Radar signal detection, identification and spectrum coordination method and system based on time-frequency characteristics

Abstract

This invention discloses a radar signal detection, identification, and spectrum coordination method and system based on time-frequency features, belonging to the field of data processing technology. The method includes: acquiring a two-dimensional time-frequency matrix; generating a Boolean matrix using local energy gradient and noise floor ratio; extracting signal parameters through connected component boundary projection; constructing a heterogeneous feature vector containing continuous physical and discrete logic; determining the identification category through hybrid distance metric calculation; fusing independent time-frequency domain measurements based on the inverse of error variance weights to obtain high-precision parameters; calculating the direction of arrival based on the phase difference sequence; deriving coordinated power and generating a coordinated signal by combining the identification category and the unidirectional / bidirectional path loss balance relationship; calculating evaluation factors in the linear power domain and mapping evaluation weights to obtain a comprehensive score for executing closed-loop feedback control. This invention improves the integrity of signal detection under low signal-to-noise ratio and the recognition rate of radars with similar operating modes, and realizes closed-loop allocation of coordinated power based on spatial path balance.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a radar signal detection, identification, and spectrum coordination method and system based on time-frequency characteristics. Background Technology

[0002] With the increasing density of radio frequency devices, the electromagnetic environment in space is becoming more complex, creating a need for spectrum compatibility and coordination among multiple radar systems. In existing radar signal processing workflows, detection methods based on one-dimensional time-domain energy envelopes are limited in extracting pulse edges in environments with low signal-to-noise ratios or overlapping pulses. They are also susceptible to false alarms caused by sudden noise interference and have difficulty simultaneously acquiring the boundaries of the time and frequency domains, affecting the completeness of subsequent parameter acquisition.

[0003] Meanwhile, in the signal modulation type identification stage, some conventional methods independently process continuous physical parameters (such as pulse width and bandwidth) and discrete logical parameters (such as modulation pattern), without providing a unified mathematical measurement model for processing heterogeneous data characteristics. When faced with radar operating modes with similar characteristic parameters, the identification accuracy will decrease.

[0004] Furthermore, in radar signal parameter measurement, existing measurement mechanisms mostly rely on single time-domain or frequency-domain feature extraction. Time-domain measurements are susceptible to envelope glitches, resulting in time jitter, while frequency-domain measurements are limited by window function resolution. Both methods introduce measurement uncertainties, and no cross-fusion mechanism has been introduced to reduce single-domain errors. At the spectrum resource management and coordination level, some coordination methods employ open-loop control or fixed power allocation strategies. They fail to establish a physical mapping relationship between target radar return loss and coordinated signal transmission loss, and do not quantitatively assess the differentiated requirements of different radar operating modes for signal quality and suppression depth. This affects the rationality of spectrum coordination power allocation, making it difficult to achieve closed-loop adaptive optimization of resource scheduling and directly meet the control requirements in dense spectrum environments. Summary of the Invention

[0005] In view of the technical problems described in the background art, the present invention provides a radar signal detection, identification and spectrum coordination method and system based on time and frequency characteristics.

[0006] A radar signal detection, identification, and spectrum coordination method based on time-frequency features includes: acquiring a two-dimensional time-frequency matrix of a preset frequency band; generating a time-frequency Boolean matrix based on the ratio of local energy gradient to background noise energy; obtaining signal parameters of the target radar signal based on the boundary projection of the independent connected domains of the time-frequency Boolean matrix; constructing a heterogeneous feature vector containing continuous physical features and discrete logical features based on the signal parameters; performing hybrid range metric calculation based on a preset template feature library to obtain the identification category of the target radar signal; performing fusion calculation based on independent measurements in the frequency domain and time domain, combined with their respective error variance inverse weights, to obtain parameter measurement results; acquiring a phase difference sequence between receiving channels; obtaining the signal arrival direction based on the phase difference sequence and spatial coordinate system; determining the corresponding coordination mode based on the identification category; constructing a one-way and two-way path loss balance correlation based on the measured signal received power and path loss parameters; deriving the coordination power and generating the corresponding coordination signal; acquiring signal state variables in the linear power domain to calculate the coordination evaluation factor; performing weight matrix mapping based on the identification category to obtain the evaluation weight; obtaining a comprehensive coordination effect score based on the coordination evaluation factor and evaluation weight; and performing closed-loop feedback control based on the comprehensive coordination effect score.

[0007] Optionally, a two-dimensional time-frequency matrix of a preset frequency band is obtained, a time-frequency Boolean matrix is ​​generated based on the ratio of local energy gradient to background noise energy, and signal parameters of the target radar signal are obtained based on the boundary projection of the independent connected components of the time-frequency Boolean matrix. This includes: obtaining discrete-time signals and performing discrete Fourier transform to construct a two-dimensional time-frequency matrix composed of time window index and frequency band index; calculating the local energy gradient of the central cell within the two-dimensional sliding detection window, determining the state of the corresponding coordinate point as logically true when the ratio of the local energy gradient to the background noise energy of the surrounding reference cells reaches a set threshold, and generating a time-frequency Boolean matrix; performing connected component labeling on the time-frequency Boolean matrix to obtain independent connected components, performing orthogonal projection onto the time axis and frequency axis respectively, using the time difference between the start and end points of the non-zero projection interval on the time axis as the pulse width, and using the upper and lower bound frequency difference of the non-zero projection interval on the frequency axis as the signal bandwidth.

[0008] Optionally, a heterogeneous feature vector containing continuous physical features and discrete logic features is constructed based on signal parameters. A hybrid range metric calculation is performed based on a preset template feature library to obtain the identification category of the target radar signal. This includes: constructing a target signal feature vector, the elements of which contain numerical values ​​of continuous physical features and discrete logic features; obtaining template feature vectors from the template feature library, the template feature vectors containing calibration values ​​of continuous physical features and template discrete logic values; calculating the hybrid distance between the target signal feature vector and the template feature vector, the hybrid distance including a weighted Euclidean distance component based on statistical standard deviation and a penalty function component triggered by the comparison state of discrete logic feature values; and selecting the template category corresponding to the minimum hybrid distance as the identification category.

[0009] Optionally, based on the frequency-domain independent measurements and the time-domain independent measurements, a fusion calculation is performed using their respective inverse error variance weights to obtain the parameter measurement results. This includes: obtaining the error variances of the frequency-domain independent measurements and the time-domain independent measurements of the parameters respectively; using the error variance of the time-domain independent measurements as the weighted numerator of the frequency-domain independent measurements, and the sum of the error variances of the two as the weighted common denominator; and performing a cross-weighted summation of the frequency-domain independent measurements and the time-domain independent measurements based on the determined weighted numerator and weighted common denominator to output the parameter measurement results.

[0010] Optionally, a unidirectional and bidirectional path loss balance correlation is constructed based on the measured signal received power and path loss parameters, and the coordinated power is derived. This includes: obtaining the measured signal received power and the set target signal-to-interference suppression ratio; constructing a bidirectional path loss model corresponding to the attenuation of the target radar's own echo power, and a unidirectional path loss model corresponding to the attenuation of the coordinated signal radiated power; under the constraint of a constant signal-to-interference ratio at the receiver, constructing a unidirectional and bidirectional path loss balance correlation that eliminates spatial distance variables, and obtaining the coordinated power based on the measured signal received power, the preset radar cross section constant, spatial constant, transmit gain, receive gain, wavelength, and target signal-to-interference suppression ratio.

[0011] Optionally, the evaluation weights are obtained by performing weight matrix mapping based on the identification category, including: obtaining a baseline weight matrix containing multiple sets of different bias ratios, wherein the number of column vectors in the baseline weight matrix is ​​consistent with the total number of preset identification categories; converting the identification categories into one-hot encoded column vectors in numerical form; performing matrix multiplication of the baseline weight matrix and the one-hot encoded column vectors, and outputting an evaluation weight vector composed of quality loss weights and power suppression weights.

[0012] Optionally, the signal state variables in the linear power domain are obtained to calculate the coordination evaluation factor. A comprehensive coordination effect score is obtained based on the coordination evaluation factor and evaluation weights. Closed-loop feedback control is then performed based on the comprehensive coordination effect score, including: obtaining the linear signal-to-noise ratio of the target radar signal before and after coordinated signal radiation, and obtaining a signal quality loss factor based on their ratio; obtaining the equivalent coordinated linear power and the target echo linear power at the receiver, and obtaining a coordination suppression factor based on their ratio; mapping the coordination suppression factor, which ranges from zero to positive infinity, to a normalized constraint interval based on a negative exponential function mapping mechanism; multiplying the signal quality loss factor by the quality loss weight and then multiplying it by the mapped coordination suppression factor with the power suppression weight, and performing a weighted summation to obtain the comprehensive coordination effect score; when the comprehensive coordination effect score is lower than a preset control threshold, generating an iterative step feedback command for increasing the coordinated power output.

