Wavelet transform-based multifunction radar state switching point detection method and system

Abstract

The present application relates to a wavelet transform-based multifunctional radar state switching point detection method and system, which comprises the following steps: collecting a multifunctional radar pulse signal sequence with multiple working modes, obtaining a single radar emitter pulse signal sequence after feature parameter extraction and pulse sorting; detecting and processing the single radar emitter pulse signal sequence for missing pulses and false pulses and normalizing to obtain a preprocessed single radar pulse signal sequence; detecting the state switching point based on the multi-dimensional parameters of the preprocessed single radar pulse signal sequence to obtain a multi-dimensional parameter mutation point matrix; determining the final mutation point of the single radar emitter pulse signal sequence based on the multi-dimensional parameter mutation point matrix to realize the state switching point detection of the single radar emitter pulse signal sequence and sequence segmentation based thereon. The single working state radar pulse signal sequence obtained by the present application provides data support for working state recognition.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of radar reconnaissance, and particularly relates to a method and system for detecting a state switching point of a multifunctional radar based on wavelet transform, which can be applied to electronic reconnaissance and electronic countermeasure systems. BACKGROUND

[0002] With the development of electronic technology and the significant increase in the complexity of battlefield electromagnetic environment, electronic warfare has become the main battlefield of modern high-tech war. As one of the important links of electronic reconnaissance, radar mode recognition focuses on extracting the parameter characteristic law contained in the radar signal as the basis for recognition, and then accurately identifying the enemy radar working state and developing threat assessment and tactics.

[0003] However, with the continuous progress of radar technology, the proportion of multifunctional radars is increasing, among which phased array radars with multiple working modes are the most. Compared with traditional radars, multifunctional phased array radars (MPAR) have the characteristics of rapid and agile beam and signal, complex modulation pattern, strong adaptability, etc., and can adjust their working modes and parameters in real time according to the battlefield environment and needs, and can sequentially perform a large number of different tasks. This makes the pulse signals intercepted by the radar reconnaissance system often contain multiple radar working states, the modulation patterns and parameters of the pulse signals in each working state are different, and the duration of the working state is also different. If the continuous pulse sequence without state segmentation processing is directly used as a sample for training or recognition, it will lead to a large result error. Therefore, it is necessary to preprocess the intercepted radar pulse signal sequence, accurately and quickly detect the radar working state switching point by certain means, and segment the radar signal pulse sequence of a single working state, which is of great significance for subsequent working state recognition and interference strategy formulation.

[0004] According to the analysis of the basic working principle of multifunctional radars, multifunctional radars can adaptively schedule tasks, change the beam direction, waveform parameters, etc. according to the current instructions, and convert the working mode, change the frequency, pulse repetition interval (PRI), etc. according to different task requirements. Therefore, the switching of different working states of multifunctional radars must be accompanied by changes in signal parameters, even if there is signal parameter overlap, it is impossible for all parameters to completely overlap. For the received radar signal pulse sequence, selecting appropriate pulse parameters for analysis can determine the position of the switching point, which can provide reasonable and reliable data and important basis for identifying different working modes of multifunctional radars. SUMMARY

[0005] The application aims at intercepted multifunctional radar pulse signal sequences containing various radar working states, which are difficult to be directly used as samples for training and recognition in subsequent intelligent processing, and provides a wavelet transform-based multifunctional radar state switching point detection method and system.

[0006] To achieve the above-mentioned application purposes, the application adopts the following technical solutions:

[0007] A wavelet transform-based multifunctional radar state switching point detection method, characterized by comprising the following steps:

[0008] S1, collecting multifunctional radar pulse signal sequences with various working modes, and obtaining single radar emitter pulse signal sequences after feature parameter extraction and pulse sorting;

[0009] S2, detecting and processing missing pulses and false pulses of the single radar emitter pulse signal sequences and performing normalization to obtain preprocessed single radar pulse signal sequences;

[0010] S3, performing state switching point detection based on multi-dimensional parameters of the preprocessed single radar pulse signal sequences, generating first layer detail coefficients by respectively performing 4-layer discrete wavelet decomposition on each independent dimensional parameter and reconstruction, recording local maximum values of the detail coefficients in each window through a sliding window to obtain a set of possible mutation points of the dimensional parameter, and obtaining a multi-dimensional parameter mutation point matrix by repeating the above operation for each dimensional parameter;

[0011] S4, determining the final mutation point of the single radar emitter pulse signal sequences based on the multi-dimensional parameter mutation point matrix, and realizing state switching point detection of the single radar emitter pulse signal sequences and sequence segmentation based on the same.

