represents the original received signal at point k, i.e., the position of the mobile platform. At this point, the baseband IQ sampling data actually received by the antenna includes amplitude and phase information; j is a mathematical constant used to represent the phase of the signal, satisfying the following conditions: ; The wavelength of the signal, that is, the wavelength of the radio wave. where c is the speed of light and f is the signal frequency; The slant range indicates the distance from the target interference source ( ) to the kth sampling point The straight-line distance between them; This is the phase compensation factor; This process is equivalent to deploying a giant linear array L with a length of trajectory length L in space, which is much larger than the physical antenna aperture, thus forming a virtual array; At this time, the width of the synthesized beam , Indicates the synthesized beamwidth, which refers to the angular width of the main lobe of the synthesized antenna array; L is the virtual array aperture length, which refers to the total length of the moving platform trajectory; For example, the larger the L value of the straight-line distance of a drone's flight or the unfolded length of its S-shaped trajectory, the narrower the beam. Advantage: Moving a small antenna a certain distance is equivalent to building a huge antenna of length L; It is worth noting that the width of the synthesized beam It is much smaller than the beamwidth of a physical antenna, thus enabling the differentiation of interference sources in the main lobe direction from nearby reflected waves. The map-assisted correction unit is used to load a high-precision 3D city model, perform line-of-sight (LOS) analysis to identify non-line-of-sight (NLOS) paths, and introduce positive time delay deviation correction or reduce the weight of NLOS paths in the solution. The fingerprint matching unit is used to store and retrieve the pre-collected radio fingerprint database, and match real-time signal features with the fingerprint database to assist in positioning.

[0011] Specifically, the adaptive beamforming processing unit operates as follows: Real-time monitoring of the noise covariance matrix in a noisy environment; Linear constraints are set based on the approximate DOA of the interference source; The weighted vector is updated through an iterative algorithm to maximize the output signal-to-noise ratio while suppressing reflected signals from the sides of tall buildings and the ground. The setting of the linear constraint conditions includes the following steps: First, let the approximate direction of the interference source be... The direction of multipath reflection is The guide vector is , The array response in this direction is represented, and the constraint equations are established as follows: Perform main lobe constraint to maintain gain: ensure that the gain remains constant in the direction of the interference source. The gain is 1 (or a constant) ):

[0012] in, H represents the conjugate transpose of the weight vector, where H stands for Hermitian transpose, i.e., transpose first and then conjugate, used to calculate power or inner product; The steering vector pointing in the direction of the interference source; To suppress interference, null constraints are applied in the multipath reflection direction. The gain is 0:

[0013] in, The steering vector indicating the direction of interference; Apply derivative constraints: To make the main lobe more robust, The first derivative is set to 0 at this point, denoted as a flat main lobe:

[0014] Under the constraints of main lobe, null, and derivative, minimizing the array output power, i.e., minimizing noise and residual interference, is as follows: in, For minimization operation, Let covariance matrix be the variance matrix. Let C be the array output power, C be the constraint matrix, and f be the response vector. The optimal weight vector is solved using the Lagrange multiplier method. ,and The minimization operation must satisfy the following conditions: It is carried out under constraints.

[0015] Specifically, in the OFDM cyclic prefix processing unit, the length of the cyclic prefix is ​​configured to be greater than the maximum delay spread of the target investigation area channel, and is adjustable in the range of 2μs to 10μs.

[0016] Specifically, the high-precision time-frequency estimation unit uses high-order statistical analysis and takes advantage of the insensitivity of the fourth-order cumulant to Gaussian noise to suppress spatial correlation clutter under low signal-to-noise ratio and accurately extract TDOA and FDOA parameters. Specifically, the operation of the motorized synthetic aperture unit is as follows: Plan S-shaped or circular flight paths; The phase center position of each sampling point is recorded using a high-precision POS system; Phase compensation and coherent superposition are performed on data collected from different locations to form an equivalent giant antenna array with a size equal to the trajectory length; Specifically, in the anti-multipath antenna array unit, the preset geometric structure includes a uniform linear array or a circular array; An automatic method for locating and troubleshooting radio interference sources includes the following steps: S1. Hardware Deployment and Preprocessing: First, deploy the monitoring station network in the investigation area and load the three-dimensional building model. Start the anti-multipath antenna array and spatial diversity antenna. Use the front-end bandpass filter and choke structure to filter out out-of-band noise and suppress low-angle multipath reflections to complete hardware initialization. S2. Spatial Beamforming and Diversity Combining: Based on the antenna array signal deployed in S1, an omnidirectional scan is performed to estimate the coarse DOA of the interference signal. The MVDR / LCMV algorithm is run to calculate the weighting coefficients, synthesize a highly directional main beam to filter out multipath interference, and form nulls in the multipath reflection direction. The multipath signals of spatial diversity are combined using the maximum ratio combining (MRC) algorithm to perform signal-to-noise ratio weighted combining and output a high signal-to-noise ratio synthesized signal. S3, Digital Domain Multipath Cancellation and Signal Recovery: The high signal-to-noise ratio synthesized signal output from S2 is demodulated using orthogonal frequency division multiplexing, cyclic prefix is ​​removed, and fast Fourier transform is performed. The channel frequency response is estimated using the pilot signal, and single-tap equalization is performed in the frequency domain to transform the flat fading channel. The Rake receiver is started to search for multipath components with energy exceeding the threshold on the time axis, and phase alignment and weighted combining are performed. Finally, spread spectrum despreading and channel encoding and decoding are performed to correct burst errors and recover the original data. The specific steps for estimating the channel frequency response using pilot signals and performing single-tap equalization in the frequency domain are as follows: FFT Transform: After OFDM demodulation, the time-domain signal... After FFT transformation to the frequency domain, we obtain ; Where k is the subcarrier index; Channel estimation: Extracting pilot subcarrier data in the frequency domain, as well as known transmitted pilot values. The received value is Calculate the channel frequency response estimate:

[0017] For data subcarriers, via pilot points Perform linear interpolation or spline interpolation to obtain channel estimates for all subcarriers. ; Single-tap equalization: This involves directly performing complex division on each subcarrier in the frequency domain, i.e., a single-tap filter, as shown in the following formula:

[0018] IFFT: Converts the equalized frequency domain signal... Perform an IFFT transform to recover the original symbols in the time domain; Principle: OFDM decomposes a broadband frequency selective fading channel into multiple parallel narrowband flat fading sub-channels. Therefore, each sub-channel only needs a complex coefficient or a single tap to fully compensate for channel distortion.