[0013] A radar signal detection, recognition, and spectrum coordination system based on time-frequency features is also provided, comprising: a radar signal detection and interception module, used to acquire a two-dimensional time-frequency matrix of a preset frequency band, generate a time-frequency Boolean matrix based on the ratio of local energy gradient to background noise energy, and acquire signal parameters of the target radar signal based on the boundary projection of the independent connected components of the time-frequency Boolean matrix; a signal recognition module, used to construct a heterogeneous feature vector containing continuous physical features and discrete logical features based on the signal parameters, perform hybrid distance metric calculation based on a preset template feature library, and acquire the recognition category of the target radar signal; and a parameter measurement module, used to measure independent measurements in the frequency domain and time domain, combined with their respective inverse error variances. The system performs fusion calculations based on numerical weights to obtain parameter measurement results; the direction-finding module obtains the phase difference sequence between receiving channels and uses the phase difference sequence and spatial coordinate system to obtain the signal direction of arrival; the spectrum coordination module determines the corresponding coordination mode based on the identified category, constructs a unidirectional and bidirectional path loss balance correlation based on the measured signal received power and path loss parameters, deduces the coordination power, and generates the corresponding coordination signal; the effect evaluation module obtains the signal state variables in the linear power domain to calculate the coordination evaluation factor, performs weight matrix mapping based on the identified category to obtain the evaluation weight, obtains the comprehensive coordination effect score based on the coordination evaluation factor and evaluation weight, and performs closed-loop feedback control based on the comprehensive coordination effect score.

[0014] Optionally, the radar signal detection and interception module is further configured to: acquire discrete-time signals and perform discrete Fourier transform to construct a two-dimensional time-frequency matrix; calculate the local energy gradient of the central cell within the two-dimensional sliding detection window, and determine the state of the corresponding coordinate point as logically true when the ratio of the local energy gradient to the background noise energy of the surrounding reference cells reaches a set threshold, thereby generating a time-frequency Boolean matrix; perform connected component labeling on the time-frequency Boolean matrix to obtain independent connected components, perform orthogonal projection onto the time axis and frequency axis respectively, use the time difference between the start and end points of the non-zero interval of the projection on the time axis as the pulse width, and use the upper and lower bound frequency difference of the non-zero interval of the projection on the frequency axis as the signal bandwidth.

[0015] Optionally, the effect evaluation module is further used to: obtain the linear signal-to-noise ratio of the target radar signal before and after coordinated signal radiation to obtain a signal quality loss factor; obtain the ratio of the equivalent coordinated linear power at the receiver to the linear power of the target echo to obtain a coordinated suppression factor; map the coordinated suppression factor to a normalized constraint interval based on a negative exponential function mapping mechanism; calculate a comprehensive coordinated effect score by weighted summation of the signal quality loss factor, the mapped coordinated suppression factor, and the evaluation weight obtained from the mapping; and generate an iterative step feedback instruction for increasing the coordinated power when the comprehensive coordinated effect score is lower than a preset control threshold.

[0016] The beneficial effects of this invention are reflected in: In the entire radar signal detection, recognition, and spectrum coordination method based on time-frequency features, firstly, a two-dimensional spectrogram energy morphology detection mechanism is constructed. This mechanism generates a time-frequency Boolean matrix using the local energy gradients of the central unit and background noise, and performs boundary projection of independent connected domains. This transforms the one-dimensional time threshold determination into a two-dimensional spatial connectivity determination, reducing the impact of sudden noise and overlapping interference, and obtaining the pulse width and bandwidth. Furthermore, a hybrid distance metric model fusing continuous physical features and discrete logical features is constructed. This model reduces the difference in physical dimensions by introducing a weighted Euclidean distance component based on statistical standard deviation, and constrains the logical features using a penalty function component triggered by the matching state, improving the recognition rate of radar operating modes with similar features. Finally, a fusion mechanism using the inverse weight of the dual-domain error variance is introduced. This cross-weights the independent measurements in the time and frequency domains, ensuring that data sources with smaller error variances receive a corresponding weight in the fusion result, reducing numerical jitter caused by single-domain measurement spikes. Finally, by extracting the phase difference sequence between the receiving channels and combining it with the spatial coordinate system, an analytical channel for the direction of signal arrival was constructed, which provided spatial boundary constraints for the deduction of coordinated power based on spatial path loss and improved the robustness of the radar signal analysis front-end link. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0018] Figure 1 This is a schematic diagram illustrating the steps of the radar signal detection, identification, and spectrum coordination method based on time-frequency features of the present invention. Figure 2 This is a schematic diagram of a portion of steps S1 in the radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics of the present invention; Figure 3 This is a schematic diagram of a portion of step S2 in the radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics of the present invention; Figure 4 This is a schematic diagram of part of step S6 in the radar signal detection, identification and spectrum coordination method based on time-frequency characteristics of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0021] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0022] This invention provides a radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics, such as... Figure 1 As shown, in one specific embodiment, the method includes: S1. Obtain a two-dimensional time-frequency matrix of a preset frequency band, generate a time-frequency Boolean matrix based on the ratio of local energy gradient to background noise energy, and obtain the signal parameters of the target radar signal based on the boundary projection of the independent connected domains of the time-frequency Boolean matrix.

[0023] S2. Construct a heterogeneous feature vector containing continuous physical features and discrete logical features based on the signal parameters, perform hybrid distance metric calculation based on a preset template feature library, and obtain the identification category of the target radar signal.

[0024] S3. Based on the independent measurements in the frequency domain and the independent measurements in the time domain, perform a fusion calculation by combining their respective error variance reciprocal weights to obtain the parameter measurement results.

[0025] S4. Obtain the phase difference sequence between the receiving channels, and obtain the signal direction of arrival based on the phase difference sequence and the spatial coordinate system.

[0026] S5. Determine the corresponding coordination mode according to the identified category, construct the one-way and two-way path loss balance correlation based on the measured signal received power and path loss parameters, deduce the coordination power and generate the corresponding coordination signal.

[0027] S6. Obtain the signal state variables in the linear power domain to calculate the coordination evaluation factor, perform weight matrix mapping according to the identified category to obtain the evaluation weight, obtain the comprehensive coordination effect score according to the coordination evaluation factor and the evaluation weight, and perform closed-loop feedback control according to the comprehensive coordination effect score.

[0028] In this embodiment, it should be noted that, in S1, in order to address the technical problems of existing detection methods based on one-dimensional time-domain energy envelope, such as limited pulse edge extraction and susceptibility to false alarms caused by sudden noise interference in environments with low signal-to-noise ratio or pulse overlap, and the difficulty in simultaneously acquiring time-frequency boundaries affecting the integrity of subsequent parameter interception, S1 provides a two-dimensional spectral energy morphology detection mechanism.

[0029] In the specific logic execution, the monitoring antenna front end continuously and discretely samples the preset frequency band from 1200MHz to 1300MHz at a sampling rate of 600MHz. It then frames the data using a 1024-point sliding window function and performs a discrete Fourier transform to generate a two-dimensional time-frequency matrix. The calculated background noise energy is approximately... When the target radar signal enters the detection window, the local energy gradient rises to By comparing the ratio of the local energy gradient to the background noise energy, a logically true time-frequency Boolean matrix is ​​output at the corresponding time-frequency coordinate point. Then, orthogonal boundary projection is performed on the independent connected components of this matrix, and the length of the non-zero interval of the time axis projection is mapped to... The pulse width, along with the upper and lower bounds of the non-zero interval of the frequency axis projection (1240MHz and 1260MHz), maps to a signal bandwidth of 20MHz. This processing logic transforms the one-dimensional time threshold determination into a two-dimensional spatial connectivity determination, reducing the impact of burst noise and overlapping interference, and acquiring synchronous interception data of pulse width and bandwidth.

[0030] In S2, taking into account the multi-dimensional parameters captured synchronously, and addressing the technical problem that conventional methods in the signal modulation type identification process independently handle continuous physical parameters and discrete logical parameters without providing a unified mathematical metric model for processing heterogeneous data features, which leads to a decrease in identification accuracy when facing radar operating modes with similar features, S2 constructs a hybrid distance metric logic for heterogeneous feature vectors.