[0012] As a preferred solution, in the step S1, the multifunctional radar pulse signal sequences are represented by pulse description word (PDW) sequence composed of extracted feature parameters, and the feature parameters include carrier frequency (RF), pulse width (PW), amplitude (PA), pulse repetition interval (PRI) and duty cycle (DC);

[0013] The multifunctional radar pulse signal sequences are represented by the following vector S:

[0014]

[0015] wherein, PDW i =[RF i ,PW i ,PA i ,PRI i ,DC i ] T i=1,2,…,n represents the pulse description word of the i-th received pulse.

[0016] As a preferred solution, the step S2, the detection and processing of the missing pulse, includes:

[0017] For the i-th value PRI in the PRI sequence i , the average value M of the two PRI before and after it is calculated;

[0018] If PRI i > 2.5M, the current PRI i is considered as an error PRI value generated by two missing pulses, and PRI i / 3 and M are taken in turn to interpolate twice before the error PRI value, and PRI i is corrected by subtracting the interpolated values twice;

[0019] If 1.5M < PRI i < 2.5M, the current PRI i is considered as an error PRI value generated by one missing pulse, and M is taken to interpolate once before the error PRI value, and PRI i is corrected by subtracting M.

[0020] As a preferred solution, the step S2, the detection and processing of the false pulse, includes:

[0021] The average value M2 of the PRI sequence is calculated, and if PRI i < 0.8M2 and PRI i+1 < 0.8M2, the two consecutive PRIs are considered as error PRIs generated by a false pulse, and the two PRI values are added as the correction value of the false pulse and replace the two consecutive PRIs.

[0022] As a preferred solution, for the interpolation of the missing pulse, the average value of the parameters of the two pulses before and after the error pulse is selected as the filling;

[0023] The duty cycle parameter is calculated and updated according to the values after the processing of the false pulse and the missing pulse;

[0024] Finally, the parameters of each dimension are normalized.

[0025] As a preferred solution, the step S3, the detection of the state switching point, includes the following steps:

[0026] S31, for the pre-processed single radar pulse signal sequence, it is considered as a 5-dimensional PDW time sequence X = {x1, x2, …, x i , …, x n} ∈ R 5×n , the sequence has multiple switching points and the positions are unknown, where x ia multi-dimensional parameter vector of the PDW time series at the i-th pulse, n is the length of the time series;

[0027] After the independent discrete wavelet 4-layer haar wavelet decomposition and reconstruction of any one-dimensional parameter sequence using the following formula to generate the detail coefficients, the first layer detail coefficient d1 is selected,

[0028] [d, a] = wavedec(data, 4, db3);

[0029] d1 = wrcoef('d', d, a, db3, 1);

[0030] wherein wavedec(data, 4, db3) represents 4-layer decomposition of data data using db3 wavelet, d represents the detail coefficient after wavelet decomposition, and a represents the approximation coefficient after wavelet decomposition; wrcoef('d', d, a, db3, 1) represents the operation of reconstructing the first layer detail coefficient using db3 wavelet, and d1 represents the reconstructed first layer detail coefficient;

[0031] S32, sliding from the beginning of the first layer detail coefficient sequence generated by reconstruction to the end of the detail coefficient in a fixed window length, recording the position coordinates of the maximum detail coefficient value of each sliding window; wherein the local maximum of the detail coefficient represents a mutation of the signal at this time;

[0032] S33, for each dimension parameter of X = {x1, x2, …, x i ,…,x n} ∈ R 5×n , the steps S31 and S32 are performed to obtain a multi-dimensional parameter mutation point matrix T S :

[0033]

[0034] wherein, 1 ≤ q ≤ l represents the pulse position coordinates corresponding to the maximum value of the detail coefficient of each dimension parameter in the q-th sliding window, l is the number of sliding times of the sliding window.

