Satellite same-frequency mixed signal modulation identification method

Abstract

This invention provides a method for identifying the modulation scheme of mixed satellite signals at the same frequency. It combines signal cyclic spectrum analysis with a neural network to jointly identify the modulation schemes of any two mixed satellite signals. The signal modulation types include bpsk, qpsk, π / 4dqpsk, uqpsk, and 8psk. Maximum probability template matching is then performed to effectively identify any two mixed modulation schemes of commonly used satellite communication signals, including bpsk, qpsk, π / 4dqpsk, uqpsk, and 8psk. The method consists of two phases: a preparation phase and an application phase. The preparation phase generates the neural network dataset and trains the neural network, while also generating the target matrix set. The application phase is used for the modulation identification of the mixed signals.

Description

Technical Field

[0001] This invention relates to the field of signal modulation technology, and more particularly to a method for identifying satellite co-frequency mixed signal modulation. Background Technology

[0002] With the rapid development of satellite communication technology, new services and systems are constantly emerging, and spectrum resources are becoming increasingly scarce. In non-cooperative satellite communication, the phenomenon of multiple narrowband signals aliasing in the time and frequency domains is inevitable. Conventional time-domain, frequency-domain, and spatial-domain methods are basically ineffective in identifying signal modulation methods when receiving signals with a single antenna. At this time, a new method for identifying mixed signal modulation to deal with interference from co-channel signals is needed.

[0003] Today, satellite communication technology is not only widely used in information transmission, but also plays a crucial role in public safety and national defense. In the civilian sector, the basis for parameter estimation and location of stolen signals is the identification of the signal modulation mode. In the military field, satellite communication also plays a vital role. In electronic warfare, electromagnetic interference methods such as signal suppression are commonly used to disrupt enemy communication equipment, and one very effective method is co-channel interference. When such interference occurs, accurately identifying the signal modulation mode in a complex electromagnetic environment to suppress the interference is a crucial foundation for ensuring communication.

[0004] Current methods for identifying modulation of mixed signals at the same frequency include: first, separating the signals and then identifying each individual signal; however, at low signal-to-noise ratios, the signal separation error is large, leading to a significant decrease in identification accuracy and making effective modulation identification impossible. Another method involves establishing a virtual multidimensional matrix and using underdetermined blind source separation to separate the signals, recover the transmitted signals, and identify the modulation scheme using characteristic parameters such as the fourth-order cumulant; however, this method is significantly affected by frequency domain aliasing. Yet another method extracts classification features directly from the cyclic domain of the mixed waveform using the cyclic domain characteristics of the signal, directly identifying the clearly separable feature domains. One approach is to effectively identify cyclic spectrum peaks by reducing background noise in the cyclic spectrum, and then use a threshold to identify the peak position of the cyclic frequency; however, the amplitude of the spectral peaks changes with the signal power, making the method of setting a threshold to identify the presence of spectral peaks a significant drawback. Another approach is to simplify the cyclic spectrum of the signal and combine it with deep learning to achieve mixed signal modulation identification; however, this method is computationally complex and computationally intensive, and because the characteristics of the signal cyclic frequency are similar for different modulation schemes, the total number of signal classes that can be identified by this method is limited. Summary of the Invention

[0005] In response to the technical problems mentioned in the background section, this invention provides a method for identifying the modulation schemes of co-channel satellite signals. By combining signal cyclic spectrum analysis and neural networks, it can jointly identify the modulation schemes of any two co-channel satellite signals. The modulation types include bpsk, qpsk, π / 4dqpsk, uqpsk, and 8psk. This invention provides an effective theoretical basis for suppressing co-channel satellite interference in both civilian and military fields, and has significant practical implications.