[0019] S4. High-precision joint estimation of time and frequency parameters: Establish a multipath signal model including direct wave and reflected wave, input the digital signal recovered in S3 and the three-dimensional building model loaded in S1, construct a two-dimensional mutual ambiguity function CAF and perform two-dimensional search in time delay and frequency offset, so as to find the maximum concentration point of signal energy in both time delay and Doppler frequency dimensions, achieve joint optimal estimation, use the fourth-order cumulant algorithm to suppress non-Gaussian noise, search for TDOA / FDOA peak, and output accurate time difference of arrival (TDOA) and frequency difference of arrival (FDOA) values; S5. Multi-station collaborative calculation and preliminary positioning: Based on the TDOA / FDOA data extracted in S4, the synchronization time of each monitoring station in S1, and the coordinates of the monitoring stations, each monitoring station achieves time synchronization through GPS / BeiDou. The central server collects TDOA / FDOA data and solves the hyperboloid and ellipsoid equations simultaneously. It uses the Chan algorithm or Taylor series expansion method to solve the two-dimensional coordinates and error ellipse of the interference source, outputs the two-dimensional latitude and longitude coordinates and error ellipse of the interference source, and evaluates the reliability of the positioning results to determine whether it falls into the shadow area or the error is too large. If the result falls into the signal shadow area or the error exceeds the threshold, S6 is triggered to fill the blind spot. The credibility assessment includes the following steps: Geometric accuracy factor, i.e., GDOP / DOP analysis: First, calculate the condition number of the geometric configuration matrix formed by the monitoring station and the interference source. If the DOP value is too large, such as DOP value > 6, it indicates that the distribution of the monitoring stations is unreasonable, such as collinearity, and the positioning results are unreliable. Residual analysis: Substitute the solved coordinates back into the TDOA / FDOA equation system to calculate the difference between the theoretical and measured values, i.e., calculate the residuals.

[0020] If the root mean square residual (RMS) exceeds the preset threshold, such as a distance error > 50 meters, the solution is considered to have failed or been severely affected by NLOS. Shadow Area / NLOS Judgment: Based on the 3D building model, determine whether the line connecting the candidate interference source point and the monitoring station passes through the building. If it is determined to be NLOS (non-line-of-sight), the result is directly marked as low confidence. Multi-station consistency test: Compare the results calculated from different combinations of monitoring stations, such as station 1-2-3 and station 1-2-4. If the difference between the two sets of results exceeds the threshold, it indicates that there is a gross error and the result is unreliable.

[0021] It is worth noting that the decision-making logic is as follows: If DOP is normal AND residual is small AND LOS check passes → reliable, output the result; If the DOP is large OR the residual is large OR the NLOS judgment is unreliable, trigger S6 to perform a maneuver to fill in the blind spot; S6. Mobile blind spot filling and synthetic aperture imaging: When the initial positioning result is in the shadow area or the error is too large, dispatch the UAV or monitoring vehicle to the suspicious area determined by S5, perform multi-point coherent sampling along the predetermined trajectory and record the position and attitude data, perform phase center position calibration on the discrete sampling points, compensate the received signals of each point to the same reference point phase and then perform coherent integration processing to form a synthetic array pattern with a narrow beamwidth to penetrate the obstruction and accurately determine the direction of the interference source, output high-precision direction information or corrected coordinates, and input the more accurate direction information or coordinates obtained by S6, as well as the three-dimensional building model of S1, into S7 for final correction; S7. Map-Assisted Correction and Fingerprint Matching: Based on a 3D city model, line-of-sight analysis is performed to identify non-line-of-sight (NLOS) paths between monitoring stations and candidate interference source locations. The length difference between the direct path and the diffraction model path is calculated and used as the positive time delay deviation correction TDOA measurement value. Alternatively, the path can be assigned a low weight in the solution. The multipath delay spectrum and field strength distribution characteristics of the real-time signal are extracted and matched with a pre-stored radio fingerprint database using KNN or SVM. Finally, the accurate 3D coordinates of the interference source are output, including floor information and optimal investigation path guidance.

[0022] Specifically, the NLOS error correction in step S7 includes: Calculate the direct path length between the monitoring station and the candidate location. ; Calculate the shortest reflection / diffraction path length based on the building diffraction model. ; The difference between the two Convert to time deviation ; Will The positive correction bias is added to the TDOA measurement or the NLOS measurement is directly removed.

[0023] Specifically, the pore size processing in step S6 includes: Plan S-shaped or circular flight / driving trajectories; A high-precision POS system is used to record the phase center positions of N discrete sampling points on the trajectory. ); The received signal at the kth sampling point Perform phase compensation And perform coherent integral summation: ; in Slope distance; An equivalent giant antenna array with a size equal to the trajectory length L is formed, and a synthesized narrow beam is used to distinguish the main lobe from the adjacent reflected lobe.

[0024] Compared with related technologies, the automatic radio interference source location and investigation system and method provided by the present invention have the following beneficial effects: 1. This invention utilizes an anti-multipath antenna array and an adaptive beamforming algorithm. The system can form a high-gain main beam in the direction of the interference source in real time, while forming a deep null in the direction of multipath reflection. This is equivalent to actively filtering out strong reflected waves from the walls of tall buildings in physical space, effectively avoiding the direction finding results being deflected by false reflected signals, significantly improving the accuracy of DOA estimation in complex environments. Furthermore, through spatial diversity reception and maximum ratio combining (MRC) technology, it ensures that even if some antennas are blocked or in deep fading points, the system can still recover a high signal-to-noise ratio signal. Combined with the antenna physical structure with chokes and smothering plates, low-angle multipath reflections are suppressed from the source, effectively solving the problem that traditional equipment cannot distinguish between direct waves and reflected waves, resulting in serious deviations in direction finding results or even pointing in the wrong direction. This greatly improves the accuracy and reliability of direct wave identification. 2. This invention utilizes OFDM cyclic prefix technology to limit multipath delay within the guard interval. Combined with a Rake receiver, multipath components are converted from interference to gain and combined, effectively eliminating inter-symbol interference and ensuring reliable demodulation of broadband signals in complex reflection environments. Furthermore, it employs a two-dimensional mutual ambiguity function (CAF) algorithm based on fourth-order cumulants, which utilizes the difference in higher-order statistical characteristics between the signal and Gaussian background noise. Even in low signal-to-noise ratio scenarios where the signal is submerged by noise, it can still accurately extract the time difference of arrival (TDOA) and frequency difference (FDOA). 3. In this invention, by establishing a system-level collaborative positioning module, multi-station collaborative TDOA / FDOA joint positioning is utilized, combined with the mobile synthetic aperture technology of UAVs or monitoring vehicles. The system can not only utilize fixed stations, but also actively fill in the blind spots of high-rise buildings. The synthetic aperture technology is equivalent to turning a small antenna into a large aperture array, which greatly improves the angular resolution and realizes the positioning of hidden interference sources. Furthermore, a high-precision three-dimensional city model is introduced for line-of-sight analysis, automatically identifying non-line-of-sight (NLOS) paths and introducing positive time delay deviation correction or weight reduction processing, which effectively corrects the positioning error caused by signal diffraction / reflection. At the same time, a radio fingerprint database is used for feature matching to assist in confirming floors or even specific rooms, thereby improving the positioning accuracy from the street level to the building or room level. Attached Figure Description