[0035] As a preferred solution, in the step S4, the final mutation point set of the pulse time series is determined, which is defined as:

[0036]

[0037] wherein y s (1 ≤ s ≤ p) is the s-th final mutation point position, and p is the number of switching points.

[0038] As a preferred solution, in the step S4, the sequence is divided into radar pulse subsequences S' in different working modes according to the obtained state switching points. h h = 1, 2, …, p + 1, p is the number of switching points.

[0039] The application further provides a wavelet transform-based multifunctional radar state switching point detection system applying the detection method according to any one of the preceding solutions, and the detection system comprises:

[0040] A collection module is configured to collect a multifunctional radar pulse signal sequence having multiple working modes.

[0041] A feature extraction and pulse sorting module is configured to extract feature parameters and sort pulses of the multifunctional radar pulse signal sequence, so as to obtain a single radar emitter pulse signal sequence.

[0042] A preprocessing module is configured to detect and process missing pulses and false pulses of the single radar emitter pulse signal sequence and perform normalization, so as to obtain a preprocessed single radar pulse signal sequence.

[0043] A mutation point matrix construction module is configured to detect state switching points based on multi-dimensional parameters of the preprocessed single radar pulse signal sequence, to generate first layer detail coefficients by respectively performing four-layer discrete wavelet decomposition and reconstruction on each independent dimensional parameter, and to obtain a set of possible mutation points of the dimensional parameter by recording local maximum values of the detail coefficients in each window through a sliding window, and to obtain a multi-dimensional parameter mutation point matrix by repeating the above operation for each dimensional parameter.

[0044] A state switching point detection module is configured to determine final mutation points of the single radar emitter pulse signal sequence based on the multi-dimensional parameter mutation point matrix, so as to detect state switching points of the single radar emitter pulse signal sequence.

[0045] A sequence division module is configured to divide the single radar emitter pulse signal sequence according to the state switching points.

[0046] Compared with the prior art, the application has the following technical effects:

[0047] The wavelet transform-based multifunctional radar state switching point detection method of the application is capable of detecting switching points between different working modes through wavelet multi-resolution analysis after preprocessing, in the case that the intercepted complex radar pulse sequence contains multiple working mode changes, so as to divide the pulse sequence data to obtain a single working state radar pulse signal sequence, and to provide reasonable and reliable data for subsequent working state recognition. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1is the overall implementation flowchart of the embodiment of the present application;

[0049] Figure 2 is the processing flowchart of the missing pulse and false pulse of the embodiment of the present application;

[0050] Figure 3 is the partial parameter variation diagram of the pulse sequence of the multifunctional radar in different working modes of the embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to more clearly illustrate the embodiments of the present application, the specific embodiments of the present application will be described below with reference to the drawings. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained from these drawings without creative labor, and other embodiments can also be obtained.

[0052] As shown in Figure 1 , the detection method of the state switching point of the multifunctional radar based on wavelet transform of the embodiment of the present application comprises the following steps:

[0053] S1, obtaining the pulse signal sequence of the multifunctional radar with multiple working states (i.e. working modes) through the electronic reconnaissance system, and obtaining the pulse signal sequence of a single radar radiation source after feature parameter extraction and pulse sorting.

[0054] Among them, the pulse signal sequence of the multifunctional radar can be represented by the pulse description word (Pulse Description Word, PDW) sequence composed of the extracted feature parameters of radio frequency (Radio Frequency, RF), pulse width (Pulse Width, PW), pulse amplitude (Pulse Amplitude, PA), pulse repetition interval (Pulse Repetition Interval, PRI) and duty cycle (Duty Cycle, DC);

[0055] Therefore, the pulse sequence of the multifunctional radar of the embodiment of the present application can be represented by the following vector S:

[0056]

[0057] Among them, PDW i =[RF i ,PW i ,PA i ,PRI i ,DC i ] T , i = 1, 2, …, n represents the pulse description word of the i-th pulse.

[0058] S2, considering the noise, environment and other factors interference caused by the intercepted radar pulse signal missing pulse, false pulse, these two cases have a great relevance to the subsequent radar pulse sequence switching point detection, so before the switching point detection, first of all, the single radar emitter pulse signal sequence is detected and processed and normalized.

