A modulation period estimation method and system

Abstract

The present invention provides a modulation period estimation method and system, which relates to the technical field of communication and radar signal analysis. The method includes: receiving communication and radar signals to obtain a first characteristic graph matrix; inputting the first characteristic graph matrix into a first estimation and classification network model to obtain a first classification result, and then outputting a first discrimination result; continuously looping to obtain a preliminary estimation period; based on the communication and radar signals, obtaining a modulation period spectrum; within the preliminary estimation period, calculating a corresponding differential signal based on the modulation period spectrum, and then obtaining a second characteristic graph matrix; inputting the second characteristic graph matrix into a second estimation and classification network model to obtain a second classification result, and then outputting a second discrimination result to obtain the modulation period. The present invention solves the problem of how to directly estimate the modulation period for communication and radar signals of multiple different typical signal modulation systems without knowing the signal modulation system and without adding an additional modulation system identification step.

Description

Technical Field

[0001] The present invention relates to the technical field of communication and radar signal analysis, and in particular to a modulation period estimation method and system. Background Art

[0002] To facilitate information transmission, both communication and radar systems modulate signals at regular intervals, known as the symbol period for communication signals or the pulse repetition period for radar signals. This period is collectively referred to as the modulation period. When monitoring received communication and radar signals, the key to extracting information under non-cooperative conditions lies in estimating their modulation parameters. One of the most critical modulation parameters is the modulation period. Understanding the modulation period facilitates the estimation of other parameters and information under non-cooperative conditions.

[0003] Currently, various methods are available for analyzing and estimating the modulation period of communication and radar signals. However, these signals employ a variety of modulation schemes, including multi-level quadrature amplitude modulation (mQAM), multi-level phase shift keying (mPSK), multi-level frequency shift keying (mFSK), pulse Doppler, and pseudo-code phase modulation. Each modulation scheme requires different signal analysis methods. Therefore, modulation identification must be performed prior to signal analysis and processing, complicating the entire signal processing process and increasing the difficulty of hardware implementation and modulation period estimation. Summary of the Invention

[0004] Based on this, it is necessary to provide a modulation period estimation method and system to address the above technical problems, so as to solve the problem of how to directly estimate the modulation period of communication and radar signals of various different typical signal modulation systems without knowing the signal modulation system and without adding additional modulation system identification steps.

[0005] The present invention provides a modulation period estimation method, the method comprising:

[0006] S100: receiving communication and radar signals to obtain a first feature map matrix; inputting the first feature map matrix into a constructed first estimation classification network model to obtain a first classification result; performing discrimination on the first classification result and outputting a first discrimination result;

[0007] S200: Continuously loop S100 until the preset conditions are met Then, the index position of the last jump point in the first discrimination result is searched within the preliminary search interval to obtain a preliminary estimated period;

[0008] S300: obtaining a modulation periodic spectrum based on the communication and radar signals; calculating a corresponding differential signal based on the modulation periodic spectrum within the preliminary estimation period; and constructing a covariance matrix based on the differential signal to obtain a second characteristic map matrix;

[0009] S400: Inputting the second feature map matrix into the constructed second estimation classification network model to obtain a second classification result; performing discrimination on the second classification result and outputting a second discrimination result;

[0010] S500: Obtain a modulation period according to the maximum value of the second discrimination result.

[0011] Furthermore, obtaining the first feature map matrix includes:

[0012] S101: Order ,set up Value and value; Indicates the length of continuous detection The number of segments of the point signal, Indicates the number of maximum singular values ​​selected;

[0013] S102: The front of the communication and radar signal points are discarded to obtain the first signal, based on the previous The point value constructs the dimension The Hankel matrix of

[0014] S103: Perform singular value decomposition on the Hankel matrix to obtain the corresponding left singular matrix, singular value diagonal matrix, and right singular matrix; select the first singular value matrix in the singular value diagonal matrix. maximum singular values, and reconstructing the Hankel matrix using the corresponding left singular matrix and the right singular matrix to obtain a first covariance matrix;

[0015] S104: If , then let , keep looping S102~S103 until , the corresponding The first covariance matrices are superimposed and averaged to obtain Hankel cumulative covariance matrix in dimensions;

[0016] S105: Normalize the Hankel cumulative covariance matrix to obtain the first feature map matrix.

