Waveform parameter extraction and database building method and system for phased array radar radiation source signals

Abstract

The invention discloses a waveform parameter extraction and database building method for phased array radar radiation source signals, and relates to the technical field of radar signals, and the method comprises the steps: extracting a plurality of to-be-measured paths and pulse widths corresponding to each wave position for pulse signals of each wave position in the phased array radar radiation source signals; according to all the to-be-measured paths corresponding to each beam position and each target hidden Markov model, determining a PRI stagger pattern of each beam position; determining the wave position feature vectors belonging to the same waveform pattern and the grouping result of each waveform pattern according to each wave position feature vector and a preset wave position pulse width interval; and taking each waveform pattern grouping result as each waveform unit, modeling the characteristic parameters of each waveform unit, determining a characteristic parameter model of each waveform unit, and constructing a radar waveform library. According to the invention, a systematic description system can be established for the phased array radar radiation source signal features, and the subsequent radar working mode recognition task can be better completed.

Description

Technical Field

[0001] The present application relates to the technical field of radar signal, and particularly to a method and system for extracting waveform parameters and building a library of phased array radar radiation source signals. Background Art

[0002] After nearly a decade of rapid development of active phased array technology, phased array radars have shown obvious advantages in terms of scanning speed, accuracy, reliability, and flexibility. However, for phased array radar radiation source signals, their working environment and signal characteristics bring many challenges. The current related technologies can only achieve pulse clustering through the characteristic parameters of a single pulse, and have no knowledge of what waveforms there are for specific radar signals. Therefore, phased array radar signals cannot be comprehensively described, which in turn makes it difficult to complete the subsequent sorting and recognition tasks of radar working modes. Summary of the Invention

[0003] The purpose of the present application is to provide a method and system for extracting waveform parameters and building a library of phased array radar radiation source signals, which can build a radar waveform library, achieve a comprehensive description of phased array radar signals, and thus better complete the subsequent sorting and recognition tasks of radar working modes.

[0004] To achieve the above purpose, the present application provides the following solutions:

[0005] In the first aspect, the present application provides a method for extracting waveform parameters and building a library of phased array radar radiation source signals, including: for the pulse signals of each wave position in the phased array radar radiation source signals, extracting multiple paths to be measured and pulse widths corresponding to each wave position; each path to be measured includes multiple PRI stagger values and the timing rules between the PRI stagger values; determining the PRI stagger pattern of each wave position according to all the paths to be measured corresponding to each wave position and each target hidden Markov model; the target hidden Markov model is trained according to multiple historical paths; each target hidden Markov model represents a PRI stagger pattern; the PRI stagger pattern characterizes the historical basic path to which the path to be measured belongs; taking the PRI stagger pattern of each wave position and the pulse width of each wave position as the feature vectors of each wave position, and determining the feature vectors of the wave positions belonging to the same waveform pattern and the grouping results of each waveform pattern according to the feature vectors of each wave position and the preset wave position pulse width interval; taking the grouping results of each waveform pattern as each waveform unit, and modeling the characteristic parameters of each waveform unit to determine the characteristic parameter model of each waveform unit; the characteristic parameters of the waveform unit at least include the frequency, pulse width, PRI stagger value, bandwidth, modulation type, waveform connection time, number of waveform pulses, and number of waveform occurrences corresponding to the waveform pattern; constructing a radar waveform library based on the characteristic parameter models of each waveform unit.

[0006] Second aspect, the present application provides a waveform parameter extraction and library building system for phased array radar radiation source signals; an extraction module, configured to extract a plurality of paths to be measured and pulse widths corresponding to each wave position for the pulse signals of each wave position in the phased array radar radiation source signals; each path to be measured includes a plurality of PRI stagger values and the timing rules between the PRI stagger values; a determination module, configured to determine the PRI stagger pattern of each wave position according to all the paths to be measured corresponding to each wave position and each target hidden Markov model; the target hidden Markov model is trained according to a plurality of historical paths; each target hidden Markov model represents a PRI stagger pattern; the PRI stagger pattern characterizes the historical basic path to which the path to be measured belongs; the historical basic path is determined based on a plurality of historical paths; a waveform pattern extraction module, configured to use the PRI stagger pattern of each wave position and the pulse width of each wave position as the feature vectors of each wave position, and determine the wave position feature vectors belonging to the same waveform pattern and the grouping results of each waveform pattern according to the wave position feature vectors and a preset wave position pulse width interval; a modeling module, configured to use the grouping results of each waveform pattern as each waveform unit, and model the characteristic parameters of each waveform unit to determine the characteristic parameter model of each waveform unit; the characteristic parameters of the waveform unit at least include the frequency, pulse width, PRI stagger value, bandwidth, modulation type, waveform connection time, number of waveform pulses, and number of waveform occurrences corresponding to the waveform pattern; a library building method module, configured to build a radar waveform library based on the characteristic parameter models of each waveform unit.