[0059] Specifically, as Figure 2 shown, the above step S2 includes the following steps:

[0060] S21, first of all, the detection and processing of missing pulse: for the i th value PRI i in the PRI sequence, calculate the average value M of the two PRI before and after it, if PRI i > 2.5M, the current PRI i can be regarded as two missing pulse generated error PRI value, take PRI i / 3 and M in the error PRI value in turn for two times interpolation, while PRI i subtract the value of two times interpolation after correction; if 1.5M < PRI i <2.5M, the current PRI i can be regarded as a missing pulse generated error PRI value, take M in the error PRI value before interpolation, while PRI i subtract M for correction.

[0061] S22, for the detection and processing of false pulse: first calculate the mean value M2 of PRI sequence, if PRI i <0.8M2 and PRI i+1 <0.8M2, the two consecutive PRI is regarded as false pulse generated error, the two PRI values are added as false pulse correction value and replace the two consecutive PRI value.

[0062] S23, because the missing pulse, false pulse will not affect the rest of the parameters, for the interpolation of missing pulse, the average value of the parameters of the two pulses before and after the error pulse is selected as the filling; the duty cycle parameter is calculated and updated according to the value after the false pulse and missing pulse processing; finally, the normalized processing is carried out on each dimension parameter, and the preprocessed single radar pulse signal sequence is obtained.

[0063] S3, based on the multi-dimensional parameters of the preprocessed single radar pulse signal sequence, the state switching point detection is carried out, the first layer detail coefficient is generated by respectively carrying out 4 layer decomposition and reconstruction on each independent dimension parameter, and the local maximum value of the detail coefficient in each window is recorded by sliding window, the mutation point set that may exist in the parameter is obtained, and the above operation is repeated for each dimension parameter, and the multi-dimensional parameter mutation point matrix can be obtained.

[0064] The specific steps of the above step S3 are as follows:

[0065] S31, considering that there is parameter overlap between different working modes of the multifunction radar, and there is also signal parameter variation within the same working mode, if only the switching point detection is performed on a single parameter of the pulse, it is unreliable, for example, in the case of Figure 3 If only the switching point detection is performed on the PRI parameter, many false switching points in the working mode will be detected and the real switching points will be missed, which will cause great error to the subsequent working mode recognition. In order to avoid the interference of false switching points on the single-dimensional pulse parameter sequence and miss the real switching points as little as possible, the multi-dimensional parameters of the preprocessed pulse sequence will be comprehensively detected. The pulse sequence of the multifunction radar is regarded as a multi-dimensional time sequence, and a method of multi-dimensional time sequence analysis is used for analysis. For the structure change detection of the multi-dimensional time sequence, first, the mutation points on each independent dimension are determined respectively, and the switching points are selected as the final mutation points of the entire multi-dimensional time sequence through significance comparison or voting strategy.

[0066] S32, generally, for any signal f(t)∈L 2 The specific form of the discrete wavelet transform (DWT) of (R) is as follows:

[0067]

[0068] Wherein, j, k are any positive integers, and j is the number of layers of decomposition (scale parameter), k is the translation parameter, is a discrete wavelet function, a0>1, t0>1, from the above formula, it can be seen that the wavelet transform has a unique “zoom” property through scale change and translation change, which can well extract the local frequency band characteristics of the signal, and is very suitable for analyzing non-stationary signals. Since the wavelet transform has the time-frequency localization property, the position and size of the singular point (mutation point) of the signal can be detected and analyzed by the local maximum of the wavelet transform coefficient module.

[0069] S33, for the preprocessed radar pulse PDW sequence, it can be regarded as a 5-dimensional time sequence X={x1,x2,…,x i ,…,x n}∈R 5×n The sequence has multiple switching points and the positions are unknown, wherein x i represents the multi-dimensional parameter vector of the PDW time sequence at i time (i.e. the i th pulse), n is the length of the time sequence.

[0070] The following formula is used to detect any one-dimensional parameter sequence (for example, the carrier frequency {x1RF x2 RF ,…,x i RF ,…,x n RF After performing independent discrete wavelet four-level wavelet decomposition and reconstructing to generate detail coefficients, the first level detail coefficient d1 is selected.