[0017] Furthermore, the first estimation classification network model The build includes:

[0018] S111: Establishing a signal sample library;

[0019] S112: Order ; Select the first signal ;in, The label is , the modulation period is ; Redefine the received signal: , , ;

[0020] S113: Select Value; if , then in the interval Randomly select an integer value as Value; if , , then in the interval Randomly select an integer value as Value; if , then in the interval Randomly select an integer value as value;

[0021] S114: Based on the received signal Get the first feature map matrix , and label it , added to the first feature map sample library, and let ;

[0022] S115: Continuously looping S112 to S114 until all first feature map matrices corresponding to the signals in the signal sample library are obtained, thereby completing the establishment of the first feature map sample library;

[0023] S116: Obtaining the first estimation classification network model based on the first feature map sample library training .

[0024] Furthermore, the first classification result is judged, and outputting the first judgment result includes:

[0025] Determine the first classification result Is it less than the judgment standard? , the first discrimination result is output as: ;in, is a step function; when When the discriminator The first discrimination result outputted is 0, otherwise the first discrimination result outputted is 1.

[0026] Furthermore, the preliminary estimation period The calculation method is:

[0027] ;

[0028] in, , Indicates the number of reference points in the preliminary search interval; Indicates the index position of the last transition point in the first discrimination result; Indicates the judgment criteria.

[0029] Furthermore, obtaining the modulated periodic spectrum includes:

[0030] S301: Set the row value of the dynamic matrix and column values ,in, ;

[0031] S302: Take the front signal points , arrange the point values ​​in column order, and get OK Dynamic matrix of columns;

[0032] S303: Divide the dynamic matrix into two parts evenly to obtain the first part and Part II ; and based on the first part and the second part Correspondingly, two covariance matrices are obtained;

[0033] S304: Calculate the sum of the modulus values ​​of each element in the two covariance matrices respectively, and take the larger value of the two;

[0034] S305: If , then let , keep looping S302~S304 until , and obtain the modulation periodic spectrum.

[0035] Furthermore, obtaining the second feature map matrix includes:

[0036] Based on multiple differential signals, multiple maximum points of the corresponding differential signals are obtained, and the indexes of the multiple maximum points are arranged in the order of the values ​​in the differential signals to obtain an index set. ; Select the index collection The element value in ,use OK Dynamic matrix construction of columns dimensional covariance matrix, and normalize the covariance matrix to obtain the second feature map matrix.

[0037] Furthermore, the second estimation classification network model The build includes:

[0038] S401: Establishing a signal sample library;

[0039] S402: Order ; Select the first signal ;in, The label is , the modulation period is ; Redefine the received signal: , , ;

[0040] S403: Select Value; if , then let ,in, is a rational fraction less than 1; if , then let ,in, is a rational fraction greater than 1 and less than 5; if , then let ; , then let ,in, for A random integer value within ; if , then let ,in, is any random integer value other than the four cases in S403;

[0041] S404: Take the front signal points , arrange the point values ​​in column order, and get OK A dynamic matrix of columns; based on the dynamic matrix, a covariance matrix is ​​obtained, and the covariance matrix is ​​standardized to obtain a second feature map matrix ; and label it , added to the second feature map sample library, and let ;

[0042] S405: Continuously looping S403 to S404 until all second feature map matrices corresponding to the signals in the signal sample library are obtained, thereby completing the establishment of the second feature map sample library;

[0043] S406: Obtaining the second estimation classification network model based on the second feature map sample library training .

[0044] Furthermore, the modulation period The calculation method is:

[0045] ;

[0046] in, Indicates the second classification result; is a step function; Indicates the number of multiple maximum value points of the corresponding differential signal; Indicates the maximum value point of the differential signal; Indicates the maximum value of the second classification result.

[0047] On the other hand, the present invention further provides a modulation period estimation system, comprising a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of any one of the above methods.

[0048] In general, the present invention provides a modulation period estimation method and system, which can achieve the following beneficial effects compared with the prior art:

[0049] (1) The present invention can directly estimate the modulation period of communication and radar signals of various typical signal modulation systems without knowing the signal modulation system and without adding additional modulation system identification steps. In addition, the method proposed by the present invention only requires one training for both the rough estimation first estimation classification network model and the precise estimation second estimation classification network model. Subsequently, no matter how many signals are subjected to modulation period estimation, only the trained estimation classification network model needs to be used, without the need for repeated training. This makes the entire estimation process simpler, more convenient, and faster.

[0050] (2) On the one hand, the present invention utilizes the structural characteristics of the dynamic matrix under the modulation period dimension, and obtains the modulation period spectrum by summing the module values ​​of all elements of the covariance matrix after segmentation to highlight this special structural feature; on the other hand, the Hankel cumulative covariance matrix feature map constructed by the present invention is When taking different values ​​(that is, when being greater than different multiples of the modulation period), it has different structural characteristics; therefore, a rough estimate of the modulation period of the signal can be made, and then these structural characteristics can be used to roughly determine a suitable preliminary estimation period, so as to avoid the search interval of the selected modulation period being too small or too large without any prior information, thereby increasing the estimation efficiency and accuracy.