[0006] The technical means employed in this invention are as follows:

[0007] A method for identifying satellite co-frequency mixed signal modulation includes the following steps:

[0008] For any two combinations of input signals, namely binary phase shift keying (bpsk), quadrature phase shift keying (qpsk), π / 4 quadrature phase shift keying (π / 4dqpsk), unbalanced quadrature phase shift keying (uqpsk), and 8-phase shift keying (8psk), perform second-order cyclic moments m. 20 and fourth-order cyclic moment m 40 Calculation;

[0009] Based on the second-order cyclic moment m under each different code rate modulation scheme combination 20 The 2f of the calculation result c , 2f c +f m2 The fourth cyclic moment m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 At this point, determine whether a peak generation target matrix exists;

[0010] Second-order cyclic moment m under each different code rate modulation scheme combination 20 The 2f of the calculation result c , 2f c +f m2 and fourth-order cyclic moment m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 The training dataset is generated centered around the peak data.

[0011] Receive the input signal and represent the input signal as:

[0012] r(t) = r1(t) + r2(t) + n(t)

[0013] Where r1(t) and r2(t) represent the respective modulation signals of the two mixed signals, and n(t) represents the noise;

[0014] Calculate the m of the received signal 21 Cyclic moments are used to determine the code rate of each of the two signals.

[0015] Find m 21 The maximum value and its index [m1, i1], and the second largest value and its index [m2, i2] in the cyclic moments, where i1 and i2 are the cyclic frequencies of the two mixed signals, denoted as... f m2 ;

[0016] Calculate the m of the signal 20 and m 40 The cyclic moment, so as to be in m 20 The 2f of the calculation result c , 2f c +f m2 place and m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 M data points are selected to the left and right of the center to generate the input dataset. A neural network is used to identify whether there is a peak in the training dataset. A 1×8 matrix is ​​generated based on the output of the neural network.

[0017] Based on the mixed signal target matrix, the maximum likelihood probability density of the matrix is ​​calculated based on the generated 1×8 matrix.

[0018] If the target matrix with the highest probability is [1 1 1 1 1 0 1 0], then calculate the signal C. 20 C 21 Higher-order cumulants, according to C 20 / C 21 Determine the modulation method.

[0019] Furthermore, the second-order cyclic moment m is applied to any two combinations of the input signal binary phase shift keying (bpsk), quadrature phase shift keying (qpsk), π / 4 quadrature phase shift keying (π / 4dqpsk), unbalanced four-phase phase shift keying (uqpsk), and 8-phase shift keying (8psk). 20 and fourth-order cyclic moment m 40 The calculations include:

[0020] m 20 (f,τ)= <r(t)r(t+τ)e -j2πft > t

[0021] m 40 (f,τ1,τ2,τ3,)= <r(t)r(t+τ1)r(t+τ2)r(t+τ3)e -j2πft > t

[0022] Where r(t) represents the input signal, τ, τ1, τ2, τ3 represent the time delay, and f represents the cycle frequency. t Indicates time average.

[0023] Furthermore, the second-order cyclic moment m under each different code rate modulation scheme combination... 20 The 2f of the calculation result c , 2f c +f m2 The fourth cyclic moment m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 At this point, determine whether a peak generation target matrix exists, including:

[0024] In the cyclic moment m 20 A signal with characteristics will be at m 20 2f in the cyclic spectrum c , 2f c +f m2 The peak value appears at [location].

[0025] In the cyclic moment m 40 A signal with characteristics will be at m 40 4f in the cyclic spectrum c , 4f c +f m2 , 4f c +0.5f m2 The peak value appears at [location].

[0026] According to [2f] c , 2f c +f m2 4f c , 4f c +f m2 , 4f c +0.5f m2 [8] Does the target matrix exist for generating peaks? If there are peaks, the position within [] is 1; if there are no peaks, it is 0.

[0027] Furthermore, the second-order cyclic moment m under each different code rate modulation scheme combination... 20 The 2f of the calculation result c , 2f c +f m2 and fourth-order cyclic moment m 40 The 4f of the calculation result c 4f c +f m1 4f c +f m2 , 4f c +0.5f m2 The training dataset is generated centered around the peak data and includes:

[0028] With 2f c , 2f c +f m2 4f c , 4f c +f m2 , 4f c +0.5f m2 For each of the M points to the left and right of the center, a training dataset is generated, and a training dataset is generated at each peak.