[0031] In the specific logic execution, the intercepted 1250MHz center frequency and 20MHz bandwidth will be used. Pulse repetition period, Pulse width is used as a continuous physical feature, and a discrete logical feature value of 1 representing linear frequency modulation is extracted. When performing distance measurement with the corresponding calibration value in the template library, the statistical standard deviation of each continuous physical feature is set to 0.05 times the feature calibration value as the denominator. A weighted Euclidean distance component is calculated using a weighting coefficient of 0.25 to reduce dimensional differences between different physical features. Simultaneously, an asymmetric penalty function component triggered by a state is introduced for the discrete logical feature. Since the current logical feature value is consistent with the template, this penalty function component outputs 0. Finally, a mixed distance of 0 is calculated, which is used to determine the identification category as imaging mode radar. This logic fuses heterogeneous data through a unified metric model, reducing the interference of multidimensional dimensional differences on the identification process and improving the recognition rate of radar operating modes with similar features.

[0032] In S3, after obtaining the radar identification category, the measurement of radar signal parameters relies on single time-frequency domain feature extraction. This leads to time-domain measurements being susceptible to envelope glitches, causing time jitter, and frequency-domain measurements being limited by window function resolution, resulting in uncertainty. Furthermore, the lack of a cross-fusion mechanism to reduce single-domain errors presents technical problems. S3 introduces a fusion logic with the inverse of the variance weight of dual-domain errors.

[0033] In the refined measurement of pulse repetition period, frequency-domain independent measurement values ​​are obtained. The corresponding error variance, converted to a signal-to-noise ratio, is 0.01; simultaneously, time-domain independent measurements are acquired. The corresponding error variance is 0.04. During the fusion calculation, the time-domain error variance of 0.04 is used as the weighted numerator of the frequency-domain measurements, and the frequency-domain error variance of 0.01 is used as the weighted numerator of the time-domain measurements. The sum of their variances, 0.05, is used as the common denominator. Through this cross-weighted summation calculation based on the reciprocal of the error variance, the fused parameter measurement result is output as follows: This computational logic gives a higher weight to data sources with small error variance in the fusion results, reduces numerical jitter caused by single-domain measurement spikes, and provides stable time-domain reference data for subsequent spatial path loss extrapolation and spectrum resource allocation.

[0034] In S4, in order to enable the aforementioned stable time-domain reference data to be effectively applied to the allocation of resources in physical space, S4 solves the technical problem that the lack of spatial boundary constraints in the radar signal analysis link leads to subsequent power extrapolation being deviated from the support of physical space orientation.

[0035] In the collaborative operation of a multi-channel receiver, the phase difference sequence caused by the spatial path difference between the receiving channels is extracted by parallel acquisition of discrete time series from each receiving channel. Then, using a preset spatial coordinate system, the phase difference sequence is input into a spatial angle analytical model to calculate and output the direction of arrival of the target radar signal. Specifically, the following spatial angle analytical model is used to calculate the signal direction of arrival: ,in, This represents the spatial angle between the direction of the incoming signal wave and the normal to the monitoring antenna. This represents the equivalent value of the phase difference after averaging the phase difference sequences between the receiving channels; Indicates the wavelength of the target radar signal; This represents the physical spacing between antenna array elements in the receiving channels in the spatial coordinate system; It represents pi (π).

[0036] In specific operations, the physical spacing of the antenna array elements is configured. Half wavelength The equivalent value of the phase difference between the receiving channels is extracted through multi-channel interference logic. Approximately Substituting the values ​​into the spatial angle analytical model, the spatial angle between the direction of the incoming signal wave and the normal to the monitoring antenna is approximately... .Should The spatial azimuth data directly corresponds to the physical spatial constraint attributes in a specific direction, providing necessary preliminary directional data for subsequently determining the transmit and receive gains of the coordinated signal. Through this step of spatial azimuth analysis, the correspondence between the signal's time-frequency domain characteristics and physical spatial orientation is established, constructing a complete radar signal analysis pre-link, and providing basic spatial boundary constraints for subsequent physical-level power extrapolation based on the spatial path loss model.

[0037] In S5, after mastering the target radar's operating mode and spatial location, some coordination methods employ open-loop control or fixed power allocation strategies. However, the lack of a physical mapping relationship between the target radar's return loss and the coordination signal transmission loss affects the rationality of spectrum coordination power allocation. S5 constructs a single- and two-way path loss balance deduction logic. Specifically, after obtaining the target radar signal's identification category, the corresponding coordination mode is determined according to a preset mode mapping table. For example, when the identification category is "imaging mode radar," the system determines the coordination mode as "wideband noise suppression mode" to disrupt the coherence required for high-resolution imaging; when the identification category is "pulse Doppler tracking radar," the system determines the coordination mode as "comb spectrum coverage mode."

[0038] In the specific logic execution, the equivalent baseband target radar signal received power after frequency down-conversion via the receiving link is obtained as follows: The target signal-to-interference ratio (SIR) is set to 10. Under the constant SIR constraint, the received power and its value are set to... The radar cross section constant, pi, and the space constant of 1.58 are multiplied together as the numerator of the derivation. The transmit gain of 31.62 and the receive gain of 10 are then multiplied together with... The squares of the wavelengths are multiplied together to form the denominator. By multiplying the ratio of the numerator to the denominator by the signal-to-interference suppression ratio of 10, the equivalent reference coordination power used for digital baseband generation is derived as follows: Based on this, a baseband coordination signal with corresponding power is generated. This baseband coordination signal is then physically amplified by an RF power amplifier before transmission. This calculation logic cancels out the unknown spatial distance variable through equations, establishes a corresponding physical mapping relationship, eliminates the direct dependence on the actual target distance, and ensures the objectivity and rationality of spectrum resource scheduling and power allocation among multiple radars.

[0039] In S6, after coordinating signal radiation, S6 provides a unified dynamic weight quantification evaluation logic to address the technical problem that existing technologies have failed to quantify and evaluate the differentiated requirements of different radar operating modes for signal quality and suppression depth, which makes it difficult to achieve closed-loop adaptive optimization of resource scheduling.

[0040] In the specific logic execution, the linear signal-to-noise ratio (SNR) of the target radar before and after coordination was measured to decrease from 316.2 to 3.16 in the linear power domain, and the signal quality loss factor was calculated to be 0.01. Simultaneously, the ratio of equivalent received power to target echo power was measured, and the coordination suppression factor was calculated to be 20. Based on the identified imaging mode category, matrix mapping was performed to obtain a quality loss weight of 0.7 and a power suppression weight of 0.3. In the scoring calculation, a negative exponential function with a mapping constant of 0.1 was used to map the coordination suppression factor to the normalized constraint interval to eliminate dimensional conflicts, thereby calculating a comprehensive coordination effect score of 0.2665. Since this score is lower than the preset control threshold of 0.35, an iterative step feedback command was generated to adjust the original... The coordination power increases in increments of 1.5 dB to This closed-loop control logic enables the quantitative processing and dynamic mapping of evaluation indicators, meeting the control requirements in dense spectrum environments.

[0041] In summary, the entire radar signal detection, recognition, and spectrum coordination method based on time-frequency features firstly constructs a two-dimensional spectral energy morphology detection mechanism. This mechanism generates a time-frequency Boolean matrix using the local energy gradients of the central unit and background noise, and performs boundary projection of independent connected domains. This transforms the one-dimensional time threshold determination into a two-dimensional spatial connectivity determination, reducing the impact of sudden noise and overlapping interference, and obtaining the pulse width and bandwidth. Furthermore, a hybrid distance metric model fusing continuous physical features and discrete logical features is constructed. This model reduces the difference in physical dimensions by introducing a weighted Euclidean distance component based on statistical standard deviation, and constrains the logical features using a penalty function component triggered by the matching state, improving the recognition rate of radar operating modes with similar features. Finally, a fusion mechanism using the inverse weight of the dual-domain error variance is introduced. This cross-weights the independent measurements in the time and frequency domains, ensuring that data sources with smaller error variances receive a corresponding weight in the fusion result, reducing numerical jitter caused by single-domain measurement spikes. Finally, by extracting the phase difference sequence between the receiving channels and combining it with the spatial coordinate system, an analytical channel for the direction of signal arrival was constructed, which provided spatial boundary constraints for the deduction of coordinated power based on spatial path loss and improved the robustness of the radar signal analysis front-end link.

[0042] like Figure 2 As shown, in one specific embodiment, S1 includes: S11, acquiring a discrete-time signal and performing a discrete Fourier transform to construct the two-dimensional time-frequency matrix composed of a time window index and a frequency band index.