[0051] (3) Based on the feature map and its special structural characteristics, this paper proposes to use convolutional neural network (ResNet101) for training under large sample conditions, which realizes the The characteristic graph under the value condition is classified and identified, making the estimation of the modulation period faster and more accurate.

[0052] (4) The dynamic covariance matrix feature map constructed by the present invention has the dimension When taking different values ​​(integer multiples or fractional multiples of the modulation period), it also has different structural characteristics, especially When taking different integer multiples of the modulation period, the discrimination is greater, based on which the modulation period can be distinguished under multiple different integer multiples. The values ​​are classified and identified in order to accurately find the real modulation period.

[0053] (5) Since the modulation period spectrum may take local maxima at integer multiples and fractional multiples of the modulation period, it is impossible to directly estimate the modulation period by searching for the maximum point, which will lead to the problem of fuzzy period estimation. Therefore, the present invention combines the modulation period spectrum with the structural characteristics of the dynamic covariance matrix, finds multiple local maximum points as candidate values ​​of the true modulation period through the modulation period spectrum, and then classifies and identifies the dynamic covariance matrix feature map of the dimensions corresponding to the multiple candidate values ​​to find the true modulation period, thereby eliminating the fractional and integer multiples of the modulation period from the multiple candidate values, solving the problem of fuzzy period estimation and achieving better estimation effect.

[0054] (6) Although candidate values ​​of the modulation period can be obtained by searching the modulation period spectrum to find multiple local maximum points, it is difficult to determine the range of the search without any prior information. Therefore, the present invention utilizes the structural characteristics of the Hankel cumulative covariance matrix and classifies and identifies the Hankel cumulative covariance matrix feature map to find the value of M that is greater than a certain integer multiple of the modulation period, thereby determining a search interval that includes at least an integer multiple of the modulation period in [N, 10N], thereby greatly improving the efficiency of modulation period estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0056] Figure 1 1 is a schematic diagram of a modulation period estimation method and system provided by the present invention;

[0057] Figure 2 This is a schematic diagram of a method for obtaining a first characteristic graph matrix of a modulation period estimation method and system provided by the present invention;

[0058] Figure 3 This is a schematic diagram of a method for establishing a first estimation classification network model of a modulation period estimation method and system provided by the present invention;

[0059] Figure 4 This is a schematic diagram of a method for estimating a modulation period and a method for obtaining a modulation period spectrum of the system provided by the present invention;

[0060] Figure 5 It is a schematic diagram of a method for establishing a second estimation classification network model of a modulation period estimation method and system provided by the present invention;

[0061] Figure 6 This is a schematic diagram of dynamic matrix partitioning of a modulation period estimation method and system provided by the present invention;

[0062] Figure 7 This is a schematic diagram of a modulation period spectrum of a modulation period estimation method and system provided by the present invention;

[0063] Figure 8 It is a schematic diagram of a characteristic diagram of the Hankel cumulative covariance matrix generated by different modulation system signals under different M values ​​of a modulation period estimation method and system provided by the present invention;

[0064] Figure 9 It is a schematic diagram of the dynamic covariance matrix characteristics generated by different modulation system signals under different M values ​​of a modulation period estimation method and system provided by the present invention. DETAILED DESCRIPTION

[0065] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings and embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0066] It should be noted that, in the description of the embodiments of the present invention, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a method, step, or system comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such method, step, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the method, step, or system comprising the element.

[0067] like Figure 1 As shown, the present invention provides a modulation period estimation method and system, the method comprising:

[0068] S100: receiving communication and radar signals to obtain a first feature map matrix; inputting the first feature map matrix into a constructed first estimation classification network model to obtain a first classification result; performing discrimination on the first classification result and outputting a first discrimination result.

[0069] Specifically, if Figure 2 As shown, the first feature map matrix The gains can include:

[0070] S101: Order ,set up Value and value; Indicates the length of continuous detection The number of segments of the point signal, Indicates the number of maximum singular values ​​to be selected.

[0071] S102: Transmit communication and radar signals forward points are discarded to obtain the first signal. The point value constructs the dimension The Hankel matrix of ; where , .

[0072] As an example, the dimensions are The Hankel matrix of is:

[0073] .

[0074] S103: Perform singular value decomposition on the Hankel matrix, that is, , and then get the corresponding left singular matrix , singular value diagonal matrix , right singular matrix ; Select the singular value diagonal matrix Center front The largest singular value, and use the corresponding left singular matrix and right singular matrices Reconstruct the Hankel matrix and obtain the first covariance matrix.