[0029] Each generated training dataset is represented as: [C(kM)...C(k)...C(k+M)], where k represents the index of the peak cyclic frequency of the cyclic moment, C(k) represents the amplitude at k, and M represents half of the set sampling data point window, with a total number of points of 2M+1.

[0030] Further, the calculation of the received signal m 21 Cyclic moments are used to determine the code rate of each of the two signals. The calculation expression is as follows:

[0031] m 21 (f,τ)= <r(t)r * (t+τ)e -j2πft > t

[0032] Where, r *(t) represents the conjugate of the input signal.

[0033] Furthermore, the step of finding m 21 The maximum value and its index [m1, i1], and the second largest value and its index [m2, i2] in the cyclic moments, where i1 and i2 are the cyclic frequencies of the two mixed signals, denoted as... f m2 ,include:

[0034] m 21 The cyclic spectrum of the cyclic moment contains two peaks. The horizontal axis corresponding to each peak represents the code rate of the two signals. Find the two peaks and find the code rate of the two signals according to the horizontal axis index of the two peaks.

[0035] Furthermore, based on the mixed signal target matrix and the generated 1×8 matrix, the maximum likelihood probability density of the matrix is ​​calculated, and the calculation expression is as follows:

[0036]

[0037] Where A represents the target matrix, B represents the matrix output by the neural network, and n is the number of elements in the matrix.

[0038] Furthermore, if the maximum probability target matrix is ​​[1 1 1 1 1 0 1 0], then the signal C is calculated. 20 C 21 Higher-order cumulants, according to C 20 / C 21 Determining the modulation method includes:

[0039] If the target matrix with the highest probability is [1 1 1 1 1 0 1 0], then calculate the signal C. 20 C 21 Higher-order cumulants are as follows:

[0040] C 20 =E[r(t)r(t)]

[0041] C 21 =E[r(t)r * (t)]

[0042] Where E[·] represents the mean.

[0043] According to C 20 / C 21 Determine the modulation method: if it is greater than or equal to 0.9, it is a mixed signal of bpsk and bpsk; if it is greater than or equal to 0.5 and less than 0.9, it is a mixed signal of bpsk and uqpsk; if it is less than 0.5, it is a mixed signal of uqpsk and uqpsk.

[0044] Compared with the prior art, the present invention has the following advantages:

[0045] 1. This invention proposes for the first time a method combining neural networks with the cyclic frequency of the modulation signal's cyclic moment. By using neural networks to identify single spectral peaks at the cyclic frequency where the modulation signal may have peaks, a discrimination matrix is ​​generated, which greatly increases the accuracy of identification while reducing the amount of computation.

[0046] 2. In the process of modulation identification of mixed signals at the same frequency using signal cyclic moments, the cyclic spectral features of bpsk and uqpsk are similar and cannot be effectively identified. Furthermore, signals with peaks at the carrier frequency's cyclic frequency can cause other signals without peaks at the carrier frequency to also have peaks, leading to modulation identification errors. To address this problem, this invention proposes for the first time a mixed signal modulation identification method that uses the maximum likelihood of the probability density of the decision matrix for template comparison, combined with the decision matrix generated by spectral peak identification.

[0047] 3. This invention enables accurate identification of the modulation type of mixed signals in any two mixed forms of commonly used PSK type signals for satellites. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram illustrating the process of generating the neural network input training dataset and the final determination of the target matrix for the mixed signal bar modulation method in this invention.

[0050] Figure 2 This is a schematic diagram illustrating the application process of the satellite co-frequency mixed signal modulation identification method provided by the present invention.

[0051] Figure 3 A schematic diagram showing the setup of the neural network input training dataset for this invention.