[0007] According to the specific embodiments provided by the present application, the following technical effects are disclosed in the present application:

[0008] For the pulse signals of each wave position in the phased array radar radiation source signals, this application extracts multiple paths to be measured and pulse widths corresponding to each wave position, that is, the characteristic parameters of the subsequent waveform patterns are extracted and the radar waveform library is established in units of wave positions, avoiding the problem that the related technology can only achieve pulse clustering through the characteristic parameters of a single pulse. Then, according to the hidden Markov models of each target trained by multiple historical paths, the PRI stagger patterns of each wave position are determined. Furthermore, the PRI stagger patterns of each wave position and the pulse widths of each wave position are used as the feature vectors of each wave position. In this way, two important characteristic parameters in the waveform pattern are extracted, and the feature vectors of the wave positions belonging to the same waveform pattern and the grouping results of each waveform pattern are determined. Further, the grouping results of each waveform pattern are used as each waveform unit, and the characteristic parameters of each waveform unit are modeled to determine the characteristic parameter model of each waveform unit; finally, based on the characteristic parameter models of each waveform unit, the constructed radar waveform library is obtained by extracting the characteristic parameters of the waveform units from the phased array radar signals in units of wave positions. Among them, the parameter feature extraction of the waveform unit is to perform parameter modeling on the characteristics of each dimension of the pulse group in units of waveforms and calculate its statistical law characteristics, and then store the parameter model in the radar waveform library to achieve unified management of different radar models. This application can establish a systematic description system for the characteristics of phased array radar radiation source signals and better complete the subsequent radar working mode recognition tasks. Brief Description of the Drawings

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0010] Figure 1 It is a schematic flowchart of a method for extracting waveform parameters and building a library for phased array radar radiation source signals provided in an embodiment of the present application;

[0011] Figure 2 It is a technical approach diagram for extracting waveform parameters and building a library for phased array radar radiation source signals provided in an embodiment of the present application;

[0012] Figure 3 It is a schematic diagram of a visual graphic interface assisting manual selection of PRI stagger values provided in an embodiment of the present application;

[0013] Figure 4 It is a distribution diagram of PRI stagger values before and after mapping provided in an embodiment of the present application; Figure 4 Among them, (a) is the distribution diagram before PRI stagger value mapping; Figure 4 Among them, (b) is the distribution diagram after PRI stagger value mapping;

[0014] Figure 5 This is a schematic diagram of the extracted waveform grouping result provided in the embodiment of the present application. Detailed implementation manners

[0015] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0016] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below in conjunction with the accompanying drawings and specific implementation manners.

[0017] Currently, existing algorithms can only achieve pulse clustering through the characteristic parameters of a single pulse and have no knowledge of which waveforms a specific radar has. In order to more comprehensively describe phased array radar signals and better complete subsequent radar working mode recognition tasks, it is necessary to extract waveform unit features of phased array radar signals in units of wave positions and establish a radar waveform library. For the extraction of waveform unit features, it is necessary to perform parameter modeling on the features of each dimension of the pulse group in units of waveforms and calculate their statistical law features, and then store the parameter model in the radar waveform library to achieve unified management of different radar models. Through the above method, a systematic description system can be established for the characteristics of phased array radar radiation source signals, so that when facing the problem of overlapping signal characteristic parameters, deeper associations between different pulses and different parameters can be used to provide a more distinguishable feature for subsequent radar signal sorting and recognition. At the same time, in the face of the situation where signal pulses may be incomplete or lost, through the mutual association between pulse signals in the same waveform, the lost pulse signals can be complemented to a certain extent in the parameter dimension, reducing the interference of signal loss on subsequent radar sorting and recognition.

[0018] As Figure 1 shown, the present application provides a method for extracting waveform parameters and building a library of phased array radar radiation source signals, including:

[0019] Step 101: For the pulse signals of each wave position in the phased array radar radiation source signal, extract multiple paths to be measured and pulse widths corresponding to each wave position; each path to be measured includes multiple PRI stagger values and the timing rules between the PRI stagger values.

[0020] Among them, the Pulse Repetition Interval (PRI) refers to the time interval between two consecutive pulses. Different radar models may have different PRI settings, and even the same radar model may adjust the PRI under different operating modes. The change of PRI can affect the range resolution and the maximum unambiguous range of the radar.

[0021] The Pulse Width (PW) refers to the duration of a single pulse. A shorter pulse width can provide better range resolution but will reduce the energy of the signal. Different radar models and application scenarios may select different pulse width settings.

[0022] Step 102: Determine the PRI stagger pattern of each wave position according to all the paths to be measured corresponding to each wave position and each target Hidden Markov Model; the target Hidden Markov Model is trained based on multiple historical paths; each target Hidden Markov Model represents a PRI stagger pattern.

[0023] Among them, the PRI stagger pattern characterizes the historical basic path to which the path to be measured belongs.

[0024] In some embodiments, step 102 specifically includes: Step 201: For any path to be measured corresponding to any wave position, determine the target Hidden Markov Model that matches the path to be measured in the model library; Step 202: The model library includes each target Hidden Markov Model; Step 203: Take the PRI stagger pattern represented by the matching target Hidden Markov Model as the PRI stagger pattern of the path to be measured; Step 204: Determine the PRI stagger pattern of each wave position according to the PRI stagger patterns of all the paths to be measured.

[0025] In some embodiments, before step 201, steps 301 - 305 are further included.

[0026] Step 301: Obtain all the historical paths corresponding to each wave position; each historical path includes multiple historical PRI stagger values and the timing rule between each historical PRI stagger value.

[0027] Step 302: Cluster all the historical paths according to the preset cluster range to determine multiple clusters.