[0025] Figure 1 System block diagram of the automatic radio interference source location and investigation system provided by the present invention; Figure 2 This is a flowchart illustrating the automatic location and investigation method for radio interference sources provided by the present invention. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0027] Please refer to the following: Figures 1 to 2 ,in, Figure 1 System block diagram of the automatic radio interference source location and investigation system provided by the present invention; Figure 2 This is a flowchart illustrating the automatic location and investigation method for radio interference sources provided by the present invention.

[0028] In some embodiments, such as Figure 1 As shown, an automatic radio interference source location and investigation system includes a front-end hardware enhancement module for suppressing multipath effects and noise at the signal receiving end. The module includes: The anti-multipath antenna array unit consists of multiple antenna elements deployed according to a preset geometric structure, used to collect space signals; The adaptive beamforming processing unit, connected to the anti-multipath antenna array unit, is used to run the minimum variance distortionless response (MVDR) or linear constrained minimum variance (LCMV) algorithm. It calculates the weighting coefficients in real time based on the signal angle of arrival (DOA) to form a main beam pointing towards the interference source and to form nulls in the multipath reflection direction. The spatial diversity receiving unit, through multiple independent receiving antennas deployed in different spatial locations and the maximum ratio combining (MRC) circuit, is used to combat fast signal fading. The high-performance filtering unit uses a bandpass filter and a choke coil structure to resist multipath antennas, thereby filtering out out-of-band interference and suppressing low-angle reflected waves. The core algorithm optimization module is used for multipath separation and demodulation of the received signal in the digital domain. The module includes: The OFDM cyclic prefix processing unit is used to remove the cyclic prefix and perform FFT transformation to convert the time-domain multipath channel into a frequency-domain flat fading channel, and perform channel equalization based on the pilot signal. The Rake multipath merging unit is used to search and identify multiple multipath components whose energy exceeds the dynamic threshold on the time axis, i.e., identify multiple multipath components with an average noise power of 3-5 times. It assigns an independent branch to each component, estimates the phase offset of each branch using pilot symbols, generates a phase rotation factor through a numerically controlled oscillator for coherent alignment, and calculates weighting coefficients based on the instantaneous signal-to-noise ratio (SNR) of each branch for vector summation to merge multipath energy. The spread spectrum despreading and channel coding / decoding unit is used to perform correlation despreading using pseudo-random codes to concentrate signal energy, and to perform error correction decoding and deinterleaving using Turbo codes or LDPC codes. The high-precision time-frequency estimation unit uses high-order statistical analysis to establish signal models for direct and reflected waves. It employs a two-dimensional mutual ambiguity function (CAF) algorithm based on fourth-order cumulants to accurately estimate the time difference of arrival (TDOA) and frequency difference (FDOA) in a noisy environment. A system-level collaborative positioning module is used to achieve multi-dimensional stereo coverage and precise positioning. The module includes: The multi-station collaborative positioning unit is used to control at least three monitoring stations to synchronize their time, collect the time difference of arrival (TDOA) or frequency difference (FDOA) data of each station, and construct a system of equations for hyperboloid and ellipsoid to solve for the coordinates of the interference source. The mobile synthetic aperture unit is used to control UAVs or monitoring vehicles to move along a predetermined S-shaped or circular trajectory. It uses a high-precision POS system to record the phase center position of each sampling point. The phase center position is calibrated by Lever Arm compensation for discrete sampling points. After compensating the received signals of each point to the same reference point phase, coherent integration is performed to form a giant antenna array with a size equal to the trajectory length, and synthesize an array pattern with a narrow beamwidth. The map-assisted correction unit is used to load a high-precision 3D city model, perform line-of-sight (LOS) analysis to identify non-line-of-sight (NLOS) paths, and introduce positive time delay deviation correction or reduce the weight of NLOS paths in the solution. The fingerprint matching unit is used to store and retrieve the pre-collected radio fingerprint database, and match real-time signal features with the fingerprint database to assist in positioning.

[0029] The specific operation of the adaptive beamforming processing unit is as follows: Real-time monitoring of the noise covariance matrix in a noisy environment; Linear constraints are set based on the approximate DOA of the interference source; The weighted vector is updated through an iterative algorithm to maximize the output signal-to-noise ratio while suppressing reflected signals from the sides of tall buildings and the ground. Setting linear constraints involves the following steps: First, let the approximate direction of the interference source be... The direction of multipath reflection is The guide vector is , The array response in this direction is represented, and the constraint equations are established as follows: Perform main lobe constraint (gain preservation): ensure that the main lobe is confined in the direction of the interference source. The gain is 1 (or a constant) ):

[0030] in, H represents the conjugate transpose of the weight vector, where H stands for Hermitian transpose, i.e., transpose first and then conjugate, used to calculate power or inner product; The steering vector pointing in the direction of the interference source; Perform null constraint (interference suppression): force it in the multipath reflection direction The gain is 0:

[0031] in, The steering vector indicating the direction of interference; Apply derivative constraints: To make the main lobe more robust, in The first derivative is set to 0 at this point, denoted as a flat main lobe:

[0032] Under the constraints of main lobe, null, and derivative, minimizing the array output power, i.e., minimizing noise and residual interference, is as follows: in, For minimization operation, Let covariance matrix be the variance matrix. Let C be the array output power, C be the constraint matrix, and f be the response vector. The optimal weight vector is solved using the Lagrange multiplier method. ,and The minimization operation must satisfy the following conditions: It is carried out under constraints.