[0071] [d,a]=wavedec(data,4,db3);

[0072] d1 = wrcoef('d', d, a, db3, 1);

[0073] In this code, wavedec(data,4,db3) means performing a 4-level decomposition of the data using the db3 wavelet, where d represents the detail coefficients after wavelet decomposition and a represents the approximation coefficients after wavelet decomposition; wrcoef('d',d,a,db3,1) means using the db3 wavelet to reconstruct and generate the first level of detail coefficients, where d1 represents the first level of detail coefficients after reconstruction.

[0074] S34. Further, starting from the first position of the first layer of detail coefficient sequence generated by reconstruction, a fixed window length is used to slide backward until the window ends at the position containing the detail coefficients. The coordinates of the position of the maximum detail coefficient value in each sliding window are recorded. As can be seen from the analysis in S31, the local maximum value of the detail coefficient indicates that there may be a sudden change in the signal at this moment.

[0075] S35. For X = {x1, x2, ..., x...} i ,…,x n}∈R 5×n Processing each dimension in step S3 yields the multidimensional parameter mutation point matrix T. S :

[0076]

[0077] in, 1≤q≤l represents the pulse position coordinates corresponding to the maximum value of the detail coefficients of each dimension parameter within the q-th sliding window. l represents the number of times the sliding window has been slid.

[0078] S4. The obtained multidimensional parameter mutation point matrix is ​​comprehensively analyzed. The switching point is selected as the final mutation point of the entire multidimensional time series through strategies such as significance comparison or voting. This enables the detection of the state switching point of the radar pulse signal sequence and the sequence segmentation to obtain a single radar state pulse sequence.

[0079] The set of final abrupt change points in the pulse time series is defined as follows:

[0080]

[0081] wherein y s (1≤s≤p) is the s-th final mutation point position, and p is the number of switching points.

[0082] The specific algorithm flow is shown in Table 1.

[0083] Table 1 Algorithm flow

[0084]

[0085] wherein the detection algorithm parameters include a sliding window length, a sliding step, a wavelet function selection, and a decomposition layer number.

[0086] Finally, according to the obtained radar state switching points, the radar pulse sequence is functionally segmented to obtain radar pulse subsequences S' in different working modes. h , h=1,2,…,p+1, and p is the number of switching points.

[0087] Based on the wavelet transform-based multi-functional radar state switching point detection method of the embodiment of the present application, the embodiment of the present application further provides a wavelet transform-based multi-functional radar state switching point detection system, which comprises a collection module, a feature extraction and pulse sorting module, a preprocessing module, a mutation point matrix construction module, a state switching point detection module, and a sequence segmentation module.

[0088] Specifically, the collection module of the embodiment of the present application acquires a multi-functional radar pulse signal sequence with multiple working states (i.e., working modes) through an electronic reconnaissance system.

[0089] The feature extraction and pulse sorting module of the embodiment of the present application is used to extract feature parameters and sort pulses of the multi-functional radar pulse signal sequence to obtain a single radar emitter pulse signal sequence.

[0090] wherein the multi-functional radar pulse signal sequence can be represented by a pulse description word (PDW) sequence composed of extracted feature parameters including a radio frequency (RF), a pulse width (PW), a pulse amplitude (PA), a pulse repetition interval (PRI), and a duty cycle (DC);

[0091] Therefore, the multi-functional radar pulse sequence of the embodiment of the present application can be represented by the following vector S:

[0092]

[0093] Among them, PDW i =[RF i PW i PA i ,PRI i DC i ] T ,i=1,2,…,n represents the pulse description word of the i-th pulse.

[0094] The preprocessing module in this embodiment of the invention is used to detect and process missing pulses and false pulses in a single radar radiation source pulse signal sequence, and then normalize it to obtain a preprocessed single radar pulse signal sequence. Specifically, considering the possibility of missing pulses and false pulses caused by interference from noise, environment, and other factors in the intercepted radar pulse signal, these two situations are highly relevant to the correct detection of subsequent radar pulse sequence switching points. Therefore, before performing switching point detection, the single radar radiation source pulse signal sequence is first processed for missing pulses and false pulses, and then normalized. Figure 2 As shown, the process includes the following:

[0095] (1) First, detect and process the leaked pulse: For the i-th value PRI in the PRI sequence i Calculate the average value M of the two preceding and following PRI values. If PRI i If the current PRI is >2.5M, then the current PRI i The error PRI value can be considered as generated by two leak pulses, and PRI is taken as... i / 3 and M are interpolated twice before the error PRI value, while PRI is... i Correct the value after subtracting the two interpolated values; if 1.5M <PRI i If the current PRI is less than 2.5M, then the current PRI... i The error PRI value can be considered as generated by a leaky pulse. M is taken and interpolated before the error PRI value, while PRI is... i Subtract M to make corrections.