[0043] S12. Calculate the local energy gradient of the central cell within the two-dimensional sliding detection window. When the ratio of the local energy gradient to the background noise energy of the surrounding reference cells reaches a set threshold, the state of the corresponding coordinate point is determined to be logically true, and the time-frequency Boolean matrix is ​​generated.

[0044] S13. Perform connected component labeling on the time-frequency Boolean matrix to obtain independent connected components, and perform orthogonal projection onto the time axis and frequency axis respectively. Use the time difference between the start and end points of the non-zero interval of the projection on the time axis as the pulse width, and use the upper and lower bound frequency difference of the non-zero interval of the projection on the frequency axis as the signal bandwidth, thereby completing the interception of signal parameters.

[0045] In this embodiment, it should be noted that in S11, discrete-time signal acquisition and discrete Fourier transform are performed. In a specific application scenario, the monitoring antenna front end continuously and discretely samples a preset frequency band of 1200MHz to 1300MHz at a sampling rate of 600MHz to acquire discrete-time signals containing target state information. To convert the one-dimensional time series into a multi-dimensional analysis domain, a sliding window function with a length of 1024 points is used to truncate the original discrete sequence into frames, and a discrete Fourier transform is performed frame by frame to calculate the complex spectrum values ​​of each frequency band, thereby constructing a two-dimensional time-frequency matrix composed of time window index and frequency band index. This calculation logic maps single time amplitude information to a two-dimensional time-frequency space, showing the frequency distribution characteristics of the signal at different time scales, solving the problem of feature overlap in existing one-dimensional time-domain analysis in complex electromagnetic environments, and providing basic data structure support for the implementation of subsequent multi-dimensional detection mechanisms.

[0046] In S12, the logic for calculating the local energy gradient and generating the time-frequency Boolean matrix within a two-dimensional sliding window is implemented. The detection window slides within the two-dimensional time-frequency matrix, comparing the energy difference between the central cell and surrounding reference cells. The measured background noise energy in the scene is approximately 1.2 x 10⁻¹¹ watts. When the target radar signal enters, the calculated local energy gradient increases to 1.5 x 10⁻⁹ watts. Specifically, the local energy gradient of the central cell within the two-dimensional sliding detection window can be calculated using the following differential operator model. : ,in, Representing time-frequency coordinates The local energy gradient at the central unit; This indicates the energy value of the central unit; and These represent the energy values ​​of adjacent windows on the time axis; and These represent the energy values ​​of adjacent frequency bands on the frequency axis.

[0047] Furthermore, it is determined whether the ratio of the local energy gradient to the background noise energy of the surrounding reference cells reaches a set morphological threshold. This morphological threshold is determined based on statistical analysis of a limited number of historical idle spectrum observations. Specifically, multiple sets of extreme values ​​of background noise energy gradient fluctuations during specific observation periods without target radar signals are collected. The ratio sequence of these extreme values ​​to the corresponding average noise level is calculated, and the mean of this sequence is taken with an added constant tolerance factor as the morphological threshold. For example, the average ratio of noise gradient fluctuations extracted from 72 hours of historical observation records is 80. After adding a tolerance factor of 1.5, the morphological threshold is determined to be 120. When the above ratio reaches the set morphological threshold, the state of the corresponding coordinate point is determined to be logically true; regions that do not reach the threshold are set to logically false, thereby generating a time-frequency Boolean matrix. This operational logic, by comparing energy gradients rather than single amplitude values, eliminates smoothly fluctuating background noise, converts the continuous energy spectrum into a discrete time-frequency Boolean matrix, suppresses the erosion of the signal's main contour by sudden noise, and establishes morphological conditions for subsequent pulse boundary extraction.

[0048] In S13, connected component labeling and orthogonal boundary projection were performed. A connected component labeling algorithm was applied to the logically true elements in the matrix to extract the independent connected components with the largest area, thus identifying the main shape of the target signal. Subsequently, orthogonal projection operations were performed on the time axis and frequency axis, respectively. In the scene data, the time difference between the start and end points of the non-zero projection interval on the time axis was mapped to 20 microseconds; this difference served as the extracted pulse width. The upper and lower bounds of the non-zero projection interval on the frequency axis were 1240MHz and 1260MHz, respectively, and the frequency difference of 20MHz served as the signal bandwidth. This boundary projection logic based on two-dimensional spatial connectivity eliminates the easily fluctuating amplitude envelope determination, directly extracting the time and frequency boundaries synchronously from the morphological clustering results. This obtains signal parameters unaffected by overlapping interference, ensuring the objectivity of the input features in subsequent signal recognition stages.

[0049] like Figure 3 As shown, in one specific embodiment, S2 includes: S21, constructing a heterogeneous feature vector containing continuous physical features and discrete logic features. The continuous physical features include center frequency, signal bandwidth, pulse repetition period, and pulse width; the discrete logic features are modulation patterns.

[0050] S22. Obtain template feature vectors from the template feature library, and perform a hybrid distance metric calculation to obtain the recognition category. Hybrid distance The calculation formula is as follows:

[0051] in, Represents the target signal feature vector and the first Mixed distance between template feature vectors This represents the feature index, with values ​​ranging from 1 to 4. Indicates the first Weighting coefficients of continuous physical characteristics, The first feature vector extracted from the target signal represents the... Item continuous physical characteristic values, Indicates the first The first template Item continuous physical characteristic calibration value, Indicates the first in the template library The statistical standard deviation of the characteristic parameters, This represents the penalty function component triggered by the numerical comparison of discrete logic features. The discrete logic characteristic values ​​of the target signal are represented. Indicates the first Template discrete logic values ​​of a template; when and When the two are inconsistent Taking a constant of 1000, when consistent Take 0.

[0052] In this embodiment, it should be noted that in S21, the construction logic for heterogeneous feature vectors is executed. Multiple dimensions of the target signal are extracted and assembled into a unified vector structure. Specifically, the measured 1250MHz center frequency, 20MHz signal bandwidth, 500µs pulse repetition period, and 20µs pulse width are used as continuous physical feature values. Simultaneously, combined with the time-frequency morphological variation pattern, discrete logical feature values ​​of 1 are analyzed and assigned to the target signal. These discrete logical feature values ​​represent the linear frequency modulation pattern. This feature vector construction logic integrates the continuously changing physical measurement values ​​with the discrete classification attributes of logical labels into the same mathematical structure, solving the technical problem of isolated heterogeneous data that cannot be jointly calculated. This provides complete underlying data model support for the subsequent joint measurement and type determination of radar operating modes.

[0053] In S22, a hybrid range metric calculation is implemented. Specifically, for the heterogeneous feature joint identification calculation step in S22, a hybrid range metric is performed to determine the identification category of the target radar signal. Its calculation expression is as follows: This computational process aims to solve the technical problem in existing radar signal identification where continuous physical features and discrete logical features cannot be quantitatively compared within the same mathematical space.

[0054] In terms of computational logic settings, since the various characteristics of radar signals have completely different physical dimensions and numerical magnitudes, such as the center frequency often reaching hundreds of megahertz, while the pulse width is only on the order of microseconds, if the Euclidean distance is calculated directly, the feature with a large numerical magnitude will mask the difference weight of the feature with a small numerical magnitude.

[0055] Therefore, the formula incorporates the statistical standard deviation of the k-th feature parameter from the template library. As the denominator, the k-th continuous physical feature value extracted from the target signal is... The k-th continuous physical feature calibration value of the i-th template The deviations between them are divided to achieve dimensionless processing of the features. Then, the weighting coefficients of the k-th continuous physical feature are combined... We perform square weighting and sum to obtain the square root, so that the deviations of each physical dimension can be accumulated in the form of proportional weights.

[0056] Meanwhile, to handle discrete state data that cannot be subjected to conventional addition and subtraction operations, the formula includes a penalty function triggered by comparing discrete logic feature values. When the discrete logic characteristic value of the target signal With the Template Discrete Logic Values ​​of a Template When the signals are consistent, the penalty function is assigned a value of zero; when they are inconsistent, a specific maximum constant is assigned. This maximum constant (e.g., a value of 1000) is determined based on the extreme boundary of the range of mixed distance metrics for continuous physical features. Specifically, by statistically analyzing the weighted sum of the maximum measurement deviations of typical radar continuous physical features under low signal-to-noise ratio conditions, the upper limit of the numerical envelope of this sequence is calculated, and a scalar much larger than this upper limit is set as the preset penalty function scalar for discrepancy in discrete logic features. For example, in 500 historically poor signal-to-noise ratio simulation tests, the maximum weighted Euclidean distance component of the same type of signal due to fluctuations in physical parameters typically does not exceed 50. To ensure an effective classification rejection effect when logic patterns are mismatched, the penalty function constant is set to 1000, thus forming a numerical classification boundary in the mathematical model.