[0052] Figure 4 A schematic diagram showing the setup of the neural network input training dataset for this invention. Detailed Implementation

[0053] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0054] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0055] like Figure 1-4 As shown, this invention provides a satellite co-frequency mixed signal modulation identification method, which can be divided into two main stages, the first being the preparation stage: Figure 1 The flowchart for the preparation phase mainly involves generating the neural network input training dataset and generating the final target matrix for determining the mixed signal bar modulation method. This includes the following steps:

[0056] S1. Perform second-order cyclic moments m on any two combinations of the input signal: binary phase shift keying (bpsk), quadrature phase shift keying (qpsk), π / 4 quadrature phase shift keying (π / 4dqpsk), unbalanced four-phase phase shift keying (uqpsk), and eight-phase phase shift keying (8psk). 20 and fourth-order cyclic moment m 40 Calculation;

[0057] S2, the second-order cyclic moment m under each different code rate modulation scheme combination. 20 The 2f of the calculation result c 2f c +f m1 2f c +f m2 The fourth cyclic moment m 40 The 4f of the calculation result c , 4f c +f m2 , 4fc +0.5f m2 At this point, determine whether a peak generation target matrix exists;

[0058] S3, the second-order cyclic moment m under each different code rate modulation scheme combination. 20 The 2f of the calculation result c , 2f c +f m2 and fourth-order cyclic moment m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 The training dataset is generated centered around the peak data.

[0059] Application phase: Figure 2 This is a flowchart of the application phase, specifically a flowchart of the application process of the satellite co-frequency mixed signal modulation identification method provided by this invention.

[0060] S4. Receive the input signal and represent the input signal as follows:

[0061] r(t) = r1(t) + r2(t) + n(t)

[0062] Where r1(t) and r2(t) represent the respective modulation signals of the two mixed signals, and n(t) represents the noise;

[0063] S5. Calculate the m of the received signal. 21 Cyclic moments are used to determine the code rate of each of the two signals.

[0064] S6, Find m 21 The maximum value and its index [m1, i1], and the second largest value and its index [m2, i2] in the cyclic moments, where i1 and i2 are the cyclic frequencies of the two mixed signals, denoted as... f m2 ;

[0065] S7. Calculate the m of the signal. 20 and m 40 The cyclic moment, so as to be in m 20 The 2f of the calculation result c , 2f c +f m2 place and m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 Centered on point M, M (arbitrarily chosen, where M is a positive integer greater than 10 and less than 1000, and within this range, the larger M is, the higher the accuracy, but a larger M will also lead to a larger computational load) data points are selected to generate the input dataset. A neural network is used to identify whether there is a peak in the training dataset. Based on the neural network output (the neural network outputs 1 if it exists, and 0 if it does not, forming a 1×8 matrix of 1s and 0s), a 1×8 matrix (of fixed size, based on [2f]) is generated. c , 2f c +f m2 4f c , 4f c +f m2 , 4f c +0.5f m2 Do these 8 locations contain a target matrix for peak generation? If there is a peak, the position within the [] is 1; if there is no peak, it is 0.

[0066] S8. Based on the mixed signal target matrix, and the generated 1×8 matrix, calculate the maximum likelihood probability density of the matrix.

[0067] S9. If the target matrix with the highest probability is [1 1 1 1 1 0 1 0], then calculate the signal C. 20 C 21 Higher-order cumulants, according to C 20 / C 21 Determine the modulation method.

[0068] In a specific implementation, as a preferred embodiment of the present invention, in step S1, a second-order cyclic moment m is applied to any two combinations of the input signal: binary phase shift keying (bpsk), quadrature phase shift keying (qpsk), π / 4 quadrature phase shift keying (π / 4dqpsk), unbalanced four-phase phase shift keying (uqpsk), and 8-phase shift keying (8psk). 20 and fourth-order cyclic moment m 40 The calculations include:

[0069] m 20 (f,τ)= <r(t)r(t+τ)e -j2πft > t

[0070] m 40 (f,τ1,τ2,τ3,)= <r(t)r(t+τ1)r(t+τ2)r(t+τ3)e -j2πft > t

[0071] Where r(t) represents the input signal, τ, τ1, τ2, τ3 represent the time delay, and f represents the cycle frequency. t Indicates time average.