[0096] (2) Detection and processing of false pulses: First, calculate the mean M2 of the PRI sequence. If PRI i <0.8M2 and PRI i+1 If the value is less than 0.8M2, then these two consecutive PRI values ​​are considered as errors caused by spurious pulses. The two PRI values ​​are added together as the correction value for the spurious pulses and used to replace the two consecutive PRI values.

[0097] (3), because the missing pulse, false pulse will not affect the rest of the parameters, for the interpolation of missing pulse, the rest of the parameter selection error pulse before and after two pulses the average value of this parameter as fill in; Duty cycle parameter is calculated according to the false pulse, missing pulse after processing value update; Finally, each dimension parameter is normalized, the preprocessed single radar pulse signal sequence is obtained

[0098] The mutation point matrix construction module of the embodiment of the application is used for detecting state switching points based on the multi-dimensional parameters of the preprocessed single radar pulse signal sequence, generating first layer detail coefficients by respectively performing 4 layer decomposition and reconstruction of discrete wavelet on each independent dimensional parameter, and obtaining the set of possible mutation points of the dimensional parameter by recording the local maximum value of the detail coefficients in each window through a sliding window. The multi-dimensional parameter mutation point matrix can be obtained by repeating the above operation for each dimensional parameter. Specifically, the following processes are included.

[0099] The specific steps are as follows:

[0100] 1. Considering the parameter overlap between different working modes of the multifunctional radar, and the signal parameter change in the same working mode, if only the switching point of a single pulse parameter is detected, it is not reliable, for example, if only the PRI parameter is detected in Figure 3 , many false switching points in the working mode will be detected, and the real switching points will be missed, which will cause great error to the subsequent working mode recognition. In order to avoid the interference of false switching points on the single pulse parameter sequence and as few as possible to miss the real switching points, the multi-dimensional parameters of the preprocessed pulse sequence are comprehensively tested. The multifunctional radar pulse sequence is regarded as a multi-dimensional time sequence, and the method of multi-dimensional time sequence analysis is used for analysis. For multi-dimensional time sequence structure change detection, first, the mutation points are determined on each independent dimension, and the switching points are selected as the final mutation points of the whole multi-dimensional time sequence through significance comparison or voting strategy.

[0101] 2. Generally, for any signal f(t) ∈ L 2 (R), the discrete wavelet transform (DWT) has the following specific form:

[0102]

[0103] wherein j, k are any positive integer, and j is the number of layers of decomposition (scale parameter), and k is the translation parameter, For discrete wavelet function, a0>1, t0>1, from the above formula, the wavelet transform has a unique "zoom" property through scale change and translation change, and can well extract the local band feature of the signal, and is very suitable for analyzing non-stationary signals. Since the wavelet transform has the time-frequency localization property, the position and size of the singular point (abrupt point) of the signal can be detected and analyzed by the local maximum value of the wavelet transform coefficient module.

[0104] 3、For the preprocessed radar pulse PDW sequence, it can be regarded as a 5-dimensional time sequence X={x1, x2, …, x i ,…,x n}∈R 5×n , the sequence has multiple switching points and the positions are unknown, wherein x i represents the multi-dimensional parameter vector of the PDW time sequence at i time (i.e. the i-th pulse), n is the length of the time sequence.