[0057] Before performing specific feature calculations, the weight coefficients of each continuous physical feature in the model are determined. Calculation of statistical standard deviation The required multiplier factor values ​​are all obtained based on the historical measurement deviation distribution and information gain fitting. Specifically, multiple measurement datasets of similar radars under different conditions are statistically analyzed, the relative measurement error variance distribution of each physical feature is calculated, the extreme values ​​of relative errors covering the 95% confidence interval are extracted as statistical standard deviation multiplier factors, and principal component analysis is used to calculate the information gain ratio of each feature to the classification result to allocate weight coefficients. For example, analyzing 100 sets of historical measurement data yields a relative error extreme value of 4.8% for the center frequency, and an approximate calibration factor of 0.05 is used as the multiplier factor to calculate... Meanwhile, principal component analysis revealed that the information gain proportions of the four physical features were similar; therefore, the weight coefficients of each feature were set in a balanced manner. Both are 0.25.

[0058] Combined with specific data from the application scenario, the captured continuous physical characteristic values , , , Discrete logic characteristic numerical When compared with an imaging radar template, the continuous physical feature calibration values ​​of the template are... Both , , , The corresponding statistical standard deviation The value is uniformly set to 0.05 times the calibration value. , , , Weighting coefficients of each component All values ​​are set to 0.25. At this point, since the target signal features are within the tolerance matching range of the template calibration value, and the discrete logic feature values ​​are consistent, the penalty function is 0. The calculated target signal feature vector and the... Mixing distance between template feature vectors This calculation process compresses multidimensional heterogeneous data into a single distance scalar, reducing dimensional conflicts during feature comparison and improving the ability to identify and match complex radar signals.

[0059] In one specific implementation, S3 includes: acquiring independent measurements of the parameter in the frequency domain and time domain, respectively. During the fusion calculation, the error variance of the time-domain independent measurements is used as the weighted numerator of the frequency-domain independent measurements, and the sum of the error variances of both is used as the weighted common denominator. The final parameter measurement result is obtained by cross-weighting and summing the frequency-domain independent measurements and the time-domain independent measurements.

[0060] In this embodiment, it should be noted that in S3, a dual-domain parameter cross-fusion logic based on the inverse of the error variance weight is implemented. Specifically, it can be expressed as: ,in, Indicates the parameter measurement results; Represents independent measurements in the frequency domain; This represents the variance of the error of independent measurements in the frequency domain; Represents time-domain independent measurements; It represents the variance of the error of independent measurements in the time domain.

[0061] In the refined measurement scenario of pulse repetition period, the frequency domain independent measurement value is acquired as 500.02 microseconds, with a corresponding error variance of 0.01; simultaneously, the time domain independent measurement value is acquired as 499.97 microseconds, with a corresponding error variance of 0.04. During the fusion calculation, the error variance of the time domain independent measurement value (0.04) is used as the weighted numerator of the frequency domain independent measurement value, and the error variance of the frequency domain independent measurement value (0.01) is used as the weighted numerator of the time domain independent measurement value. The sum of the error variances of the two, 0.05, is used as the weighted common denominator. Through this cross-calculation mechanism of multiplying the corresponding measurement value by its weighted ratio and summing the results, the fused parameter measurement result is output as 500.01 microseconds.

[0062] This calculation logic uses error variance as a data reliability evaluation index, which gives data sources with high measurement accuracy and small error variance a higher allocation weight in the final result. It reduces numerical deviations and jitter caused by single-domain measurement glitches or resolution limitations, improves the stability of radar signal parameter analysis, and provides accurate parameter benchmarks for subsequent direction finding in associated spatial coordinate systems and for extrapolating spectral coordinated power based on single and two-way path loss.

[0063] In one specific implementation, S5 includes: obtaining coordinated power by constructing a unidirectional and bidirectional path loss balance correlation relationship. The calculation formula is as follows:

[0064] in, This represents the coordination power obtained from the simulation. This indicates the set target signal-to-interference ratio. This indicates the measured signal received power. Represents pi (π). This represents a space constant that includes atmospheric loss in space. This represents the preset radar cross section constant. Indicates transmit gain. Indicates the receive gain. Indicates the wavelength of the target radar signal.

[0065] In this embodiment, it should be noted that in S5, for the spectrum coordination power control stage, the coordination power is derived based on the unidirectional and bidirectional path loss balance correlation, and its calculation expression is as follows: This calculation process is primarily used to address the technical problem of how to calculate the coordinated radiated power that can produce a specific suppression effect at the target radar receiver when the exact spatial physical distance to the target radar is unknown.

[0066] In radar transmission equations, the attenuation of echo power is inversely proportional to the fourth power of the target distance, while the power attenuation of a coordinated unidirectional transmitted signal reaching the target radar is inversely proportional to the square of the distance. If directly estimated in an open-loop configuration, errors in spatial distance will be amplified by high powers, leading to distortion in power distribution.

[0067] Therefore, this formula utilizes the locally measured target radar signal received power by constructing an equation substitution under the constant signal-to-interference ratio constraint. The included unidirectional path loss information offsets the distance variable in the bidirectional detection path. In the specific structure, the measured signal received power... Pi Space constant including atmospheric loss in space and the preset radar cross section constant The product, as the numerator, reflects the physical reflection basis of the target radar's own echo dependence; it will coordinate the signal transmission gain. Receiver gain and the wavelength of the target radar signal The product of squares, used as the denominator, reflects the antenna and frequency characteristics of the transmit / receive link. The ratio of these two values ​​restores the base power of the equivalent interference at standard spatial distances, multiplied by the set target signal-to-interference suppression ratio. Output the coordinated power obtained from the deduction .

[0068] Furthermore, the target signal-to-interference suppression ratio in the above model Space constant including atmospheric loss in space With respect to the preset radar cross section constant All data are based on historical test calibration and prior equipment data. Specifically, the average critical power ratio required to disable the target radar system's detection capability in historical electronic countermeasures field tests is extracted as the target signal-to-interference suppression ratio. Simultaneously, the spatial constant is determined by combining measured atmospheric attenuation records of the operating frequency band, and the radar cross-section constant is extracted based on typical physical size calibration data of the target radar. For example, the average power ratio that caused the radar to be unable to effectively detect signals in 20 historical field tests of the same L-band radar is 9.5, which is rounded up to set the target signal-to-interference suppression ratio. The value is 10; refer to the L-band calibration spatial constant. The value is 1.58; the mean geometric area of ​​the antenna is extracted from the equipment database, and the radar cross-section constant is set. for .

[0069] The specific data from the application scenario is analyzed to measure the target radar signal reception power. The set target signal-to-interference ratio The space constant including atmospheric loss in space The preset radar cross section constant Coordinate signal transmission gain Receive gain The wavelength of the target radar signal Pi Calculate using constants. Perform multiplication and division operations, i.e. Divide by Coordination power obtained through simulation This calculation process eliminates the distance variable dimension through equation balancing, establishes the physical mapping between received power and transmitted power, reduces energy deviation caused by open-loop guessing, and ensures the objectivity of spectrum resource scheduling.

[0070] like Figure 4 As shown, in one specific embodiment, S6 includes: S61, obtaining the linear signal-to-noise ratio of the target radar signal before and after the coordinated signal radiation, and obtaining a signal quality loss factor based on the ratio of the two. Obtaining the equivalent coordinated linear power at the receiver and the target echo linear power, and obtaining a coordinated suppression factor based on the ratio of the two.

[0071] S62. Perform weight matrix mapping based on the identified category. Perform matrix multiplication of the baseline weight matrix and the one-hot encoded column vector of the identified category in numerical form, and dynamically output the evaluation weight vector consisting of quality loss weight and power suppression weight.

[0072] S63. Obtain a comprehensive coordination effect score. :

[0073] in, : Indicates the overall coordination effect score. : Represents the quality loss weight obtained from the mapping. : Represents the signal quality loss factor obtained in the linear domain. : Represents the power suppression weights obtained from the mapping. : Represents the natural constant. : Represents the preset mapping mechanism constant, with a value of 0.1. : Represents the coordination inhibition factor obtained in the linear domain. When scoring When the power output is below the preset control threshold, an iterative step feedback command is generated to increase the coordinated power output, thereby achieving closed-loop control.