[0028] Step 303: For any cluster, select the historical path with the highest occurrence frequency and the most types of PRI stagger values among the multiple historical paths corresponding to the cluster as the historical basic path.

[0029] Step 304: Take the historical basic path of each cluster as the PRI stagger pattern corresponding to each cluster.

[0030] Step 305: Use the PRI stagger pattern corresponding to each cluster as the PRI stagger pattern corresponding to each target Hidden Markov Model.

[0031] In some embodiments, after step 302, it further includes: using the multiple historical paths corresponding to any cluster as a training set, training a Hidden Markov Model using the log-likelihood function, and determining the trained Hidden Markov Model corresponding to the cluster; using the trained Hidden Markov Models corresponding to each cluster as the respective target Hidden Markov Models.

[0032] In some embodiments, the log-likelihood function is:

[0033] where Q is the hidden state sequence, λ is the parameter of the Hidden Markov Model; logP(O|λ) is the log-likelihood value between the path and the Hidden Markov Model; O is the observation sequence corresponding to each PRI stagger value in the path; is the total probability of observing the observation sequence O given the parameter λ of the Hidden Markov Model.

[0034] In some embodiments, step 201 specifically includes steps 401 - 402.

[0035] Step 401: For any one of the multiple paths to be measured, use the forward algorithm to calculate the log-likelihood degree between the path to be measured and each target Hidden Markov Model.

[0036] Step 402: Based on the log-likelihood degree, determine the target Hidden Markov Model corresponding to the path to be measured.

[0037] In some embodiments, step 402 specifically includes: using the dynamic time warping algorithm to calculate the distance between the test path and each historical base path; determining the target Hidden Markov Model corresponding to the path to be measured according to the log-likelihood degree and the distance between the test path and each historical base path.

[0038] Step 103: Use the PRI stagger pattern of each wave position and the pulse width of each wave position as the feature vector of each wave position, and determine the feature vector of the wave positions belonging to the same waveform pattern and the grouping result of each waveform pattern according to the feature vector of each wave position and the preset wave position pulse width interval.

[0039] In some embodiments, step 103 specifically includes: determining the wave position feature vectors that simultaneously satisfy the first preset condition and the second preset condition as the wave position feature vectors belonging to the same waveform pattern; the first preset condition is to screen out the wave position feature vectors with exactly the same PRI stagger pattern among the PRI stagger patterns of each wave position according to the PRI stagger pattern of each wave position; the second preset condition is to screen out the wave position feature vectors belonging to the same pulse width interval based on the preset wave position pulse width interval and the pulse width of each wave position; the preset wave position pulse width interval includes multiple pulse width intervals.

[0040] Step 104: Use the grouping results of each waveform pattern as each waveform unit, and model the characteristic parameters of each waveform unit to determine the characteristic parameter model of each waveform unit; the characteristic parameters of the waveform unit at least include the frequency, pulse width, PRI stagger value, bandwidth, modulation type, waveform connection time, number of waveform pulses, and number of waveform occurrences corresponding to the waveform pattern;

[0041] Step 105: Based on the characteristic parameter models of each waveform unit, construct a radar waveform library, which is used for sorting and identifying phased array radar radiation sources and analyzing working modes.

[0042] Among them, the radar waveform library models and describes the radar radiation source characteristics through mathematical language for convenient unified management and research, and can provide sorting and inference basis for subsequent signal processing work such as sorting and identifying radar radiation sources and analyzing working modes.

[0043] In practical applications, the main steps of this application are divided into four parts: PRI stagger time sequence relationship extraction, wave position characteristic parameter clustering, waveform unit parameter modeling, and radar waveform library establishment to carry out specific research work, which solves the problems of waveform parameter extraction and library building of phased array radar signals, as Figure 2 shown.

[0044] (1) PRI stagger time sequence relationship (i.e., PRI stagger pattern) extraction.

[0045] For the phased array radar pulse group of the same waveform, its PRI stagger pattern is complex and diverse, such as fixed, staggered, etc. Since the PRI stagger pattern is the key feature for extracting the waveform, it is necessary to obtain and record the PRI stagger value and its time sequence law, that is, the PRI stagger pattern.

[0046] For the PRI stagger pattern, the next PRI value (i.e., the PRI stagger value) depends on the previous PRI value. Considering each PRI value as a state, then it meets the requirement that the next state of the Hidden Markov Model (HMM) only depends on the previous state. Therefore, it can be represented by a Gaussian Hidden Markov Model, and its parameter space is Θ = (A, B, π), where A = [a ij ​M×M is the transition matrix of M states, π = (π1, π2,..., π M ) is the initial state distribution, represents the Gaussian model of each state, that is The PRI sequence P k = {p1, p2,..., p k} has the probability distribution ξ(P k ) as follows:

[0047]