[0075] It should be noted that the reconstructed Hankel matrix is:

[0076] ;

[0077] The first covariance matrix is:

[0078] .

[0079] S104: If , then let , keep looping S102~S103 until , the corresponding The first covariance matrices are superimposed and averaged to obtain The Hankel cumulative covariance matrix in dimensions.

[0080] It should be noted that the Hankel cumulative covariance matrix is: .

[0081] S105: Hankel cumulative covariance matrix Perform standardization to obtain the first feature map matrix .

[0082] It should be noted that standardization includes:

[0083] ;

[0084] ;

[0085] .

[0086] make ,Will Input to the trained first estimation classification network model , and obtain the classification results .

[0087] like Figure 3 As shown, the first estimated classification network model The build includes:

[0088] S111: Establish a signal sample library.

[0089] As a specific embodiment, the method for establishing a signal sample library includes:

[0090] S1111: Order ;

[0091] S1112: Set the category to ,make ;

[0092] S1113: Set the signal-to-noise ratio to ,make .

[0093] S1114: respectively from One type of signal system is randomly selected from each of the three types of signal systems, and the corresponding sample signal processing is performed respectively.

[0094] Since the characteristic signals extracted from different signal systems are different, in order to increase the diversity and comprehensiveness of the signal sample library, the present invention randomly selects one type of signal system from each type of signal system, performs corresponding sample signal processing on each type, and then establishes a signal sample library.

[0095] More specifically, one of the first signal systems is randomly selected and sample signal processing is performed; one of the second signal systems is randomly selected and sample signal processing is performed; and one of the third signal systems is randomly selected and sample signal processing is performed.

[0096] Among them, the first type of signal system includes: linear frequency modulation, linear frequency pseudo-code phase modulation composite, rectangular pulse shaping 2~8PSK, 2~8FSK, rectangular pulse shaping 8~32QAM, etc.; the second type of signal system includes: pulse Doppler, linear frequency modulation pulse train, sinusoidal frequency modulation pulse train, etc.; the third type of signal system includes: raised cosine pulse shaping 2~8PSK, raised cosine pulse shaping 8~32QAM, etc.

[0097] Sample signal processing includes: after selecting the signal system, randomly generating the modulation parameter value corresponding to the modulation period (including the modulation period N), and then generating the corresponding discrete time signal according to the modulation system and modulation parameters , randomly insert multiple points of 0 value in the front end of the signal (the number of 0 value points is a random integer between 1 and N), and according to the signal-to-noise ratio Add noise to the sample signal collected; use the signal after adding noise to Said that Label it " ” and then added to the signal sample library.

[0098] S1115: If ,make , go to S1116, otherwise go to S1114; for example, .

[0099] S1116: If ,make , go to S1117, otherwise go to S1113; for example, .

[0100] S1117: If , go to S1118, otherwise go to S1112; for example, .

[0101] S1118: This completes the establishment of the signal sample library. signal samples.

[0102] S112: Order ; Select the first signal ;in, The label is , the modulation period is ; Redefine the receiving signal: , , .

[0103] For example, the signal sample library has signal samples, and take the first 51840 samples from the signal sample library. signal .

[0104] S113: Select Value; if , then in the interval Randomly select an integer value as Value; if , , then in the interval Randomly select an integer value as Value; if , then in the interval Randomly select an integer value as value;

[0105] S114: Based on the received signal Get the first feature map matrix , and label it , added to the first feature map sample library, and let .

[0106] It should be noted that the first feature map matrix The method of obtaining is the same as above and will not be repeated here.

[0107] S115: Continuously loop S112 to S114 until all first feature map matrices corresponding to the signals in the signal sample library are obtained, thereby completing the establishment of the first feature map sample library.

[0108] That is to say, judgment Is it greater than the number of signals in the signal sample library? More specifically, if , to S112, and continue to cycle S112~S114; until , complete the construction of the first feature map sample library, the library contains samples.

[0109] S116: Obtaining a first estimated classification network model based on the first feature map sample library training .

[0110] For example, in the first feature map sample library generated in the above steps, each label "0" to "11" contains 4320 samples. From the samples contained in each label, 2440 samples are randomly selected for training and 600 samples are used for verification. Then, a training sample library with a sample number of 2440*12=28800 and a test sample library with a sample number of 600*12=7200 are constructed. Then, the 28800 samples in the training sample library are fed into the ResNet-101 convolutional neural network for training. The trained network is trained with It means that the network is able to classify the feature map samples fed into it, that is, output a value between "0" and "11".