[0074] In this embodiment, it should be noted that in S61, signal state variables in the linear power domain are acquired and evaluation factors are calculated. The linear signal-to-noise ratio (SNR) of the target radar signal before coordination is measured to be 316.2 in the linear power domain, and this SNR drops to 3.16 after the coordinated signal is radiated. A signal quality loss factor of 0.01 is obtained based on the ratio of these two values, thus characterizing the degree of degradation of target imaging or detection quality caused by interference. Simultaneously, the equivalent coordinated linear power at the receiver and the target echo linear power are acquired, and a coordination suppression factor of 20 is obtained based on their ratio, thus characterizing the power coverage depth of the coordinated signal on the target echo. This logic of directly calculating dimensionless ratios in the linear power domain avoids the mathematical ambiguity and physical dimension distortion caused by direct addition, subtraction, multiplication, and division in the logarithmic domain. It transforms the complex electromagnetic suppression effect into a standardized evaluation factor that can be directly used for weighted calculations in subsequent scoring models. The signal quality loss factor and coordination suppression factor obtained above together constitute the coordination evaluation factor used for quantitative assessment.

[0075] In S62, dynamic weight matrix mapping logic based on the identified category is executed. The system pre-configures a baseline weight matrix containing multiple sets of different bias ratios. The values ​​of each bias ratio in this matrix are obtained using the analytic hierarchy process (AHP) based on a finite number of historical electromagnetic spectrum expert evaluation data. Specifically, for each target radar's operating category, multiple domain experts are collected to score the relative importance of signal quality degradation and absolute power suppression. A judgment matrix is ​​constructed, and the eigenvector corresponding to the largest eigenvalue is calculated. After normalization, this eigenvector serves as the baseline weight. For example, historical statistics show that imaging mode radars are highly sensitive to phase and signal continuity. In 50 sets of expert comprehensive scoring data, the relative importance of signal quality loss is determined to be far higher than that of simple power suppression. Based on this, the corresponding bias ratio for imaging mode radar is derived as a quality loss weight of 0.7 and a power suppression weight of 0.3. When S2 confirms that the target is an imaging mode radar, the identified category is converted into a one-hot encoded column vector.

[0076] Furthermore, by performing matrix multiplication of the baseline weight matrix and the one-hot encoded column vector, an evaluation weight vector focusing on signal quality degradation is dynamically extracted, resulting in a quality loss weight of 0.7 and a power suppression weight of 0.3. This operational logic converts discrete radar mode classification results into continuous weight allocation coefficients, addressing the deficiency of not being able to finely adapt to the differences in anti-interference sensitivity of different radar systems. This allows subsequent resource scheduling effect evaluation to generate adaptive biases based on the specific operating mechanism of the target radar, ensuring the rationality and objectivity of the evaluation system.

[0077] In S63, a comprehensive coordination effect score calculation and feedback control were implemented. Specifically, for the closed-loop control and resource scheduling evaluation stage in S63, a comprehensive coordination effect score was calculated to execute feedback control, and its calculation expression is as follows: This computational process aims to resolve the dimensional conflicts in multi-target evaluation indicators during radar spectrum coordination, as well as the technical problem that fixed weights are difficult to adapt to the differentiated sensitivities of different radar systems.

[0078] In spectrum management, the signal quality loss factor is obtained in the linear domain. Typically a small ratio between 0 and 1, while the coordination inhibition factor obtained in the linear domain The range could expand from a dozen to hundreds. If the two are directly weighted and added together, the large value of the coordination inhibition factor will mask the weighting effect of the signal quality loss factor, causing the evaluation model to become unstable.

[0079] Therefore, the formula introduces a natural constant-based factor for the coordination inhibition factor. The negative exponential function mapping mechanism utilizes This structure smoothly compresses input data ranging from zero to positive infinity into a normalized constraint interval of 0 to 1. The preset mapping mechanism constant... The preset control thresholds required for subsequent closed-loop operations are all obtained based on nonlinear regression of historical adversarial effectiveness data. Specifically, historical test datasets covering different coordination power degradation states are extracted, and nonlinear regression training is performed with the manually determined suppression success rate as the dependent variable. The model convergence coefficient is extracted as the mapping mechanism constant. The scores corresponding to the successful suppression probability contour lines are extracted as preset control thresholds; for example, by fitting and calculating using 500 sets of historical records, the optimal attenuation coefficient is found to be 0.098, which is approximately the constant of the calibration mapping mechanism. The value is set to 0.1, and the model output value corresponding to a 90% success rate is extracted as 0.345. This value is then rounded up to determine the preset control threshold as 0.35. After dimensional alignment, the formula is combined with the quality loss weights obtained through mapping. Power suppression weights obtained by mapping Perform a weighted summation and output a comprehensive coordination effect score. .

[0080] The signal quality loss factor is calculated in the linear power domain based on operational data from the application scenario. The obtained coordinating inhibitory factor Since the preceding steps identify the signal as an imaging mode radar, a quality loss weight biased towards signal quality is obtained through mapping. and power suppression weights And set preset mapping mechanism constants. To calculate using the formula, first calculate the negative exponent term. Thus, the value of the mapped suppression term is obtained as follows: Then weight allocation is performed. Plus Finally, the overall coordination effect score was calculated. This operational logic transforms the heterogeneous suppression state into a stable control scalar, providing a quantitative basis for generating iterative step feedback instructions when the score falls below the control threshold, thus establishing a feedback data chain for spectrum management.

[0081] This invention also provides a radar signal detection, identification, and spectrum coordination system based on time-frequency features. The system is used to implement a radar signal detection, identification, and spectrum coordination method based on time-frequency features. The system includes: The radar signal detection and interception module is used to acquire a two-dimensional time-frequency matrix of a preset frequency band, generate a time-frequency Boolean matrix based on the ratio of local energy gradient to background noise energy, and obtain the signal parameters of the target radar signal based on the boundary projection of the independent connected domains of the time-frequency Boolean matrix. The signal recognition module is used to construct a heterogeneous feature vector containing continuous physical features and discrete logical features based on signal parameters, and to perform hybrid distance metric calculations based on a preset template feature library to obtain the recognition category of the target radar signal. The parameter measurement module is used to perform fusion calculations based on independent frequency domain measurements and independent time domain measurements, combined with their respective error variance inverse weights, to obtain parameter measurement results. The direction finding module is used to acquire the phase difference sequence between the receiving channels and to obtain the direction of arrival of the signal wave based on the phase difference sequence and the spatial coordinate system. The spectrum coordination module is used to determine the corresponding coordination mode according to the identification category, construct the unidirectional and bidirectional path loss balance correlation based on the measured signal received power and path loss parameters, deduce the coordination power and generate the corresponding coordination signal. The effect evaluation module is used to obtain the signal state quantities in the linear power domain to calculate the coordination evaluation factor, perform weight matrix mapping according to the identified category to obtain the evaluation weight, obtain the comprehensive coordination effect score according to the coordination evaluation factor and the evaluation weight, and perform closed-loop feedback control according to the comprehensive coordination effect score.

[0082] In one specific implementation, the radar signal detection and interception module is further configured to: acquire discrete-time signals and perform discrete Fourier transform to construct a two-dimensional time-frequency matrix; calculate the local energy gradient of the central cell within the two-dimensional sliding detection window, and determine the state of the corresponding coordinate point as logically true when the ratio of the local energy gradient to the background noise energy of the surrounding reference cells reaches a set threshold, thereby generating a time-frequency Boolean matrix; perform connected component labeling on the time-frequency Boolean matrix to obtain independent connected components, perform orthogonal projection onto the time axis and frequency axis respectively, use the time difference between the start and end points of the non-zero projection interval on the time axis as the pulse width, and use the upper and lower bound frequency difference of the non-zero projection interval on the frequency axis as the signal bandwidth.

[0083] In one specific implementation, the effect evaluation module is further configured to: obtain the linear signal-to-noise ratio of the target radar signal before and after coordinated signal radiation to obtain a signal quality loss factor; obtain the ratio of the equivalent coordinated linear power at the receiver to the linear power of the target echo to obtain a coordinated suppression factor; map the coordinated suppression factor to a normalized constraint interval based on a negative exponential function mapping mechanism; calculate a comprehensive coordinated effect score by weighted summation of the signal quality loss factor, the mapped coordinated suppression factor, and the evaluation weight obtained from the mapping; and generate an iterative step feedback instruction for increasing the coordinated power when the comprehensive coordinated effect score is lower than a preset control threshold.