[0048] Among them, a ij is the state transition probability, indicating the probability that the hidden state transfers from state i to state j at the next moment, that is, the next state is j and the current state is i; i is the starting state index. In a ij , it represents the hidden state number at the current moment t, ranging from 1 to M; j is the target state index, and in a ij , it represents the hidden state number at the next moment t + 1, ranging from 1 to M; is the observation (emission) probability distribution of state m. It describes the probability distribution followed by the observed PRI value (p k ) when the system is in the hidden state m. Here it is a Gaussian (normal) distribution, that is μ m is the mean of the transition matrix of the m-th state; is the variance of the transition matrix of the m-th state; p k is the k-th observed value. The k-th PRI value in the sequence, p k = {p1, p2....p k} is the observed PRI sequence, containing k PRI values; α1(m) is the forward probability belonging to a certain state m at the initial moment (t = 1); α k (m) is the forward probability belonging to a certain state m at the final moment (t = k), referring to the joint probability of observing the entire sequence p k and being in state m at moment k; α t+1 (m) is the forward probability belonging to a certain state m at moment t + 1; a jm is the state transition probability. It has the same meaning as the previously defined a ij , only with different index letters. It is used in the recurrence formula of α t+1 (m); is the probability density of observing p t+1 in state m. That is, it represents the probability density value of observing the PRI value p t+1 when the hidden state is m; p t+1 is the observed value at the next moment t + 1; The probability density of observing p1 in state m. It represents the probability density value of observing the first PRI value p1 when the hidden state is m; π m Is the initial probability distribution of state m; α t (j) is the forward probability of state j at time t, that is, when calculating α t+1 (m), the forward probabilities of all possible previous states j at time t are required; t is the time index. It represents the time step or serial number in the sequence (for example, the t-th PRI value); m is the state index, representing one of the M hidden states (from 1 to M), m = 1, 2,..., M; M is the total number of hidden states, representing the number of different hidden states defined in the model; Is the probability density function of the Gaussian distribution, that is, the PRI stagger pattern is recorded as the state transition path. That is, each PRI stagger pattern can be regarded as a path, and the state of each node on this path is the PRI value. The specific process of PRI stagger pattern extraction and classification based on HMM is as follows:

[0049] Step 1: Data preprocessing: Path extraction and typical value mapping.

[0050] ① Extract path data.

[0051] Extract the PRI stagger values of all paths from the training set. Assume that the training set contains N paths, and the path set is denoted as Each path P i Is represented as a node sequence:

[0052] Where, x j Represents the j-th PRI typical value in the path; n i Is the number of PRI typical values.

[0053] ② Visualization graphical interface to assist in manually selecting possible typical values of PRI parameters and typical value mapping.

[0054] Design a visualization graphical interface software that can intuitively display all PRI values in all wave positions, and then manually select the range of each possible PRI typical value cluster, and calculate the mean value of the node values of each cluster as its typical value. The visualization graphical interface is as follows Figure 3 As shown.

[0055] The explanation of the value is that the wave position is the spatial position pointing of the electromagnetic wave beam. In order to complete a certain task, the radar will send several radar pulses to a wave position, and the set of parameter patterns of these pulse sequences is called the waveform. The phased array radar continuously sends groups of pulses to each wave position. Therefore, the radar echo received by the receiver contains the pulse sequences of each wave position. In actual engineering, there will be errors in estimating the echo radar parameters. In this scenario, the PRI is calculated based on the difference in the time of arrival (toa). Therefore, it will be affected by error factors such as signal-to-noise ratio, local clock calibration, and toa extraction algorithms. These factors will cause a Gaussian distribution around the same PRI value. When extracting the PRI timing relationship, in order to avoid the influence of numerical differences caused by errors, it is necessary to map the actual PRI to a typical value and use the typical value to represent the PRI values within a certain range.

[0056] After obtaining the possible typical PRI values, they are stored in a unified data file. For each PRI value in the PRI path corresponding to each wave position, the likelihood distance is calculated to determine whether it can be mapped to a certain typical value. After the mapping is completed, the typical value mapping path is obtained. The PRI parameter distribution diagrams before and after the mapping are as Figure 4 shown.

[0057] Step 2: Basic path extraction. Select the most representative path as the basic path.

[0058] ① Path style clustering.

[0059] Regard the node sequence of the path as the feature vector F i , and perform similarity comparison clustering on all path features {F1, F2,..., F N} according to the rule of whether the numerical values of the path nodes (i.e., PRI stagger values) are the same, that is, N cluster clustering clusters are obtained:

[0060] Among them, each cluster represents a PRI stagger style.

[0061] ② Determine the basic path.

[0062] The principles for determining the basic path are as follows:

[0063] Frequency analysis: Calculate the occurrence frequency of each path and select the path with the highest frequency.

[0064] Node integrity: Select the path with the largest number of node types and the most complete path structure as the basic path.

[0065] Extract the basic paths from each path clustering cluster according to frequency analysis and node integrity, and select the path with the most occurrences and the most complete nodes as the basic path to represent this PRI stagger pattern.

[0066] Let B m represent the basic path in the m-th clustering cluster:

[0067] where len(P i ) is the path length and K weight is the balance coefficient.

[0068] ③ Manually process the paths with pulse loss and merging.

[0069] For the cases where the path structure changes due to pulse loss or merging in the test path and cannot be correctly assigned to a certain cluster, it is manually analyzed and processed and the cluster is updated.

[0070] Manual processing: Combine the path features and context information to manually determine which basic path cluster the test path should belong to.

[0071] Cluster update: Merge the paths confirmed manually into the corresponding cluster, and update the path set and basic path of the cluster.

[0072] Step 3: HMM model training.

[0073] To simulate the error factor model in actual applications, the input is the original path rather than the mapped path. The role of the mapped path is: first, to extract the basic path as a representative; second, to facilitate creating the HMM model training set for path clustering. One PRI stagger pattern corresponds to one HMM model. The stagger value and timing law can be described through the HMM model.