[0072] In a specific implementation, as a preferred embodiment of the present invention, in step S2, the second-order cyclic moment m according to each different code rate modulation scheme combination is... 20 The 2f of the calculation result c , 2f c +f m2 The fourth cyclic moment m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 At this point, determine whether a peak generation target matrix exists, including:

[0073] In the cyclic moment m 20 A signal with characteristics will be at m 20 2f in the cyclic spectrum c 2f c +f m1 2f c +f m2 The peak value appears at [location].

[0074] In the cyclic moment m 40 A signal with characteristics will be at m 40 4f in the cyclic spectrum c 4f c +f m1 4f c +f m2 , 4f c +0.5f m2 The peak value appears at [location].

[0075] According to [2f] c , 2f c +f m2 4f c , 4f c +f m2 , 4f c +0.5f m2 The target matrix is ​​generated based on whether there is a peak at position []. If there is a peak, the position within the bracket is 1; otherwise, it is 0. In this embodiment, the generated target matrix is ​​shown in Table 1.

[0076] Table 1 shows the target matrix used for maximum likelihood probability calculation after the neural network output matrix.

[0077]

[0078]

[0079] In a specific implementation, as a preferred embodiment of the present invention, in step S3, the second-order cyclic moment m for each different code rate modulation scheme combination is... 20 The 2f of the calculation result c , 2f c +f m2 and fourth-order cyclic moment m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 The training dataset is generated centered around the peak data and includes:

[0080] With 2f c , 2f c +f m2 4f c , 4f c +f m2 , 4f c +0.5f m2 For each of the M points to the left and right of the center (M is a positive integer greater than 10 and less than 1000, and the larger M is within this range, the higher the accuracy, but the larger M is, the greater the computational cost), a training dataset is generated, and a training dataset is generated at each peak.

[0081] Each generated training dataset is represented as: [C(kM)...C(k)...C(k+M)], where k represents the index of the peak cyclic frequency of the cyclic moment, C(k) represents the amplitude at k, and M represents half of the set sampling data point window, with a total of 2M+1 points. For example... Figure 3 As shown, the mixed signal m of bpsk and uqpsk is... 20 For example, in a cyclic matrix, select the data and input it into a classification neural network, setting the output to 1. Figure 4 As shown, the mixed signal m of bpsk and qpsk is used. 20Example of a cyclic moment. Inputting data with a peak into a classification neural network and setting the output to 1, and inputting data without a peak into the classification neural network and setting the output to 0, train a set of neural networks capable of recognizing single peaks using this type of dataset.

[0082] In a specific implementation, as a preferred embodiment of the present invention, in step S5, the m of the received signal is calculated. 21 Cyclic moments are used to determine the code rate of each of the two signals. The calculation expression is as follows:

[0083] m 21 (f,τ)= <r(t)r * (t+τ)e -j2πft > t

[0084] Where, r * (t) represents the conjugate of the input signal.

[0085] In a specific implementation, as a preferred embodiment of the present invention, in step S6, m is found. 21 The maximum value and its index [m1, i1], and the second largest value and its index [m2, i2] in the cyclic moments, where i1 and i2 are the cyclic frequencies of the two mixed signals, denoted as... f m2 ,include:

[0086] m 21 The cyclic spectrum of the cyclic moment contains two peaks. The horizontal axis corresponding to each peak represents the code rate of the two signals. Find the two peaks and find the code rate of the two signals according to the horizontal axis index of the two peaks.

[0087] In a preferred embodiment of the present invention, in step S8, the maximum likelihood probability density of the matrix is ​​calculated based on the generated 1×8 matrix according to the mixed signal target matrix. The calculation expression is as follows:

[0088]

[0089] Where A represents the target matrix, B represents the matrix output by the neural network, and n is the number of elements in the matrix.