[0105] An arbitrary one-dimensional parameter sequence (such as the carrier frequency {x1 RF ,x2 RF ,…,x i RF ,…,x n RF}) is independently decomposed by 4-layer wavelet and reconstructed to generate the detail coefficient, and the first layer detail coefficient d1 is selected,

[0106] [d, a] = wavedec(data, 4, db3);

[0107] d1 = wrcoef('d', d, a, db3, 1);

[0108] Wherein, wavedec(data, 4, db3) represents that the data is decomposed by 4 layers using db3 wavelet, d represents the detail coefficient after wavelet decomposition, and a represents the approximation coefficient after wavelet decomposition; wrcoef('d', d, a, db3, 1) represents the operation of reconstructing the first layer detail coefficient using db3 wavelet, and d1 represents the first layer detail coefficient after reconstruction.

[0109] 4、Further, a fixed window length is used to slide from the beginning of the first layer detail coefficient sequence generated by reconstruction to the end, and the position coordinates of the maximum detail coefficient value of each sliding window are recorded. From the analysis of S31, it can be seen that the local maximum value of the detail coefficient represents that the signal may have an abrupt change at this time.

[0110] 5、For X={x1, x2, …, x i ,…,x n}∈R 5×n The processing of step S3 in each dimension can obtain a multi-dimensional parameter mutation point matrix T S :

[0111]

[0112] wherein, 1≤q≤l represents the pulse position coordinate corresponding to the maximum value of the detail coefficient of each dimension parameter in the qth sliding window, l is the number of sliding times of the sliding window.

[0113] The state switching point detection module of the embodiment of the application is used for determining the final mutation point of the single radar radiation source pulse signal sequence based on the multi-dimensional parameter mutation point matrix, and realizes the state switching point detection of the single radar radiation source pulse signal sequence. Specifically, the obtained multi-dimensional parameter mutation point matrix is comprehensively analyzed, and the switching point is selected as the final mutation point of the entire multi-dimensional time sequence through the strategies such as significance comparison or voting, so as to realize the state switching point detection of the radar pulse signal sequence and perform sequence segmentation based on the same, and obtain the single radar state pulse sequence.

[0114] wherein, the final mutation point set of the pulse time sequence is determined as:

[0115]

[0116] wherein y s (1≤s≤p) is the s th final mutation point position, and p is the number of switching points.

[0117] The specific algorithm flow is shown in Table 1:

[0118] Table 2 algorithm flow

[0119]

[0120] wherein, the above detection algorithm parameters include the sliding window length, the sliding step, the wavelet function selection, and the decomposition layer number.

[0121] The sequence segmentation module of the embodiment of the application is used for segmenting the single radar radiation source pulse signal sequence according to the state switching point, and obtains the radar pulse sub-sequence S' h , h=1, 2, …, p+1, and p is the number of switching points.

[0122] The above only describes the preferred embodiments and principles of the application in detail, and for ordinary skilled persons in the art, the specific implementation manner can be changed according to the idea provided by the application, and these changes should be regarded as the protection scope of the application.

Claims

1. A method for detecting a multi-function radar state switching point based on wavelet transform, characterized in that, The method comprises the following steps: S1, collecting a multi-functional radar pulse signal sequence with multiple working modes, and obtaining a single radar emitter pulse signal sequence after feature parameter extraction and pulse sorting; S2, detecting and processing missing pulses and false pulses of the single radar emitter pulse signal sequence, and performing normalization to obtain a pre-processed single radar pulse signal sequence; S3, detecting a state switching point based on multi-dimensional parameters of the pre-processed single radar pulse signal sequence, performing four-layer discrete wavelet decomposition on each independent dimensional parameter, reconstructing to generate first layer detail coefficients, recording local maximum values of the detail coefficients in each window through a sliding window, and obtaining a mutation point set that may exist in the dimensional parameter, repeating the above process to obtain a multi-dimensional parameter mutation point matrix; S4, determining a final mutation point of the single radar emitter pulse signal sequence based on the multi-dimensional parameter mutation point matrix, and realizing state switching point detection of the single radar emitter pulse signal sequence and sequence segmentation based on the same.

2. The method of claim 1, wherein, In the step S1, the multi-functional radar pulse signal sequence is represented by a pulse description word (PDW) sequence composed of extracted feature parameters, and the feature parameters include a carrier frequency (RF), a pulse width (PW), an amplitude (PA), a pulse repetition interval (PRI), and a duty cycle (DC); The sequence of multifunction radar pulse signals is represented by the following vector : ； wherein represents the pulse description word of the received first pulse.