[0074] ① Prepare the training data.

[0075] Organize the mapped path data into the HMM training format, and regard each node value in the path as the observation sequence O, O = {o1, o2, …, o Num} where o Num is the Num-th node value in the observation sequence; Num is the number of all typical values in all paths in the training set.

[0076] ② Train the model.

[0077] Number of hidden states: Set to the number of typical values that appear in the training set.

[0078] Training objective: Maximize the following log-likelihood function:

[0079] Among them, Q is the hidden state sequence, and λ is the parameter of the hidden Markov model; logP(O|λ) is the log-likelihood value between the path and the hidden Markov model; O is the observation sequence corresponding to each PRI stagger value in the path; is the total probability of observing the observation sequence O under the given parameter λ of the hidden Markov model.

[0080] The Baum-Welch algorithm is used to estimate the HMM parameter λ. Among them, the HMM parameters after training are (A, pi, B_means, B_variance). Among them, A is the transition relationship matrix between PRI typical values (i.e., states), and B is the probability distribution corresponding to the typical values, including B_means and B_variance; here, the Gaussian distribution includes two parameters, the mean (B_means) and the variance (B_variance), and pi is π.

[0081] ③ The PRI stagger patterns are statistically stored in the database.

[0082] After all the PRI stagger patterns in the training set are trained, the HMM models corresponding to the obtained PRI stagger patterns are numbered in a unified data format and stored in the radar waveform library. The HMM model parameters after training.

[0083] Specifically, the specific method for calculating the similarity between the test path and each HMM model in the model library is as follows:

[0084] Suppose the test path P test , calculate its similarity with each HMM model in the model library, and the similarity calculation steps are as follows:

[0085] Step 1. Log-likelihood of HMM: Use the forward algorithm to calculate the log-likelihood of the test path: First, initialize α1(i) = π i b i (o1), 1 ≤ i ≤ N. Among them, α1(i) is the forward probability of state i at the initial time; b i (o1) is the generation probability of state i for the observed value o1; π i is the probability of being in state i at the initial time. Among them, the forward probability: 1 < t ≤ Len ob , 1 ≤ j ≤ N.

[0086] Among them, a ji is the transition probability from state j to state i, that is, the j-th row and i-th column of A, and b i (o t ) is the generation probability of state i for the observed value o t . Among them, the state is the PRI typical value. Len obis the length of the observation sequence, and N is the number of hidden states. Among them, the observed value is the PRI stagger value. α t-1 α(i) is the forward probability of state i at the previous moment t - 1; α t α(j) is the forward probability of state j at time t.

[0087]

[0088] Among them, μ j and σ j are the mean and variance parameters of the Gaussian distribution corresponding to state j, respectively.

[0089] The log-likelihood function is:

[0090] Among them, P(O|λ) is the total probability of observing the observation sequence O under the condition of the given model parameter λ; logP(O|λ) is the log-likelihood value of the basic path of the observation sequence O relative to the given model parameter λ.

[0091] It should be noted that is the joint probability of observing the sequence O at time t and state i; among them, is the sum of the forward probabilities of reaching all possible states i (from 1 to N, that is, the total number of states N) at the last moment T, and what is obtained is the total probability P(O|λ) of observing the entire observation sequence O. That is, the final output of the forward algorithm; among them

[0092] Step 2: Path dynamic similarity: Use the Dynamic Time Warping (DTW) algorithm to calculate the distance between the test path and the basic path:

[0093] Among them, DTW(P test , B m ) is the distance between the test path and the basic path; is the value of the i-th node in the test path; is the value of the j-th node in the basic path.

[0094] Among them, P aligned is the aligned path.

[0095] Step 3: Comprehensive scoring: Combine the HMM log-likelihood and the dynamic similarity of the path to calculate the total score Score(P test , B m ) of the path to be tested belonging to each basic path.

[0096] Score(P test , B m ) = w1·logP(Otest |λ m ) - w2·DTW(P test , B m )。

[0097] Among them, P(O test |λ m ) is the probability of observing the path O to be measured under the condition of the given model parameter λ m ; w1 and w2 are both weight coefficients, taking values in the range of 0 - 1, and the sum of the two weight coefficients = 1. test

[0098] Step 4: Path classification decision: Classify the test path into the basic path with the highest comprehensive score and mark the corresponding PRI stagger pattern as the HMM model number corresponding to this basic path.

[0099]

[0100] If the comprehensive scores of all basic paths are relatively low, mark the PRI stagger pattern number as unknown, and then manually confirm whether to create a new basic path.

[0101] (2) Pulse position feature parameter clustering.

[0102] Based on the pulse width distribution range and the PRI stagger pattern, perform feature parameter clustering on each divided pulse position to determine the possible waveform patterns. The implementation plan is as follows:

[0103] ① Construction of pulse position feature vector.

[0104] Each pulse position feature vector will be represented by a pulse width interval and a PRI stagger pattern. Among them, the pulse width interval is determined according to the established pulse width prior knowledge base to determine which pulse width interval the pulse width of the pulse position belongs to, and is represented as the pulse width interval number; the PRI stagger pattern is represented by the number of the PRI stagger pattern, and discrete numbers are used.