[0084] To further clarify the operating mechanism and physical quantization process of the technical solution of this invention, the following analysis will be conducted in detail on the underlying derivation logic of the radar signal detection, identification and spectrum coordination method and system based on time-frequency characteristics, using a scenario containing specific parameters and data.

[0085] In a radio frequency protection monitoring scenario at a satellite remote sensing ground station in a suburban city, high-frequency pulse interference from L-band synthetic aperture radar (SAR) exists around the station. The ground station system needs to perform radar signal detection and identification and spectrum coordination tasks based on time-frequency characteristics. In the radar signal detection and interception phase of S1, the monitoring antenna front end uses... The sampling rate is used to continuously and discretely sample the preset frequency band from 1200MHz to 1300MHz. According to the logic in steps S11 to S13, the system frames the acquired discrete signal using a sliding window function of length 1024 points and performs a discrete Fourier transform to generate a two-dimensional time-frequency matrix. During morphological detection, the system calculates the background noise energy to be approximately... When the target radar signal enters the detection window, the local energy gradient rises to When the ratio of the two exceeds a preset morphological threshold, the system randomly outputs logical true at the corresponding time-frequency coordinate point. Subsequently, in S14, the system performs orthogonal boundary projection on the largest connected component in the time-frequency Boolean matrix, and measures the time axis projection interval as follows: The upper and lower bounds of the frequency axis projection are 1240MHz and 1260MHz, respectively. Therefore, the system successfully intercepted the preliminary parameters of the signal, namely the pulse width. The signal bandwidth is 20MHz.

[0086] In the signal identification stage (S2), the system constructs a feature vector of the target signal based on the intercepted parameters. The center frequency is measured at this point. signal bandwidth Pulse repetition period Pulse width And through preliminary analysis, its discrete logic characteristic values ​​were determined. (Represents linear frequency modulation). In S22, the system retrieves a typical L-band imaging radar template from the template library, and its calibration values... Corresponding to The corresponding statistical standard deviation The value is uniformly set to 0.05 times the eigenvalue. The system calculates the deviation between the target signal and the template by performing a hybrid distance metric: ,in, For mixed distance; This is the feature index, with a value range of 1 to 4; These are normalized weighting coefficients, with each component taking a value of 0.25. These are the continuous physical characteristic values ​​of the target signal; For template feature calibration values; The standard deviation is the statistical value. It is an asymmetric penalty function. In this scenario, because... Consistent with the template, The value is set to 0, and the mixing distance is calculated. If the value is less than the discrimination threshold, the system determines that the identification category is imaging mode radar.

[0087] In the parameter measurement step of S3, to further improve the accuracy of the data, the system performs fine-grained fusion of the pulse repetition period. First, the frequency domain repetition period is measured. The corresponding error variance is converted to signal-to-noise ratio as follows: Secondly, the time-domain repetition period was measured. The corresponding error variance is Finally, the system uses the reciprocal of the error variance weighting logic for fusion: 0.04 is used as the weighting numerator for the frequency domain values, and 0.01 is used as the weighting numerator for the time domain values, with the common denominator being the sum of their variances, i.e., 0.05. The calculated weighted fused parameter measurement results are as follows: This process effectively utilizes the accuracy advantage of frequency domain measurements under stable signals, avoids the deviations caused by time domain glitches in single-domain measurements, and provides a stable time domain reference for subsequent S5 spectrum resource allocation.

[0088] During the spectrum coordination phase of S5, the system needs to issue coordination commands based on the identified imaging mode and radar type. First, S4 measures the angle between the signal arrival direction and the antenna normal. Subsequently, the coordinated power was deduced based on the correlation between single and bidirectional path loss balance. The target radar signal received power is known. (about The target signal-to-interference suppression ratio was set. Reference radar cross section space constant (Corresponding to the approximation) (loss), coordinate transmit antenna gain (correspond ), system receive gain (correspond ),wavelength Deducing Coordination Power The calculation model is as follows: ,in, To coordinate power; The target signal-to-interference ratio; The measured target radar signal received power; Pi; This is a space constant that includes atmospheric loss in space; The radar cross section constant is denoted by . For transmit gain; For receiving gain; The wavelength is used. Substituting the data, the digital reference coordination power is calculated. Subsequently, the digital baseband module is driven to generate a broadband suppression signal with a corresponding reference power, which is then physically amplified and radiated through the radio frequency front end with a fixed system gain to achieve physical isolation of the frequency band resources.

[0089] In S6, the system enters the effect evaluation phase, monitoring the coordinated electromagnetic environment in real time within the linear power domain. In S61, monitoring revealed that the target radar's linear signal-to-noise ratio before coordination was 316.2 (corresponding to...). After coordinating signal radiation, the linear signal-to-noise ratio in this frequency band drops to 3.16 (corresponding to...). The signal quality loss factor is calculated from this. Simultaneously, the ratio of the equivalent received power of the coordinated signal to the radar signal echo power is measured in the target direction to derive the coordinated suppression factor. Since the category is identified as imaging mode in S2, S62 performs dynamic weight matrix mapping and outputs the quality loss weights. Power suppression weight In S63, the system uses a mapping mechanism to calculate the overall coordination effect score. : ,in, The overall coordination effect is scored; Weighting for quality loss; This is the signal quality loss factor; Power suppression weights; It is a natural constant; The mapping mechanism constant is set to 0.1; This is a coordinating inhibitory factor. Calculated... .

[0090] Finally, the system executes feedback control logic based on the calculated score. The system's preset closed-loop control threshold is set to 0.35, based on the current score... If the value is below the threshold, it indicates that the current coordination strength has not fully achieved the expected isolation effect for the imaging mode radar. Based on this, the system generates an iterative step feedback command to incrementally optimize the digital reference coordination power. The iteration step size set here (e.g., 1.5dB) is determined based on a comprehensive experimental optimization of the digital baseband adjustment resolution and the system's closed-loop convergence speed. Specifically, multiple different step sizes are set in a digital signal processing simulation environment. The number of iterations required for the system score to reach the preset control threshold and the steady-state fluctuation of the digital reference power are recorded at different step sizes. A suitable crossover point that balances response rate and avoids large oscillations is selected as the feedback step size. For example, data from nearly a hundred historical debugging tests show that a 1.0dB step size leads to too many convergence iterations and poor real-time response, while a 3.0dB step size easily causes numerical overshoot of the digital reference power. After comprehensively weighing a finite number of historical optimization curves, 1.5dB is determined as the optimal iteration step size. The system generates an iterative step feedback command according to S63, which adjusts the original... Coordination power The step size was incrementally adjusted, and the power value was updated to approximately [value missing]. As power increases, the system re-enters the S6 cycle for monitoring until the overall coordination effect score is recalculated. Once the 0.35 threshold is exceeded, the system stops power iteration and maintains the current coordinated output. This process fully embodies the first-principles-based closed loop of "detection-identification-deduction-evaluation-feedback," ensuring the normal operation of the ground station while achieving precise control of radar spectrum resources in complex electromagnetic environments.

[0091] The preferred embodiments of this disclosure have been described in detail above with reference to the accompanying drawings. However, this disclosure is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this disclosure, various simple modifications can be made to the technical solutions of this disclosure, and these simple modifications all fall within the protection scope of this disclosure.

[0092] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, this disclosure will not describe the various possible combinations separately.

[0093] Furthermore, various different embodiments of this disclosure can be combined in any way, as long as they do not violate the spirit of this disclosure, they should also be regarded as the content disclosed in this disclosure.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics, characterized in that, The methods include: A two-dimensional time-frequency matrix of a preset frequency band is obtained. A time-frequency Boolean matrix is ​​generated based on the ratio of local energy gradient to background noise energy. The signal parameters of the target radar signal are obtained by the boundary projection of the independent connected domains of the time-frequency Boolean matrix. Based on the signal parameters, a heterogeneous feature vector containing continuous physical features and discrete logical features is constructed. Based on the preset template feature library, a hybrid distance metric calculation is performed to obtain the identification category of the target radar signal. Based on the independent measurements in the frequency domain and the independent measurements in the time domain, a fusion calculation is performed using their respective corresponding error variance inverse weights to obtain the parameter measurement results; Obtain the phase difference sequence between receiving channels, and determine the direction of signal arrival based on the phase difference sequence and the spatial coordinate system; Based on the measured signal received power and path loss parameters, a unidirectional and bidirectional path loss balance correlation is constructed, and the coordination power is derived and the corresponding coordination signal is generated. The signal state variables in the linear power domain are obtained to calculate the coordination evaluation factor. The weight matrix is ​​mapped according to the identified category to obtain the evaluation weight. The comprehensive coordination effect score is obtained based on the coordination evaluation factor and the evaluation weight. Closed-loop feedback control is then performed based on the comprehensive coordination effect score.