[0111] As an embodiment, judging the first classification result and outputting the first judgment result includes: judging the first classification result Is it less than the judgment standard? , the output of the first discrimination result is:

[0112] ;

[0113] in, is a step function; when When the discriminator The first discrimination result outputted is 0, otherwise the first discrimination result outputted is 1.

[0114] S200: Continuously looping S100 until a preset condition is met, then searching for the index position of the last jump point in the first discrimination result within the preliminary search interval to obtain a preliminary estimated period.

[0115] It should be noted that, if the preset conditions are met, it can be , that is, in step S100, let , set the initial search range value. That is, exist The value is taken between , and the number of points taken each time is , Indicates the number of reference points in the initial search interval. For example, when Take 10, When 11 is selected, the entire coverage range is 10 to 10240.

[0116] if , then let , keep looping S100 until Only after the preset conditions are met will the index position of the last jump point in the first judgment result be searched within the preliminary search interval.

[0117] As an embodiment, obtaining the index position of the last transition point includes:

[0118] Find the first judgment result The index positions of all the transition points from 0 to 1 in the , and put them all into a set in order middle:

[0119] ;

[0120] and find out The position of the last point in , let it be :

[0121] ;

[0122] Thus, the first judgment result is obtained The index position of the last 0 to 1 transition point in .

[0123] Calculate the corresponding value , get a preliminary estimate of the cycle :

[0124] ;

[0125] in, , Indicates the number of reference points in the initial search interval; Indicates the index position of the last jump point in the first judgment result; Indicates the judgment criteria.

[0126] S300: Based on the communication and radar signals, a modulation periodic spectrum is obtained; within the preliminary estimation period, a corresponding differential signal is calculated based on the modulation periodic spectrum; and a covariance matrix is ​​constructed according to the differential signal to obtain a second characteristic map matrix.

[0127] It should be noted that if Figure 4 As shown, the modulation period spectrum is obtained by:

[0128] S301: Set the row value of the dynamic matrix and column values ,in, .

[0129] S302: Take the front signal points , arrange the point values ​​in column order, and get OK A dynamic matrix of columns.

[0130] in, OK The dynamic matrix of columns is:

[0131] .

[0132] At the same time, its covariance matrix can also be calculated: .

[0133] S303: Divide the dynamic matrix into two parts evenly to obtain the first part and Part II ; and based on the first part and Part II Correspondingly, two covariance matrices are obtained.

[0134] That is, a first covariance matrix is ​​obtained based on the first part; and a second covariance matrix is ​​obtained based on the second part.

[0135] Specifically, according to Value Calculation and , and the matrix It is divided into two parts: and ,in, Depend on Before Row composition, Depend on After Then and Multiplying them with their respective conjugate transposes results in two corresponding covariance matrices, which are:

[0136] ;

[0137] .

[0138] S304: Calculate the sum of the modulus values ​​of each element in the two covariance matrices respectively, and take the larger value of the two.

[0139] That is, take the modulus of each element of the two covariance matrices to obtain a matrix composed of the modulus values ​​of each element and , find the sum of all elements of these two matrices and take the larger value of the two :

[0140] .

[0141] S305: If , then let , keep looping S302~S304 until , and obtain the modulation periodic spectrum .

[0142] By cycling through the above steps, the modulated periodic spectrum is obtained. ,exist Within the range, calculate The differential signal , and find its maximum point :

[0143] ;

[0144] in, .

[0145] It should be noted that the second feature map matrix is ​​obtained by:

[0146] Based on multiple differential signals, multiple maximum points of the corresponding differential signals are obtained, and the indexes of the multiple maximum points are arranged in the order of the values ​​in the differential signals to obtain an index set. ;

[0147] Among them, the index collection The results include: finding There are multiple large value points in the The values ​​in the are sorted into the collection in order of size middle:

[0148] ;

[0149] .

[0150] make ; Select the index collection The element value in ,use OK Dynamic matrix construction of columns The covariance matrix of dimension , and normalize the covariance matrix to obtain the second feature map matrix.

[0151] Standardization processing includes:

[0152] ;

[0153] ;

[0154] ;

[0155] .

[0156] S400: The second feature map matrix Input to the constructed second estimation classification network model The second classification result is obtained ; Discriminate the second classification result and output the second discrimination result.

[0157] As an example, Figure 5 As shown, the second estimated classification network model The build includes:

[0158] S401: Establishing a signal sample library;

[0159] The signal sample library can be established in the same manner as the signal sample library in establishing the first estimation and classification network model.

[0160] As a specific embodiment, the method for establishing a signal sample library includes:

[0161] S4111: Order ;

[0162] S4112: Set the category to ,make ;

[0163] S4113: Set the signal-to-noise ratio to ,make .