[0105] ② Clustering conditions and determination.

[0106] The conditions for a pulse position to be considered as the same waveform are: the mean pulse width belongs to the same interval, and the mean of the pulse widths of the pulse positions needs to fall within the same pulse width interval; the PRI stagger pattern numbers are the same, and the PRI stagger pattern numbers of the pulse positions need to be exactly the same.

[0107] Based on these conditions, perform waveform pattern determination, which can be carried out according to the following steps:

[0108] Pulse width interval matching: For each pulse position, check which pulse width interval its mean pulse width belongs to.

[0109] ​PRI Jitter Pattern Matching: Compare the PRI jitter pattern numbers of each dwell with those of other dwells to ensure consistent PRI jitter patterns. If the average pulse width and PRI jitter pattern number of a dwell are the same, they are considered to belong to the same waveform pattern.

[0110] Waveform Pattern Labeling: Label each dwell with a waveform pattern number. Finally, output the waveform pattern labels to which each dwell belongs. These labels represent the waveform pattern grouping results of each dwell under the same conditions of pulse width and PRI jitter pattern number, as Figure 5 shown.

[0111] (3) Modeling of Waveform Unit Parameters.

[0112] To systematically and completely describe a waveform pattern, only the pulse width and PRI parameters were used in the previous clustering. However, to fully describe the waveform, other characteristic parameters need to be modeled and described in mathematical terms. For example, to quickly find a certain style of clothing from a pile of clothes, some prominent features such as patterns and colors can be used. But to fully describe the style of the clothing, other features such as buttons, materials, and necklines are also needed to completely express a style of clothing.

[0113] (4) Each characteristic parameter of the radiation source waveform unit can be described by a binary tuple Θ = <M, θ>, where M ∈ Μ, Μ represents the type space of the characteristic parameter, and θ represents the characteristic parameter value.

[0114] Suppose a radar radiation source has p characteristic parameters X is the characteristic parameter sequence, and the signal of a radiation source has p characteristic parameter sequences; Transpose means transpose. Then the characteristic parameters of the i-th waveform unit of the radiation source are Model the CPI(i) of the i-th waveform unit style of the radiation source as:

[0115] Among them, represents the model of the p-th characteristic parameter of the i-th waveform unit style of the radiation source, and there is:

[0116]

[0117] Among them, is the parameter value sequence of the p-th characteristic parameter of the i-th waveform unit style of the radiation source, represents the description sequence The m-th descriptor sequence in, using the subsequence method to describe the characteristic parameter is because the temporal variation characteristics of the subsequence can be used to reflect the temporal and joint variations of the characteristic parameter, is the variance sequence corresponding to the parameter value sequence, To describe The sequence number description sequence of the number of repetitions of the subsequence within a single operating mode period of the radar, Indicates The m-th subsequence in The number of repeated occurrences within a single operating mode period, Is The type description sequence of the subsequence in Express The m-th subsequence in The parameter type, such as fixed carrier frequency, staggered pulse repetition frequency, etc. The parameter value sequence Is the core of the joint parameter modeling, and its subsequence Can be modeled as:

[0118]

[0119] Among them, case1, case2, and case3 are three different types of parameter value sequences; case1 is a fixed parameter value sequence type, x1 is a fixed value, that is, this parameter value sequence has only one parameter value, so this parameter value sequence is represented by x1 to refer to that one fixed value; Case2 is a staggered parameter value sequence type; x1, x2,... x q Are q parameter values in the parameter value sequence; Case3 is an empty set; φ represents parameter incompleteness and is represented by an empty set because it cannot be modeled.

[0120] (4) Radar waveform library establishment.

[0121] Statistically analyze the waveform unit styles of radar emitters of each model, and store the characteristic parameter models modeled by each waveform unit style into the radar waveform library corresponding to their respective models for unified management. Among them, the waveform unit mentioned above is the waveform unit style.

[0122] The waveform unit parameters in the waveform library include frequency, pulse width, PRI, bandwidth, modulation type, waveform connection time, number of waveform pulses, and number of waveform occurrences. Among them, the two parameters of pulse width and PRI within the waveform are the main factors for waveform extraction. After extracting the waveform through the two parameters of pulse width and PRI within the waveform, calculate the number of waveform pulses, waveform dwell time, and connection interval between waveforms for the pulse group belonging to this waveform, and statistically analyze the mean, variance, and other statistical characteristics of all waveform unit parameters in the waveform library and save them in a unified data format in the form of the characteristic parameter model in the previous section.

[0123] In addition to implementing this application in the above manner, the following methods can also be used to implement it.

[0124] (1)In the first step of PRI stagger time series relationship extraction, the visualization graphical interface-assisted PRI typical value mapping can also be implemented using automatic clustering algorithms (such as K-means, DBSCAN) to automatically divide PRI data points into different clusters. However, in the case of complex PRI distributions and the presence of noise, the automatic clustering effect is not good.

[0125] (2)In the clustering scheme of the wave position characteristic parameters combining pulse width and PRI, more dimensions of features such as frequency and arrival angle can also be added to perform clustering using more feature information. However, in practical applications, the pulse width and PRI features have the greatest discrimination in wave position extraction, and other features may introduce redundant information and increase the computational burden.