[0033] In the OFDM cyclic prefix processing unit, the length of the cyclic prefix is ​​configured to be greater than the maximum delay spread of the target investigation area channel, and is adjustable in the range of 2μs to 10μs.

[0034] The high-precision time-frequency estimation unit uses high-order statistical analysis and takes advantage of the insensitivity of fourth-order cumulants to Gaussian noise to suppress spatial correlation clutter under low signal-to-noise ratio and accurately extract TDOA and FDOA parameters.

[0035] In anti-multipath antenna array elements, the preset geometry includes a uniform linear array or a circular array; Example 1 8-element uniform linear array UCA anti-multipath deployment This embodiment is applicable to scenarios where the monitoring station is fixedly deployed on a rooftop or high place, and the direction of the main interference source is roughly known, such as a scenario facing the center of the city. First, the hardware structure and geometric parameters are as follows: Operating frequency band: 1000 MHz, corresponding wavelength ; Array element spacing: It is worth noting that This spacing can prevent the generation of grid lobes while ensuring spatial resolution; Array coordinates: with the array center as the origin, arranged along the X-axis.

[0036] Oscillator coordinates: Where n = 1, 2, ..., 8, and the unit is units; ,in, These are the coordinates of the nth oscillator.

[0037] The specific coordinates are:

[0038] Antenna type: Single omnidirectional dipole antenna, gain 3dBi High-performance filtering unit: Bandpass filter: center frequency 1000MHz, bandwidth 20MHz, out-of-band rejection greater than 40dB; Choke structure: A metal reflector with a diameter of 0.6m is loaded at the bottom of the antenna to suppress ground reflected waves with an elevation angle of less than 10°.

[0039] Secondly, signal processing is performed: The scenario is set as follows: Direction of the target interference source: = , is the azimuth angle; The direction of the strong reflected wave is from the building on the left: =- .

[0040] Background noise: Gaussian white noise, power =1.

[0041] Step 1: Signal Acquisition and Down-Conversion Eight antenna elements simultaneously receive signals, resulting in eight analog radio frequency signals. After low noise Amplified by LNA and down-converted to a baseband signal That is, a single-channel signal.

[0042] Step 2: Covariance Matrix Estimation Collect K=1000 snapshots and calculate the covariance matrix of the received signal.

[0043]

[0044] in, , is the snapshot vector, which refers to the column vector composed of the signals received by the 8 antennas at a certain time k; K is the number of snapshots, which refers to the number of time samples collected. Here, K=1000 means that 1000 consecutive time points were taken for statistics. The purpose of taking 1000 is because noise is random. The more samples there are, the more accurate the statistical average will be, and the more stable the calculated matrix will be. k is the time index, referring to the k-th time moment or the k-th sample, and is a loop variable; Assuming the interference source power is 10W and the reflected wave power is 5W, the calculated... It contains information about signals, interference, and noise; Step 3: Calculation of MVDR weighting coefficients Calculate the weight vector based on the Minimum Variance Distortionless Response (MVDR) criterion. ;

[0045] in, As the guide vector,

[0046] Substitute, = , The calculation yields:

[0047] in, The optimal weight vector; For guiding vector; To perform a conjugate transpose on the guiding vector; The normalized spacing is 0.5, which is half a wavelength. Step 4: Forming the main beam and null The calculated complex weight coefficients Applying this to the signals in each branch, we obtain: exist With the direction and array gain kept at 1, there is no distortion response. exist- Direction, due to It contains strong interference information in that direction, and the algorithm automatically forms a deep null in that direction, at which point the gain is close to 0, suppressing reflected waves; Step 5: Spatial set division and MRC Output signal after beamforming The signals are fed into the Rake multipath combining unit. Assuming the Rake receiver identifies three multipath components, their instantaneous signal-to-noise ratios (SNRs) are as follows: Path 1 (Direct): SNR = 10 dB Path 2 (Reflection): SNR = 3 dB Path 3 (diffraction): SNR = -2 dB According to the Maximum Ratio Combination (MRC) criterion, the weighting coefficients Proportional to SNR:

[0048]

[0049]

[0050] Final output signal The signal-to-noise ratio is improved by approximately 11 dB.

[0051] Example 2 7-element uniform circular array UCA omnidirectional monitoring deployment This embodiment is applicable to monitoring vehicle or drone payloads that require 360° omnidirectional monitoring and suppression of ground multipath. First, the hardware structure and geometric parameters are as follows: Operating frequency band: 2400 MHz, corresponding wavelength ; Array radius: It is worth noting that the small radius is convenient for vehicle mounting, and the circumferential phase difference is used to determine the direction. Array element distribution: The oscillators are evenly distributed on the circumference of a circle in the XY plane, with a central point, totaling 8 points, or only 7 points on the circumference. This example uses a 7-point circumference layout.

[0052] Angle and position: n = 0, 1, ..., 6; coordinate: ; in, Let be the azimuth angle of the nth array element, referring to the angular position of the array element on the circumference, the angle of counterclockwise rotation from the positive X-axis. In a uniform circular array, the angular interval is . N is the number of array elements; Let X be the X coordinate of the nth element. Let Y be the Y-coordinate of the nth element. High-performance filtering unit: It employs a cavity filter with a center frequency of 2.4 GHz, high Q value, and high rectangular coefficient.

[0053] Absorbing materials are added to the bottom of the antenna to reduce the carrier, such as reducing coupling reflections from the roof of the monitoring vehicle; Secondly, signal processing is performed: The scenario is set as follows: The interference source is located at the azimuth angle: = Angle of elevation That is, low altitude; The ground reflected wave is located at the azimuth angle: = Angle of elevation That is, a mirror image; Since circular arrays are insensitive to elevation angles, they mainly suppress azimuth interference through nulls or by utilizing polarization characteristics. Here, we assume that the symmetry of the circular array is used to suppress azimuth interference. Disturbances in direction; Step 1: Rough estimation of DOA Using the MUSIC algorithm to perform spectral peak search on 7 signals, it was found that... and Direction has energy Concentration, consisting of the real signal and some kind of symmetrical interference; Step 2: LCMV Linear Constraints Set linear constraints: Perform main lobe constraint: ; Null constraints: Suppressing disturbances from directly behind, such as suppressing... Direction of reflected waves, forced ; Derivative constraints: In The first derivative is 0, which flattens the main lobe; Step 3: Solve for the optimal weight vector Constructing the constraint matrix and response vector: .