[0164] S4114: respectively from One type of signal system is randomly selected from each of the three types of signal systems, and the corresponding sample signal processing is performed respectively.

[0165] It should be noted that the signal system category and the corresponding sample signal processing are the same as those in the signal sample library in establishing the first estimation and classification network model, and will not be described in detail here.

[0166] S4115: If ,make , go to S2116, otherwise go to S2114; for example, .

[0167] S4116: If ,make , go to S2117, otherwise go to S2113; for example, .

[0168] S4117: If , go to S2118, otherwise go to S2112; for example, .

[0169] S4118: This completes the establishment of the signal sample library. signal samples.

[0170] S402: Order ; Select the first signal ;in, The label is , the modulation period is ; Redefine the receiving signal: , , .

[0171] For example, the signal sample library has signal samples, and take the first one from the 127008 samples in the signal sample library. signal

[0172] S403: Select value;

[0173] like , then let ,in, is a rational fraction less than 1;

[0174] like , then let ,in, is a rational fraction greater than 1 and less than 5;

[0175] like , then let ;

[0176] like , then let ,in, for A random integer value within ;

[0177] like , then let ,in, is any random integer value other than the four cases in S403.

[0178] S404: Take the front signal points , arrange the point values ​​in column order, and get OK The dynamic matrix of the column; based on the dynamic matrix, the covariance matrix is ​​obtained, and the covariance matrix is ​​standardized to obtain the second feature map matrix ; and label it , added to the second feature map sample library, and let .

[0179] in, OK Dynamic matrix of columns It can be:

[0180] ;

[0181] Covariance matrix It can be: ;

[0182] Standardization can be:

[0183] ;

[0184] ;

[0185] ;

[0186] .

[0187] S405: Continuously looping S403 to S404 until all second feature map matrices corresponding to the signals in the signal sample library are obtained, thereby completing the establishment of the second feature map sample library.

[0188] More specifically, if , to S402, and continue to cycle S402~S404; until , complete the construction of the second feature map sample library, the library contains samples.

[0189] S406: Obtaining a second estimated classification network model based on the second feature map sample library training .

[0190] For example, in the second feature map sample library generated by the above steps, each label "-1" to "12" contains 9072 samples. From the samples contained in each label, 4800 samples are randomly selected for training and 1200 samples are used for verification. Then, a training sample library with a sample number of 4800*14=67200 and a test sample library with a sample number of 1200*14=16800 are constructed. Then, the 67200 samples in the training sample library are fed into the ResNet-101 convolutional neural network for training. The trained network is used It means that the network is able to classify the feature map samples fed into it, that is, output a value between "-1" and "12".

[0191] S500: Obtaining a modulation period according to the maximum value in the second discrimination result.

[0192] It should be noted that the second discrimination result may be:

[0193] ;

[0194] if , then let , reselect the index set The element value in , and keep looping to get the second judgment result; if , then directly calculate the second discrimination result The maximum value of: ; Then calculate the modulation period .

[0195] Furthermore, the modulation period The calculation method is:

[0196] ;

[0197] in, Indicates the second classification result; is a step function; Indicates the number of multiple maximum value points of the corresponding differential signal; Indicates the maximum value point of the differential signal; Indicates the maximum value of the second classification result.

[0198] In the process of obtaining the modulation period spectrum, when the modulation period value is obtained, the dynamic matrix obtained is After being divided into two parts by row, at least one of the two divided matrices has column data that comes from the same modulation period.

[0199] like Figure 6As shown, at this time, the sum of the module values ​​of all elements of the covariance matrix of the segmentation matrix will be a large value, that is, the obtained modulation period spectrum R[M] is A local maximum value will be obtained at this point, which can be used to preliminarily distinguish and Therefore, the matrix can play a role in preliminary screening of modulation cycles. It should be noted that, if Figure 7 As shown in FIG, in addition to the local maximum at the modulation period, R[M] may also take local maximum at integer multiples and fractional multiples of the modulation period.

[0200] like Figure 8 As shown, for a periodically modulated signal, the constructed Hankel cumulative covariance matrix characteristic diagram It has two special structures. Figure 8 (a) shows the Hankel cumulative covariance matrix feature map generated by 8FSK. Each sub-graph corresponds to a different range of M values. Within this range, the shape of the feature map is roughly the same. It can be found that the larger M is, the thinner the bars on the main diagonal of the feature map are. In theory, the thickness of the bars can be used to make a rough estimate and find the appropriate M value, for example, find an M value that is approximately 4 times the modulation period. Figure 8 (a) The signals with characteristic pattern shapes also include: mfsk signal, rectangular pulse shaped mpsk signal, rectangular pulse shaped mqam signal, raised cosine pulse shaped mpsk signal, raised cosine pulse shaped mqam signal, etc. Figure 8 (b) shows the Hankel cumulative covariance matrix feature map generated by the linear frequency modulation continuous wave signal. Each sub-graph corresponds to a different range of M values. It can be found that the larger M is, the more straight lines there are in the diagonal direction of the feature map. In theory, a rough estimate can be made by the number of straight lines to find the appropriate M value. Figure 8 (b) Signals with characteristic pattern shapes also include: linear frequency modulation pulse train signals, sinusoidal frequency modulation pulse train signals, pulse Doppler signals, etc.