2. The radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics according to claim 1, characterized in that, A two-dimensional time-frequency matrix for a preset frequency band is obtained. A time-frequency Boolean matrix is ​​generated based on the ratio of local energy gradient to background noise energy. The signal parameters of the target radar signal are obtained based on the boundary projection of the independent connected components of the time-frequency Boolean matrix, including: The discrete-time signal is acquired and a discrete Fourier transform is performed to construct a two-dimensional time-frequency matrix consisting of a time window index and a frequency band index. Calculate the local energy gradient of the central cell within the two-dimensional sliding detection window. When the ratio of the local energy gradient to the background noise energy of the surrounding reference cells reaches a set threshold, the state of the corresponding coordinate point is determined to be logically true, and a time-frequency Boolean matrix is ​​generated. Perform connected component labeling on the time-frequency Boolean matrix to obtain independent connected components, and perform orthogonal projection on the time axis and frequency axis respectively. The time difference between the start and end points of the non-zero interval of the projection on the time axis is taken as the pulse width, and the upper and lower bound frequency difference of the non-zero interval of the projection on the frequency axis is taken as the signal bandwidth.

3. The radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics according to claim 1, characterized in that, Based on signal parameters, a heterogeneous feature vector containing continuous physical features and discrete logical features is constructed. A hybrid range metric calculation is performed based on a pre-defined template feature library to obtain the target radar signal's identification category, including: Construct a feature vector for the target signal. The elements of the feature vector include the numerical values ​​of continuous physical features and discrete logic features. Obtain template feature vectors from the template feature library. The template feature vectors contain continuous physical feature calibration values ​​and template discrete logic values. The mixed distance between the target signal feature vector and the template feature vector is calculated. The mixed distance includes a weighted Euclidean distance component based on statistical standard deviation and a penalty function component based on discrete logic feature numerical comparison state triggering. The template category corresponding to the minimum mixing distance is selected as the recognition category.

4. The radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics according to claim 1, characterized in that, Based on the independent measurements in the frequency domain and the independent measurements in the time domain, a fusion calculation is performed using their respective inverse error variance weights to obtain the parameter measurement results, including: Obtain the error variance of the frequency domain independent measurement value and the error variance of the time domain independent measurement value of the parameter respectively; The error variance of the time-domain independent measurement is used as the weighted numerator of the frequency-domain independent measurement, and the error variance of the frequency-domain independent measurement is used as the weighted numerator of the time-domain independent measurement. The sum of the error variances of the two is used as the weighted common denominator. Based on the determined weighted numerator and common weighted denominator, the frequency domain independent measurements and time domain independent measurements are cross-weighted and summed to output the parameter measurement results.

5. The radar signal detection, identification, and spectrum coordination method based on time-frequency features according to claim 1, characterized in that, Based on the measured signal received power and path loss parameters, a correlation between unidirectional and bidirectional path loss balance is constructed, and the coordinated power is derived, including: Acquire the measured signal received power and the set target signal-to-interference suppression ratio; Construct a two-way path loss model corresponding to the attenuation of the target radar's own echo power, and a one-way path loss model corresponding to the attenuation of the coordinated signal radiated power. Under the constraint of constant signal-to-interference ratio at the receiver, a unidirectional and bidirectional path loss balance correlation is constructed to eliminate spatial distance variables. The coordinated power is obtained based on the measured signal received power, the preset radar cross section constant, spatial constant, transmit gain, receive gain, wavelength, and target signal-to-interference suppression ratio.

6. The radar signal detection, identification, and spectrum coordination method based on time-frequency characteristics according to claim 1, characterized in that, The evaluation weights are obtained by performing a weight matrix mapping based on the identified category, including: Obtain a baseline weight matrix containing multiple sets of different bias ratios, wherein the number of column vectors in the baseline weight matrix is ​​consistent with the total number of preset recognition categories; Convert the identified categories into one-hot encoded column vectors in numerical form; Perform matrix multiplication of the baseline weight matrix and the one-hot encoded column vector, and output an evaluation weight vector consisting of quality loss weights and power suppression weights.

7. The radar signal detection, identification, and spectrum coordination method based on time-frequency features according to claim 6, characterized in that, The signal state variables in the linear power domain are acquired to calculate the coordination evaluation factor. Based on the coordination evaluation factor and evaluation weights, a comprehensive coordination effect score is obtained. Closed-loop feedback control is then executed based on the comprehensive coordination effect score, including: Obtain the linear signal-to-noise ratio of the target radar signal before and after the coordinated signal radiation, and obtain the signal quality loss factor based on the ratio of the two. Obtain the equivalent coordinated linear power at the receiver and the target echo linear power, and obtain the coordination suppression factor based on the ratio of the two. Based on the negative exponential function mapping mechanism, the coordination inhibition factor, which ranges from zero to positive infinity, is mapped to the normalized constraint interval. The signal quality loss factor is multiplied by the quality loss weight, and then multiplied by the mapped coordination suppression factor with the same power suppression weight. The result is then weighted and summed to obtain the comprehensive coordination effect score. When the overall coordination effect score is lower than the preset control threshold, an iterative step feedback instruction is generated to increase the coordination power output.

8. A radar signal detection, identification, and spectrum coordination system based on time-frequency characteristics, characterized in that, include: The radar signal detection and interception module is used to acquire a two-dimensional time-frequency matrix of a preset frequency band, generate a time-frequency Boolean matrix based on the ratio of local energy gradient to background noise energy, and obtain the signal parameters of the target radar signal based on the boundary projection of the independent connected domains of the time-frequency Boolean matrix. The signal recognition module is used to construct a heterogeneous feature vector containing continuous physical features and discrete logical features based on signal parameters, and to perform hybrid distance metric calculations based on a preset template feature library to obtain the recognition category of the target radar signal. The parameter measurement module is used to perform fusion calculations based on independent frequency domain measurements and independent time domain measurements, combined with their respective error variance inverse weights, to obtain parameter measurement results. The direction finding module is used to acquire the phase difference sequence between the receiving channels and to obtain the direction of arrival of the signal wave based on the phase difference sequence and the spatial coordinate system. The spectrum coordination module constructs a unidirectional and bidirectional path loss balance correlation based on the measured signal received power and path loss parameters, deduces the coordination power, and generates the corresponding coordination signal. The effect evaluation module is used to obtain the signal state quantities in the linear power domain to calculate the coordination evaluation factor, perform weight matrix mapping according to the identified category to obtain the evaluation weight, obtain the comprehensive coordination effect score according to the coordination evaluation factor and the evaluation weight, and perform closed-loop feedback control according to the comprehensive coordination effect score.

9. The radar signal detection, identification, and spectrum coordination system based on time-frequency characteristics according to claim 8, characterized in that, The radar signal detection and interception module is further used for: The discrete-time signal is acquired and a discrete Fourier transform is performed to construct a two-dimensional time-frequency matrix. The local energy gradient of the central cell within the two-dimensional sliding detection window is calculated. When the ratio of the local energy gradient to the background noise energy of the surrounding reference cells reaches a set threshold, the state of the corresponding coordinate point is determined to be logically true, and a time-frequency Boolean matrix is ​​generated. The time-frequency Boolean matrix is ​​marked with connected components to obtain independent connected components. Orthogonal projections are performed on the time axis and frequency axis respectively. The time difference between the start and end points of the non-zero interval of the projection on the time axis is used as the pulse width, and the upper and lower bound frequency difference of the non-zero interval of the projection on the frequency axis is used as the signal bandwidth.

10. The radar signal detection, identification, and spectrum coordination system based on time-frequency characteristics according to claim 8, characterized in that, The effectiveness evaluation module is further used for: Obtain the linear signal-to-noise ratio of the target radar signal before and after the coordinated signal radiation to obtain the signal quality loss factor; obtain the ratio of the equivalent coordinated linear power at the receiver to the target echo linear power to obtain the coordinated suppression factor; The coordination suppression factor is mapped to the normalized constraint interval based on the negative exponential function mapping mechanism; the signal quality loss factor, the mapped coordination suppression factor and the evaluation weight obtained by mapping are weighted and summed to calculate the comprehensive coordination effect score; when the comprehensive coordination effect score is lower than the preset control threshold, an iterative step feedback instruction for increasing the coordination power is generated.

Images