[0054] At this point, the Lagrange multiplier method can be used to solve the problem:

[0055] Step 4: Anti-low-angle reflection treatment Because the circular array is in the XY plane, its resolution for the Z-axis (elevation angle) is low. The system incorporates a choke structure in the high-performance filter unit to physically suppress elevation angles smaller than [the required angle]. The signal, and simultaneously, in the digital domain, using the fourth-order cumulant algorithm of a high-precision time-frequency estimation unit, distinguishes between direct waves and time delays. , and ground reflected waves: time delay ,in Very small; The signal model is:

[0056] in, The received signal is the baseband signal actually received by the antenna, including interference, reflection, and noise. The transmitted signal is the original baseband signal emitted by the interference source, such as PN code or OFDM symbol; Direct wave delay refers to the time required for a signal to travel directly from the interference source to the antenna; The reflection coefficient; Multipath delay difference, i.e., the time difference corresponding to the extra path traveled by the reflected wave compared to the direct wave, is relevant for low-angle reflections. It is so small that the direct wave and the reflected wave almost overlap in the time domain; It is Gaussian white noise; And using fourth-order cumulants Slicing, eliminating Gaussian noise Accurately calculate This allows reflected waves to be eliminated or merged as independent multipath components in the Rake merging unit.

[0057] Example 3 The following are examples of embodiments for constructing motorized synthetic aperture virtual arrays; This embodiment demonstrates how a motorized synthetic aperture unit can synthesize a large aperture through motion using a small-sized physical array; Step 1: Set the motion trajectory and parameters: Platform: Hexacopter UAV; Physical array: 2-element simple interferometer, baseline length =0.2m, installed on the belly of the aircraft; Flight trajectory: flying in a straight line along the X-axis, speed

[0058] Sampling parameters: Synthetic aperture length L = 20m, flight time is 4 seconds; Number of sampling points: N = 100 points; Sampling interval: ; POS system accuracy: Position error less than 1cm, attitude error less than .

[0059] Step 2: Virtual Array Synthesis Process Operation 1, Lever Arm Compensation: First, assume the GPS antenna phase center is in the body coordinate system. The phase center of the receiving antenna is at For the k-th sampling point, the position recorded by POS is The posture is ; Among them, POS stands for Position and Attitude System, which is used to accurately record the position and attitude of the UAV; Then, rotate and translate the phase center of the received signal to the GPS phase center: in, , is the lever arm vector, which represents the offset of the receiving antenna phase center relative to the GPS antenna phase center in the body coordinate system, and R is the rotation matrix; The rotation matrix is ​​the transformation matrix that rotates the vector from the body coordinate system to the geocentric coordinate system (ECEF, Earth-centered Earth-fixed coordinate system), which is determined by the attitude angles (yaw ψ, pitch θ, roll Φ) of the POS system. Let be the GPS position vector, representing the position coordinates of the GPS antenna in the geocentric coordinate system at the k-th sampling time. ; The corrected position vector represents the equivalent position of the receiving antenna phase center in the geocentric coordinate system after Lever Arm compensation. Operation 2, Phase Compensation and Coherence Integral: Assuming the interference source to be tested is located at If the distance is relatively far, then for the k-th sampling point, the slant distance... , The azimuth angle of the interference source; Phase compensation factor:

[0060] in, It is the phase compensation factor; It is worth noting that in actual processing, all possible [processes] need to be traversed. Beamforming; Step 3: Synthesizing the signal Adding the compensated IQ data of N=100 points together, we get:

[0061] in, The synthesized array signal represents the result of coherent superposition of the signals from all N points after phase compensation. The signal received at point k represents the raw baseband IQ data collected by the UAV at the k-th position; At this point, the physical array beamwidth is: 35.8°, extremely low resolution. It is the physical baseline length, representing the actual distance between two physical antennas; The beamwidth of the synthetic array is:

[0062] Gain boosted to: Coherent integral gain = .

[0063] Therefore, through this process, the system successfully transformed a simple interferometer with a baseline of only 0.2 meters into a giant linear array with a length of 20 meters, achieving high-precision direction finding and effectively eliminating multipath interference outside the main lobe.

[0064] Specifically, this system utilizes an anti-multipath antenna array and an adaptive beamforming algorithm. The system can form a high-gain main beam in the direction of the interference source in real time, while forming a deep null in the direction of multipath reflection. This is equivalent to actively filtering out strong reflected waves from the walls of tall buildings in physical space, effectively avoiding the direction finding results being deflected by false reflected signals, and significantly improving the accuracy of DOA estimation in complex environments. Furthermore, through spatial diversity reception and maximum ratio combining (MRC) technology, it ensures that even if some antennas are blocked or are in deep fading points, the system can still recover a high signal-to-noise ratio signal. Combined with the antenna physical structure with chokes and smothering plates, low-angle multipath reflections are suppressed from the source, effectively solving the problem that traditional equipment cannot distinguish between direct waves and reflected waves, resulting in serious deviations in direction finding results or even pointing in the wrong direction. This greatly improves the accuracy and reliability of direct wave identification. Furthermore, this invention utilizes OFDM cyclic prefix technology to limit multipath delay within the guard interval, and in conjunction with a Rake receiver, converts multipath components from interference into gain for merging, effectively eliminating inter-symbol interference and ensuring reliable demodulation of broadband signals in complex reflection environments. Moreover, it employs a two-dimensional mutual ambiguity function (CAF) algorithm based on fourth-order cumulants, which takes advantage of the difference in higher-order statistical characteristics between the signal and Gaussian background noise, and can still accurately extract the time difference of arrival (TDOA) and frequency difference (FDOA) in low signal-to-noise ratio scenarios where the signal is submerged in noise. Furthermore, this invention establishes a system-level collaborative positioning module, utilizing multi-station collaborative TDOA / FDOA joint positioning, combined with mobile synthetic aperture technology from UAVs or monitoring vehicles. The system can not only utilize fixed stations but also actively fill in blind spots in high-rise buildings. Synthetic aperture technology is equivalent to turning small antennas into large-aperture arrays, greatly improving angular resolution and enabling the positioning of concealed interference sources. In addition, a high-precision 3D city model is introduced for line-of-sight analysis, automatically identifying non-line-of-sight (NLOS) paths and introducing positive time delay deviation correction or weight reduction processing, effectively correcting positioning errors caused by signal diffraction / reflection. At the same time, a radio fingerprint database is used for feature matching to assist in confirming floors or even specific rooms, achieving an improvement in positioning accuracy from the street level to the building or room level.