[0126] (3)In the PRI stagger time series relationship extraction based on HMM, a rule-based expert system can be used to describe the PRI stagger time series relationship. For example, rules for PRI changes are set manually. However, it is difficult to cover all PRI change patterns, and it has poor adaptability and low robustness to complex PRI stagger patterns.

[0127] For the waveform unit feature extraction in this application, it is necessary to perform waveform parameter modeling on the multi-dimensional features of the pulse sequence with wave position as the unit and calculate its statistical law features, and store the parameter model in the radar waveform library to achieve unified management of different radar models. First, extract the PRI stagger time series relationship and perform clustering in combination with other pulse feature parameters (pulse width, carrier frequency, etc.); then, model according to the waveform unit parameters (frequency, pulse width, PRI, bandwidth, modulation type, waveform connection time, number of waveform pulses, number of waveform occurrences) to construct a radar waveform library. Among them, the PRI stagger time series relationship extraction is to model the PRI time series relationship using the hidden Markov model, including preparing training data, setting the number of hidden states according to the number of path typical values, constructing the objective function, and model training; the waveform unit parameter modeling is to perform statistical calculations of the mean and variance of pulse feature parameters such as frequency, pulse width, and PRI for the extracted waveform unit and calculate waveform feature parameters such as waveform connection time, number of waveform pulses, and number of waveform occurrences, and construct the above parameters into a waveform feature set to achieve waveform unit feature parameter modeling.

[0128] Due to the complex and variable signal patterns of phased array radars, as well as problems such as pulse loss and missing dimensions of pulse feature parameters, as a non-cooperating party, it is impossible to obtain the prior information of the distribution of all feature parameters. There will still be some PRI stagger values in the pulse sequences of some wave positions that are not in the PRI typical value library and cannot be mapped. Therefore, it is impossible to perform subsequent waveform extraction on the pulse sequences containing unknown PRI information.

[0129] This application provides a waveform parameter extraction and library building system for phased array radar radiation source signals, including:

[0130] An extraction module, configured to extract, for the pulse signals of each wave position in the phased array radar radiation source signals, a plurality of paths to be measured and pulse widths corresponding to each wave position; each path to be measured includes a plurality of PRI stagger values and the timing rule between the PRI stagger values.

[0131] A determination module, configured to determine the PRI stagger pattern of each wave position according to all paths to be measured corresponding to each wave position and each target hidden Markov model; the target hidden Markov model is trained according to a plurality of historical paths; each target hidden Markov model represents a PRI stagger pattern; the PRI stagger pattern characterizes the historical basic path to which the path to be measured belongs; the historical basic path is determined based on a plurality of historical paths.

[0132] A waveform pattern extraction module, configured to use the PRI stagger pattern of each wave position and the pulse width of each wave position as the feature vectors of each wave position, and determine the feature vectors of the wave positions belonging to the same waveform pattern and the grouping results of each waveform pattern according to the feature vectors of each wave position and a preset wave position pulse width interval.

[0133] A modeling module, configured to use the grouping results of each waveform pattern as each waveform unit, and model the characteristic parameters of each waveform unit to determine the characteristic parameter model of each waveform unit; the characteristic parameters of the waveform unit at least include the frequency, pulse width, PRI stagger value, bandwidth, modulation type, waveform connection time, number of waveform pulses, and number of waveform occurrences corresponding to the waveform pattern.

[0134] A radar waveform library construction method module, configured to construct a radar waveform library based on the characteristic parameter models of each waveform unit.

[0135] It should be noted that the user data (including but not limited to user device data, user personal data, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all data and data that have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of the relevant data need to comply with the relevant regulations.

[0136] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0137] The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logics, data processing logics based on quantum computing, etc., without limitation.

[0138] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.

[0139] In this article, specific examples are used to illustrate the principle and implementation manner of this application. The description of the above embodiments is only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A method for extracting waveform parameters of phased array radar radiation source signals and building a database, characterized in that Including: For the pulse signals of each wave position in the phased array radar radiation source signals, extract multiple paths to be measured and pulse widths corresponding to each wave position; each path to be measured includes multiple PRI stagger values and the timing rules between the PRI stagger values. Based on all the paths to be measured corresponding to each wave position and each target hidden Markov model, determine the PRI stagger pattern of each wave position; the target hidden Markov model is trained based on multiple historical paths; each target hidden Markov model represents a PRI stagger pattern; the PRI stagger pattern characterizes the historical basic path to which the path to be measured belongs. Take the PRI stagger pattern of each wave position and the pulse width of each wave position as the feature vectors of each wave position, and based on the feature vectors of each wave position and the preset wave position pulse width interval, determine the feature vectors of the wave positions belonging to the same waveform pattern and the grouping results of each waveform pattern. Take the grouping results of each waveform pattern as each waveform unit, and model the characteristic parameters of each waveform unit to determine the characteristic parameter model of each waveform unit; the characteristic parameters of the waveform unit at least include the frequency, pulse width, PRI stagger value, bandwidth, modulation type, waveform connection time, number of waveform pulses, and number of waveform occurrences corresponding to the waveform pattern. Based on the characteristic parameter models of each waveform unit, construct a radar waveform library, which is used for sorting and identifying phased array radar radiation sources and analyzing working modes.