[0201] Depend on Figure 8 It can be seen that based on the special structural characteristics of the Hankel cumulative covariance matrix feature map, a deep neural network can be introduced to learn the Hankel cumulative covariance matrix feature maps under a large number of different M values, and adaptively generate the optimal classifier to judge whether the M value corresponding to the current analysis dimension is greater than a certain integer multiple of the modulation period, so as to determine a class containing The search interval is an integer multiple of at least one modulation period, which improves the efficiency of accurate estimation of subsequent modulation periods.

[0202] like Figure 9As shown, the dynamic covariance matrix characteristic diagram of different modulation system signals with different values ​​of M is shown as follows: When M = pN (p is an integer greater than or equal to 1), the characteristic graph will inevitably have multiple independent blocks (square or conical) on the diagonal, and the number of blocks will increase as the value of p increases. In addition, if M = pN (where p is a rational fraction), when p>1, as p increases, the shape of the signal characteristic graph of various systems will gradually become a stick shape on the main diagonal, and the larger p is, the thinner the stick is. In each system, the characteristic graph when p<1 or p>1 also has a certain degree of discrimination.

[0203] In summary, when M takes different integer multiples of the actual modulation period, the feature maps exhibit significant discriminability, most notably in the inconsistent number of tiles in the resulting image. The previously proposed R[M] not only exhibits local maxima at the modulation period, but can also reach local maxima at integer and fractional multiples of the modulation period. Therefore, this property can be combined with R[M] to identify multiple local extreme points using the R[M] differential waveform. The feature maps corresponding to these local extreme points in the M-dimensional dimension are then classified to identify the one corresponding to the actual modulation period (i.e., M=N). This allows for distinguishing the modulation period from its integer (or fractional) multiples, and ultimately determining the modulation period N. To achieve effective classification, a deep neural network can be introduced to adaptively generate an optimal classifier by learning from a large number of feature map samples under varying signal-to-noise ratios and with varying M-dimensional dimensions.

[0204] On the other hand, the present invention also provides a modulation period estimation system, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of any of the above methods. The technical solution of the system is consistent with the technical solution of the above method and will not be repeated here.

[0205] It should be noted that for the aforementioned embodiments, for simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that this application is not limited by the order of the actions described, because according to this application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily required by this application.

[0206] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0207] In the several embodiments provided in this application, it should be understood that the disclosed methods or systems can be implemented in other ways. For example, the embodiments described above are merely illustrative, and the division of the units described is merely a logical functional division. In actual implementation, other division methods may be used, such as combining or integrating multiple units or components into another system, or ignoring or not implementing certain features.

[0208] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0209] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0210] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of this application, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. This computer software product is stored in a memory and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of this application.

[0211] Those skilled in the art will appreciate that all or part of the various circuits in the above embodiments may be implemented by instructing related hardware through a program. The program may be stored in a computer-readable memory, which may include a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0212] The above is only an exemplary embodiment of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure herein, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art that are not recorded in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

[0213] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, 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, they should be considered to be within the scope of this specification.

[0214] It will be easily understood by those skilled in the art that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A modulation period estimation method, characterized in that: The method comprises: S100: receiving communication and radar signals to obtain a first feature map matrix; inputting the first feature map matrix into a constructed first estimation classification network model to obtain a first classification result; performing discrimination on the first classification result and outputting a first discrimination result; S200: Continuously looping S100 until a preset condition is met, then searching for the index position of the last jump point in the first discrimination result within the preliminary search interval to obtain a preliminary estimated period; S300: obtaining a modulation periodic spectrum based on the communication and radar signals; calculating a corresponding differential signal based on the modulation periodic spectrum within the preliminary estimation period; and constructing a covariance matrix based on the differential signal to obtain a second characteristic map matrix; S400: Inputting the second feature map matrix into the constructed second estimation classification network model to obtain a second classification result; performing discrimination on the second classification result and outputting a second discrimination result; S500: Obtain a modulation period according to the maximum value of the second discrimination result.