3. The method of claim 2, wherein In the step S2, the detection and processing of missing pulses include: For the i-th value in the PRI sequence , the average M of the two preceding and following PRIs is calculated; If , the current error PRI value generated by two missing pulses is taken as and M are interpolated twice in sequence before the error PRI value, and after subtracting the interpolated values twice, the correction is made; If , the current error PRI value is considered to be caused by a missing pulse, and an interpolation is performed on the error PRI value with M, and is modified by subtracting M.

4. The method of claim 3, wherein, In the step S2, the detection and processing of false pulses include: Computing the mean of the PRI sequence If and then the two consecutive PRIs are considered to be false pulse generated errors, the two PRI values are added together as a false pulse correction and replace the two consecutive PRI values.

5. The method of claim 4, wherein, For interpolation of missing pulses, the average value of the parameters of the two pulses before and after the error pulse is selected as the filling value; The duty cycle parameter is calculated and updated according to the values after processing of the false pulses and the missing pulses; Finally, the parameters are normalized.

6. The method of claim 5, wherein, In the step S3, the state switching point detection comprises the following steps: S31. For the pre-processed single-pulse radar signal sequence, consider it as a 5-dimensional PDW time series The sequence has multiple switching points and their locations are unknown, where represents the multi-dimensional parameter vector of the PDW time series at the th pulse, n is the length of the time series. After four-layer haar wavelet decomposition of an arbitrary dimensional parameter sequence is performed independently using the following formula and the detail coefficients are reconstructed, the first layer detail coefficients are selected, [d, a] = wavedec (data, 4, db3); d1 = wrcoef ('d', d, a, db3, 1); wherein, wavedec (data, 4, db3) represents four-layer decomposition of data data using db3 wavelet, d represents the detail coefficients after wavelet decomposition, and a represents the approximation coefficients after wavelet decomposition; wrcoef ('d', d, a, db3, 1) represents the operation of generating the first layer detail coefficients using db3 wavelet reconstruction, and d1 represents the first layer detail coefficients after reconstruction; S32, sliding from the first position of the first layer detail coefficient sequence to the end of the detail coefficient in a fixed window length, and recording the position coordinates of the maximum detail coefficient value in each sliding window; wherein, the local maximum value of the detail coefficient represents a mutation of the signal at this time; S33、to S31, S32 are performed on each dimension parameter of the multi-dimensional parameter mutation point matrix : ； wherein, denotes the pulse position coordinate corresponding to the maximum value of the detail coefficient of the parameter in the qth sliding window, , is the number of sliding times of the sliding window.

7. The method of claim 6, wherein the step of determining the radar state comprises: In the step S4, the final mutation point set of the pulse time sequence is defined as: ； wherein is the s-th final mutation point position, p is the number of switch points.

8. The method of claim 7, wherein, In the step S4, the sequence is divided into radar pulse subsequences in different working modes according to the obtained state switching points , .

9. A system for detecting the switching point of a multi-function radar state based on wavelet transform, applying the detection method according to any one of claims 1-8, characterized in that, The detection system comprises: A collection module configured to collect a multi-functional radar pulse signal sequence with multiple working modes; The feature extraction and pulse sorting module is configured to extract feature parameters and sort pulses of the multi-function radar pulse signal sequence, and obtain a single radar emitter pulse signal sequence. The preprocessing module is configured to detect and process missing pulses and false pulses of the single radar emitter pulse signal sequence, and perform normalization to obtain a preprocessed single radar pulse signal sequence. The mutation point matrix construction module is configured to detect state switching points based on multi-dimensional parameters of the preprocessed single radar pulse signal sequence, perform four-layer discrete wavelet decomposition on each independent dimension parameter, reconstruct a first layer of detail coefficients, record local maximum values of the detail coefficients in each window through a sliding window, and obtain a set of possible mutation points of the dimension parameter. The state switching point detection module is configured to determine final mutation points of the single radar emitter pulse signal sequence based on the multi-dimensional parameter mutation point matrix, and detect state switching points of the single radar emitter pulse signal sequence. The sequence segmentation module is configured to segment the single radar emitter pulse signal sequence according to the state switching points. ​

Images