[0065] In some embodiments, reference is made to, for example Figure 2 As shown, an automatic method for locating and investigating radio interference sources includes the following steps: S1. Hardware Deployment and Preprocessing: First, deploy the monitoring station network in the investigation area and load the three-dimensional building model. Start the anti-multipath antenna array and spatial diversity antenna. Use the front-end bandpass filter and choke structure to filter out out-of-band noise and suppress low-angle multipath reflections to complete hardware initialization. S2. Spatial Beamforming and Diversity Combining: Based on the antenna array signal deployed in S1, an omnidirectional scan is performed to estimate the coarse DOA of the interference signal. The MVDR / LCMV algorithm is run to calculate the weighting coefficients, synthesize a highly directional main beam to filter out multipath interference, and form nulls in the multipath reflection direction. The multipath signals of spatial diversity are combined using the maximum ratio combining (MRC) algorithm to perform signal-to-noise ratio weighted combining, and output a high signal-to-noise ratio synthesized signal. S3, Digital Domain Multipath Cancellation and Signal Recovery: The high signal-to-noise ratio synthesized signal output from S2 is demodulated using orthogonal frequency division multiplexing, cyclic prefix is ​​removed, and fast Fourier transform is performed. The channel frequency response is estimated using the pilot signal, and single-tap equalization is performed in the frequency domain to transform the flat fading channel. The Rake receiver is started to search for multipath components with energy exceeding the threshold on the time axis, and phase alignment and weighted combining are performed. Finally, spread spectrum despreading and channel encoding and decoding are performed to correct burst errors and recover the original data. The specific steps for estimating the channel frequency response using pilot signals and performing single-tap equalization in the frequency domain are as follows: FFT Transform: After OFDM demodulation, the time-domain signal... After FFT transformation to the frequency domain, we obtain ; Where k is the subcarrier index; Channel estimation: Extracting pilot subcarrier data in the frequency domain, as well as known transmitted pilot values. The received value is Calculate the channel frequency response estimate:

[0066] For data subcarriers, via pilot points Perform linear interpolation or spline interpolation to obtain channel estimates for all subcarriers. ; Single-tap equalization: This involves directly performing complex division on each subcarrier in the frequency domain, i.e., a single-tap filter, as shown in the following formula:

[0067] IFFT: Converts the equalized frequency domain signal... Perform an IFFT transform to recover the original symbols in the time domain; Principle: OFDM decomposes a broadband frequency selective fading channel into multiple parallel narrowband flat fading sub-channels. Therefore, each sub-channel only needs a complex coefficient or a single tap to fully compensate for channel distortion.

[0068] S4. High-precision joint estimation of time and frequency parameters: Establish a multipath signal model including direct wave and reflected wave, input the digital signal recovered in S3 and the three-dimensional building model loaded in S1, construct a two-dimensional mutual ambiguity function CAF and perform two-dimensional search in time delay and frequency offset, so as to find the maximum concentration point of signal energy in both time delay and Doppler frequency dimensions, achieve joint optimal estimation, use the fourth-order cumulant algorithm to suppress non-Gaussian noise, search for TDOA / FDOA peak, and output accurate time difference of arrival (TDOA) and frequency difference of arrival (FDOA) values; S5. Multi-station collaborative calculation and preliminary positioning: Based on the TDOA / FDOA data extracted in S4, the synchronization time of each monitoring station in S1, and the coordinates of the monitoring stations, each monitoring station achieves time synchronization through GPS / BeiDou. The central server collects TDOA / FDOA data and solves the hyperboloid and ellipsoid equations simultaneously. It uses the Chan algorithm or Taylor series expansion method to solve the two-dimensional coordinates and error ellipse of the interference source, outputs the two-dimensional latitude and longitude coordinates and error ellipse of the interference source, and evaluates the reliability of the positioning results to determine whether it falls into the shadow area or the error is too large. If the result falls into the signal shadow area or the error exceeds the threshold, S6 is triggered to fill the blind spot. The credibility assessment includes the following steps: Geometric accuracy factor, i.e., GDOP / DOP analysis: First, calculate the condition number of the geometric configuration matrix formed by the monitoring station and the interference source. If the DOP value is too large, such as DOP value > 6, it indicates that the distribution of the monitoring stations is unreasonable, such as collinearity, and the positioning results are unreliable. Residual analysis: Substitute the solved coordinates back into the TDOA / FDOA equation system to calculate the difference between the theoretical and measured values, i.e., calculate the residuals.

[0069] If the root mean square residual (RMS) exceeds the preset threshold, such as a distance error > 50 meters, the solution is considered to have failed or been severely affected by NLOS. Shadow Area / NLOS Judgment: Based on the 3D building model, determine whether the line connecting the candidate interference source point and the monitoring station passes through the building. If it is determined to be NLOS (non-line-of-sight), the result is directly marked as low confidence. Multi-station consistency test: Compare the results calculated from different combinations of monitoring stations, such as station 1-2-3 and station 1-2-4. If the difference between the two sets of results exceeds the threshold, it indicates that there is a gross error and the result is unreliable.