2. A modulation period estimation method according to claim 1, characterized in that: Obtaining the first feature map matrix includes: S101: Order ,set up Value and value; Indicates the length of continuous detection The number of segments of the point signal, Indicates the number of maximum singular values ​​selected; S102: The front of the communication and radar signal points are discarded to obtain the first signal, based on the previous The point value constructs the dimension The Hankel matrix of S103: Perform singular value decomposition on the Hankel matrix to obtain the corresponding left singular matrix, singular value diagonal matrix, and right singular matrix; select the first singular value matrix in the singular value diagonal matrix. maximum singular values, and reconstructing the Hankel matrix using the corresponding left singular matrix and the right singular matrix to obtain a first covariance matrix; S104: If , then let , keep looping S102~S103 until , the corresponding The first covariance matrices are superimposed and averaged to obtain Hankel cumulative covariance matrix in dimensions; S105: Normalize the Hankel cumulative covariance matrix to obtain the first feature map matrix.

3. A modulation period estimation method according to claim 1, characterized in that: The first estimated classification network model The build includes: S111: Establishing a signal sample library; S112: Order ; Select the first signal ;in, The label is , the modulation period is ; Redefine the received signal: , , ; S113: Select Value; if , then in the interval Randomly select an integer value as Value; if , , then in the interval Randomly select an integer value as Value; if , then in the interval Randomly select an integer value as value; S114: Based on the received signal Get the first feature map matrix , and label it , added to the first feature map sample library, and let ; S115: Continuously looping S112 to S114 until all first feature map matrices corresponding to the signals in the signal sample library are obtained, thereby completing the establishment of the first feature map sample library; S116: Obtaining the first estimation classification network model based on the first feature map sample library training .

4. A modulation period estimation method according to claim 1, characterized in that: Discriminant the first classification result and output the first discrimination result includes: Determine the first classification result Is it less than the judgment standard? , the first discrimination result is output as: ;in, is a step function; when When the discriminator The first discrimination result outputted is 0, otherwise the first discrimination result outputted is 1.

5. A modulation period estimation method according to claim 1, characterized in that: The initial estimation period The calculation method is: ； in, , Indicates the number of reference points in the preliminary search interval; Indicates the index position of the last transition point in the first discrimination result; Indicates the judgment criteria.

6. A modulation period estimation method according to claim 5, characterized in that: The obtaining of the modulated periodic spectrum includes: S301: Set the row value of the dynamic matrix and column values ,in, ; S302: Take the front signal points , arrange the point values ​​in column order, and get OK Dynamic matrix of columns; S303: Divide the dynamic matrix into two parts evenly to obtain the first part and Part II ; and based on the first part and the second part Correspondingly, two covariance matrices are obtained; S304: Calculate the sum of the modulus values ​​of each element in the two covariance matrices respectively, and take the larger value of the two; S305: If , then let , keep looping S302~S304 until , and obtain the modulation periodic spectrum.

7. A modulation period estimation method according to claim 1, characterized in that: Obtaining the second feature map matrix includes: Based on multiple differential signals, multiple maximum points of the corresponding differential signals are obtained, and the indexes of the multiple maximum points are arranged in the order of the values ​​in the differential signals to obtain an index set. ; Select the index collection The element value in ,use OK Dynamic matrix construction of columns dimensional covariance matrix, and normalize the covariance matrix to obtain the second feature map matrix.

8. A modulation period estimation method according to claim 1, characterized in that: The second estimated classification network model The build includes: S401: Establishing a signal sample library; S402: Order ; Select the first signal ;in, The label is , the modulation period is ; Redefine the received signal: , , ; S403: Select Value; if , then let ,in, is a rational fraction less than 1; if , then let ,in, is a rational fraction greater than 1 and less than 5; if , then let ;like , then let ,in, for A random integer value within ; if , then let ,in, is any random integer value other than the four cases in S403; S404: Take the front signal points , arrange the point values ​​in column order, and get OK A dynamic matrix of columns; based on the dynamic matrix, a covariance matrix is ​​obtained, and the covariance matrix is ​​standardized to obtain a second feature map matrix ; and label it , added to the second feature map sample library, and let ; S405: Continuously looping S403 to S404 until all second feature map matrices corresponding to the signals in the signal sample library are obtained, thereby completing the establishment of the second feature map sample library; S406: Obtaining the second estimation classification network model based on the second feature map sample library training .

9. A modulation period estimation method according to claim 1, characterized in that: The modulation period The calculation method is: ； in, Indicates the second classification result; is a step function; Indicates the number of multiple maximum value points of the corresponding differential signal; Indicates the maximum value point of the differential signal; Indicates the maximum value of the second classification result.

10. A modulation period estimation system, comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 9.

Images