2. The method for extracting waveform parameters and building a library of phased array radar radiation source signals according to claim 1, characterized in that Based on all the paths to be measured corresponding to each wave position and each target hidden Markov model, determine the PRI stagger pattern of each wave position, specifically including: For any path to be measured corresponding to any wave position, determine the target hidden Markov model that matches the path to be measured in the model library; the model library includes each target hidden Markov model. Take the PRI stagger pattern represented by the matched target hidden Markov model as the PRI stagger pattern of the path to be measured. Based on the PRI stagger patterns of each path to be measured, determine the PRI stagger pattern of each wave position.

3. The waveform parameter extraction and database building method for phased array radar radiation source signals according to claim 2, characterized in that Before determining the target hidden Markov model that matches the path to be measured in the model library for any path to be measured corresponding to any wave position, it also includes: Obtain all the historical paths corresponding to each wave position; each historical path includes multiple historical PRI stagger values and the timing rules between the historical PRI stagger values. Cluster all the historical paths according to the preset cluster range to determine multiple clusters. For any cluster, select the historical path with the highest frequency of occurrence and the most types of PRI stagger values among the multiple historical paths corresponding to the cluster as the historical basic path. Take the historical basic paths of each cluster as the PRI stagger patterns corresponding to each cluster. Take the PRI stagger pattern corresponding to each cluster as the PRI stagger pattern corresponding to each target hidden Markov model.

4. The method for extracting waveform parameters and building a library of phased array radar radiation source signals according to claim 3, characterized in that, After clustering all the historical paths according to the preset cluster range to determine multiple clusters, it also includes: Take the multiple historical paths corresponding to any cluster as the training set, and use the logarithmic likelihood function to train the hidden Markov model to determine the trained hidden Markov model corresponding to the cluster. Take the trained hidden Markov models corresponding to each cluster as each target hidden Markov model.

5. The waveform parameter extraction and library building method for phased array radar radiation source signals according to claim 2, characterized in that For any path to be measured corresponding to any wave position, determining the target Hidden Markov Model (HMM) that matches in the model library specifically includes: For any path to be measured among multiple paths to be measured, using the forward algorithm to calculate the log-likelihood between the path to be measured and each target Hidden Markov Model; Based on the log-likelihood, determining the target Hidden Markov Model corresponding to the path to be measured.

6. The waveform parameter extraction and library building method for the phased array radar radiation source signal according to claim 5, characterized in that, Based on the log-likelihood, determining the target Hidden Markov Model corresponding to the path to be measured specifically includes: Using the dynamic time warping algorithm to calculate the distance between the test path and each historical basic path; According to the log-likelihood and the distance between the test path and each historical basic path, determining the target Hidden Markov Model corresponding to the path to be measured.

7. The waveform parameter extraction and database building method for the phased array radar radiation source signal according to claim 1, characterized in that According to each wave position feature vector and the preset wave position pulse width interval, determining the wave position feature vectors belonging to the same waveform pattern and the grouping results of each waveform pattern specifically includes: Determining the wave position feature vectors that simultaneously meet the first preset condition and the second preset condition as the wave position feature vectors belonging to the same waveform pattern; the first preset condition is to screen out the wave position feature vectors with exactly the same PRI stagger pattern among the PRI stagger patterns of each wave position according to the PRI stagger pattern of each wave position; the second preset condition is to screen out the wave position feature vectors belonging to the same pulse width interval based on the preset wave position pulse width interval and the pulse width of each wave position; the preset wave position pulse width interval includes multiple pulse width intervals.

8. The waveform parameter extraction and database building method for the phased array radar radiation source signal according to claim 4, characterized in that The log-likelihood function is: Where Q is the hidden state sequence, λ is the parameter of the hidden Markov model; logP(O|λ) is the log-likelihood value between the path and the hidden Markov model; O is the observation sequence corresponding to each PRI stagger value in the path. It is the total probability of observing the observation sequence O given the parameter λ of the hidden Markov model.

9. A waveform parameter extraction and database building system for phased array radar radiation source signals, characterized in that, Including: An extraction module, used to extract multiple paths to be measured and the pulse width corresponding to each wave position for the pulse signals of each wave position in the phased array radar radiation source signal; each path to be measured includes multiple PRI stagger values and the timing law between each PRI stagger value; A determination module, used to determine the PRI stagger pattern of each wave position according to all paths to be measured corresponding to each wave position and each target Hidden Markov Model; the target Hidden Markov Model is trained according to multiple historical paths; each target Hidden Markov Model represents a PRI stagger pattern; The PRI stagger pattern characterizes the historical basic path to which the path to be measured belongs; the historical basic path is determined based on multiple historical paths; A waveform pattern extraction module, used to use the PRI stagger pattern of each wave position and the pulse width of each wave position as each wave position feature vector, and according to each wave position feature vector and the preset wave position pulse width interval, determine the wave position feature vectors belonging to the same waveform pattern and the grouping results of each waveform pattern; A modeling module, used to use the grouping results of each waveform pattern as each waveform unit, and model the characteristic parameters of each waveform unit to determine the characteristic parameter model of each waveform unit; the characteristic parameters of the waveform unit at least include the frequency, pulse width, PRI stagger value, bandwidth, modulation type, waveform connection time, waveform pulse number, and waveform appearance times corresponding to the waveform pattern; A library building method module, used to build a radar waveform library based on the characteristic parameter models of each waveform unit.

Images