[0070] It is worth noting that the decision-making logic is as follows: If DOP is normal AND residual is small AND LOS check passes → reliable, output the result; If the DOP is large OR the residual is large OR the NLOS judgment is unreliable, trigger S6 to perform a mobile blind patch; S6. Mobile blind spot filling and synthetic aperture imaging: When the initial positioning result is in the shadow area or the error is too large, dispatch the UAV or monitoring vehicle to the suspicious area determined by S5, perform multi-point coherent sampling along the predetermined trajectory and record the position and attitude data, perform phase center position calibration on the discrete sampling points, compensate the received signals of each point to the same reference point phase and then perform coherent integration processing to form a synthetic array pattern with a narrow beamwidth to penetrate the obstruction and accurately determine the direction of the interference source, output high-precision direction information or corrected coordinates, and input the more accurate direction information or coordinates obtained by S6, as well as the three-dimensional building model of S1, into S7 for final correction; S7. Map-Assisted Correction and Fingerprint Matching: Based on a 3D city model, line-of-sight analysis is performed to identify non-line-of-sight (NLOS) paths between monitoring stations and candidate interference source locations. The length difference between the direct path and the diffraction model path is calculated and used as the positive time delay deviation correction TDOA measurement value. Alternatively, the path can be assigned a low weight in the solution. The multipath delay spectrum and field strength distribution characteristics of the real-time signal are extracted and matched with a pre-stored radio fingerprint database using KNN or SVM. Finally, the accurate 3D coordinates of the interference source are output, including floor information and optimal investigation path guidance.

[0071] Specifically, the NLOS error correction in step S7 includes: Calculate the direct path length between the monitoring station and the candidate location. ; Calculate the shortest reflection / diffraction path length based on the building diffraction model. ; The difference between the two Convert to time deviation ; Will The positive correction bias is added to the TDOA measurement or the NLOS measurement is directly removed.

[0072] Specifically, the pore size processing in step S6 includes: Plan S-shaped or circular flight / driving trajectories; A high-precision POS system is used to record the phase center positions of N discrete sampling points on the trajectory. ); The received signal at the kth sampling point Perform phase compensation And perform coherent integral summation: ; in Slope distance; An equivalent giant antenna array with a size equal to the trajectory length L is formed, and a synthesized narrow beam is used to distinguish the main lobe from the adjacent reflected lobe.

[0073] The circuits and controls involved in this invention are all existing technologies and will not be described in detail here.

[0074] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. An automatic radio interference source location and investigation system, characterized in that, Includes a front-end hardware enhancement module for suppressing multipath effects and noise at the signal receiver: The core algorithm optimization module is used to perform multipath separation and demodulation of the received signal in the digital domain. This module includes: The OFDM cyclic prefix processing unit is used to remove the cyclic prefix and perform FFT transformation to convert the time-domain multipath channel into a frequency-domain flat fading channel, and perform channel equalization based on the pilot signal. The Rake multipath merging unit is used to search and identify multiple multipath components whose energy exceeds the dynamic threshold on the time axis, i.e., identify multiple multipath components with an average noise power of 3-5 times. It assigns an independent branch to each component, estimates the phase offset of each branch using pilot symbols, generates a phase rotation factor through a numerically controlled oscillator for coherent alignment, and calculates weighting coefficients based on the instantaneous signal-to-noise ratio (SNR) of each branch for vector summation to merge multipath energy. The spread spectrum despreading and channel coding / decoding unit is used to perform correlation despreading using pseudo-random codes to concentrate signal energy, and to perform error correction decoding and deinterleaving using Turbo codes or LDPC codes. The high-precision time-frequency estimation unit uses high-order statistical analysis to establish signal models for direct and reflected waves. It employs a two-dimensional mutual ambiguity function (CAF) algorithm based on fourth-order cumulants to accurately estimate the time difference of arrival (TDOA) and frequency difference (FDOA) in a noisy environment. A system-level collaborative positioning module is used to achieve multi-dimensional stereo coverage and precise positioning. The module includes: The multi-station collaborative positioning unit is used to control at least three monitoring stations to synchronize their time, collect the time difference of arrival (TDOA) or frequency difference (FDOA) data of each station, and construct a system of equations for hyperboloid and ellipsoid to solve for the coordinates of the interference source. The mobile synthetic aperture unit is used to control UAVs or monitoring vehicles to move along a predetermined S-shaped or circular trajectory. It uses a high-precision POS system to record the phase center position of each sampling point. The phase center position is calibrated by Lever Arm compensation for discrete sampling points. After compensating the received signals of each point to the same reference point phase, coherent integration is performed to form a giant antenna array with a size equal to the trajectory length, and synthesize an array pattern with a narrow beamwidth. The map-assisted correction unit is used to load a high-precision 3D city model, perform line-of-sight (LOS) analysis to identify non-line-of-sight (NLOS) paths, and introduce positive time delay bias correction or reduce the weight of NLOS paths in the solution. The fingerprint matching unit is used to store and retrieve the pre-collected radio fingerprint database, and match real-time signal features with the fingerprint database to assist in positioning.

2. The automatic radio interference source location and investigation system according to claim 1, characterized in that, The front-end hardware enhancement module also includes: The anti-multipath antenna array unit consists of multiple antenna elements deployed according to a preset geometric structure, used to collect space signals; An adaptive beamforming processing unit, connected to the anti-multipath antenna array unit, is used to run the minimum variance distortionless response (MVDR) or linear constrained minimum variance (LCMV) algorithm, calculates the weighting coefficients in real time based on the signal angle of arrival (DOA), forms a main beam pointing towards the direction of the interference source, and forms nulls in the multipath reflection direction. The spatial diversity receiving unit, through multiple independent receiving antennas deployed in different spatial locations and the maximum ratio combining (MRC) circuit, is used to combat fast signal fading. The high-performance filtering unit uses a bandpass filter and a choke coil structure to resist multipath antennas, thereby filtering out out-of-band interference and suppressing low-angle reflected waves.

3. The automatic radio interference source location and investigation system according to claim 1, characterized in that, In the OFDM cyclic prefix processing unit, the length of the cyclic prefix is ​​configured to be greater than the maximum delay spread of the target investigation area channel, and is adjustable in the range of 2μs to 10μs.

4. The automatic radio interference source location and investigation system according to claim 1, characterized in that, The high-precision time-frequency estimation unit employs high-order statistical analysis and utilizes the insensitivity of fourth-order cumulants to Gaussian noise to suppress spatial correlation clutter under low signal-to-noise ratio, thereby accurately extracting TDOA and FDOA parameters.

5. The automatic radio interference source location and investigation system according to claim 2, characterized in that, The specific operation of the adaptive beamforming processing unit is as follows: Real-time monitoring of the noise covariance matrix in a noisy environment; Linear constraints are set based on the approximate DOA of the interference source; The weighted vector is updated through an iterative algorithm to maximize the output signal-to-noise ratio while suppressing reflected signals from the sides of tall buildings and the ground. The setting of the linear constraint conditions includes the following steps: First, let the approximate direction of the interference source be... The direction of multipath reflection is The guide vector is , The array response in this direction is represented, and the constraint equations are established as follows: Perform main lobe constraint (gain preservation): ensure that the main lobe is in the direction of the interference source. The gain is 1 (or a constant) ): in, H represents the conjugate transpose of the weight vector, where H stands for Hermitian transpose, i.e., transpose first and then conjugate, used to calculate power or inner product; The steering vector pointing in the direction of the interference source; Perform null constraint (interference suppression): force it in the multipath reflection direction The gain is 0: in, The steering vector indicating the direction of interference; Apply derivative constraints: To make the main lobe more robust, The first derivative is set to 0 at this point, denoted as a flat main lobe: Under the constraints of main lobe, null, and derivative, minimizing the array output power, i.e., minimizing noise and residual interference, is as follows: in, For minimization operation, Let covariance matrix be the variance matrix. Let C be the array output power, C be the constraint matrix, and f be the response vector. The optimal weight vector is solved using the Lagrange multiplier method. ,and The minimization operation must satisfy the following conditions: It is carried out under constraints.

6. The automatic radio interference source location and investigation system according to claim 1, characterized in that, in, In the anti-multipath antenna array unit, the preset geometry includes a uniform linear array or a circular array.

7. A method for automatically locating and investigating radio interference sources using the system described in any one of claims 1-6, characterized in that, Includes the following steps: S1. Hardware Deployment and Preprocessing: First, deploy the monitoring station network in the investigation area and load the three-dimensional building model. Start the anti-multipath antenna array and spatial diversity antenna. Use the front-end bandpass filter and choke structure to filter out out-of-band noise and suppress low-angle multipath reflections to complete hardware initialization. S2, Spatial Beamforming and Divergence Combining: Based on the antenna array signal deployed in S1, perform omnidirectional scanning, estimate the coarse DOA of the interference signal, run the MVDR / LCMV algorithm to calculate the weighting coefficients, synthesize a highly directional main beam to filter out multipath interference, and form nulls in the multipath reflection direction; The maximum ratio combining (MRC) algorithm is used to combine multiple spatial diversity signals with signal-to-noise ratio weighted merging to output a high signal-to-noise ratio synthesized signal. S3, Digital Domain Multipath Cancellation and Signal Recovery: The high signal-to-noise ratio synthesized signal output from S2 is demodulated using orthogonal frequency division multiplexing, cyclic prefix is ​​removed, and fast Fourier transform is performed. The channel frequency response is estimated using the pilot signal, and single-tap equalization is performed in the frequency domain to transform the flat fading channel. The Rake receiver is started to search for multipath components with energy exceeding the threshold on the time axis, and phase alignment and weighted combining are performed. Finally, spread spectrum despreading and channel encoding and decoding are performed to correct burst errors and recover the original data. S4. High-precision joint estimation of time and frequency parameters: Establish a multipath signal model including direct wave and reflected wave, input the digital signal recovered in S3 and the three-dimensional building model loaded in S1, construct a two-dimensional mutual ambiguity function CAF and perform two-dimensional search in time delay and frequency offset, so as to find the maximum concentration point of signal energy in both time delay and Doppler frequency dimensions, achieve joint optimal estimation, use the fourth-order cumulant algorithm to suppress non-Gaussian noise, search for TDOA / FDOA peak, and output accurate time difference of arrival (TDOA) and frequency difference of arrival (FDOA) values; S5. Multi-station collaborative calculation and preliminary positioning: Based on the TDOA / FDOA data extracted in S4, the synchronization time of each monitoring station in S1, and the coordinates of the monitoring stations, each monitoring station achieves time synchronization through GPS / BeiDou. The central server collects TDOA / FDOA data and solves the hyperboloid and ellipsoid equations simultaneously. It uses the Chan algorithm or Taylor series expansion method to solve the two-dimensional coordinates and error ellipse of the interference source, outputs the two-dimensional latitude and longitude coordinates and error ellipse of the interference source, and evaluates the reliability of the positioning results to determine whether it falls into the shadow area or the error is too large. If the result falls into the signal shadow area or the error exceeds the threshold, S6 is triggered to fill the blind spot. S6. Mobile blind spot filling and synthetic aperture imaging: When the initial positioning result is in the shadow area or the error is too large, dispatch the UAV or monitoring vehicle to the suspicious area determined by S5, perform multi-point coherent sampling along the predetermined trajectory and record the position and attitude data, perform phase center position calibration on the discrete sampling points, compensate the received signals of each point to the same reference point phase and then perform coherent integration processing to form a synthetic array pattern with a narrow beamwidth to penetrate the obstruction and accurately determine the direction of the interference source, output high-precision direction information or corrected coordinates, and input the more accurate direction information or coordinates obtained by S6, as well as the three-dimensional building model of S1, into S7 for final correction; S7. Map-Assisted Correction and Fingerprint Matching: Based on a 3D city model, line-of-sight analysis is performed to identify non-line-of-sight (NLOS) paths between monitoring stations and candidate interference source locations. The length difference between the direct path and the diffraction model path is calculated and used as the positive time delay deviation correction TDOA measurement value. Alternatively, the path can be assigned a low weight in the solution. The multipath delay spectrum and field strength distribution characteristics of the real-time signal are extracted and matched with a pre-stored radio fingerprint database using KNN or SVM. Finally, the accurate 3D coordinates of the interference source are output, including floor information and optimal investigation path guidance.

8. The automatic location and investigation method for radio interference sources according to claim 7, characterized in that, The NLOS error correction in step S7 specifically includes: calculating the direct path length between the monitoring station and the candidate location, calculating the shortest reflection / diffraction path length based on the building diffraction model, and adding the difference between the two as a positive correction deviation to the TDOA measurement value, or directly discarding the measurement value.

9. The automatic location and investigation method for radio interference sources according to claim 7, characterized in that, The synthetic aperture processing in step S6 includes: calibrating the phase center position of discrete sampling points on the trajectory of the moving platform, compensating the received signals of each point to the same reference point phase, and performing coherent integration processing to form a synthetic array pattern with a narrow beamwidth, thereby distinguishing the main lobe from the adjacent reflected lobes.

Images