[0090] In a specific implementation, as a preferred embodiment of the present invention, in step S9, if the maximum probability target matrix is ​​[1 1 1 1 1 0 1 0], then the signal C is calculated. 20 C 21 Higher-order cumulants, according to C 20 / C 21 Determining the modulation method includes:

[0091] If the target matrix with the highest probability is [1 1 1 1 1 0 1 0], then calculate the signal C. 20 C 21 Higher-order cumulants are as follows:

[0092] C 20 =E[r(t)r(t)]

[0093] C 21 =E[r(t)r * (t)]

[0094] Where E[·] represents the mean.

[0095] According to C 20 / C 21 Determine the modulation method: if it is greater than or equal to 0.9, it is a mixed signal of bpsk and bpsk; if it is greater than or equal to 0.5 and less than 0.9, it is a mixed signal of bpsk and uqpsk; if it is less than 0.5, it is a mixed signal of uqpsk and uqpsk.

[0096] Example

[0097] The input signals are bpsk and qpsk, with a bpsk signal code rate of 1.2288 Mbps and a qpsk signal code rate of 3.84 Mbps. The sampling rate is 120 MHz, the intermediate frequency is 12 MHz, the signal-to-noise ratio is 5 dB, the acquisition time is 0.5 ms, and the signal power ratio is 1:1.

[0098] First, a training dataset is generated based on the peak data at the cyclic frequency of various modulation signals. For each peak point, 10 data points to the left and right (21 points total) are used as one group of training data with an output of 1, and 10 data points to the left and right (21 points total) at points without peaks in the cyclic frequency are used as another group of training data with an output of 0. Each group contains 500 data points, generating a total of 1000 datasets. A backpropagation (BP) neural network is used as the classifier. The BP neural network has 21 input layers, 10 hidden layers, and 1 output layer.

[0099] Next, calculate the m of the signal. 21 Find the cyclic moments, including the maximum value and its index [0.06, 1228800], and the second largest value and its index [0.02, 3840000]. 1.228MHz, f m2 It is 3.84MHz.

[0100] Then calculate the m of the signal. 20 and m 40 Cyclic moments, in m 20 The 2f of the calculation result c , 2f c +fm2 place and m 40 The 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 Ten data points are selected to the left and right of the center point to generate the input dataset. A backpropagation (BP) neural network is used to identify whether there is a peak. A 1×8 matrix is ​​generated based on the neural network output: [0.9969 0.8977 0.1777 0.9024 0.1321 0.1250 0.8452 0.1578].

[0101] Finally, based on the target matrix of the mixed signal in Table 1, and using the 1×8 matrix generated in step 3, the maximum likelihood probability density of the matrix is ​​calculated to be 88.12%. Corresponding to serial number 1, the two mixed signal modulation methods are determined to be bpsk and qpsk.

[0102] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In the above embodiments of the present invention, the descriptions of each embodiment have their own emphasis; parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. It should be understood that the disclosed technical content in the several embodiments provided in this application can be implemented in other ways.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying satellite co-frequency mixed signal modulation, characterized in that, Includes the following steps: For any two combinations of input signals, namely binary phase shift keying (bpsk), quadrature phase shift keying (qpsk), π / 4 quadrature phase shift keying (π / 4dqpsk), unbalanced quadrature phase shift keying (uqpsk), and 8-phase shift keying (8psk), perform second-order cyclic moments m. 20 and fourth-order cyclic moment m 40 Calculation; Based on the second-order cyclic moment m under each different code rate modulation scheme combination 20 Calculation results 2f c +f m2 The fourth cyclic moment m 40 Calculation results 4f c +f m2 , 4f c +0.5f m2 At this point, determine whether a peak generation target matrix exists; Second-order cyclic moment m under each different code rate modulation scheme combination 20 The 2f of the calculation result c , 2f c +f m2 and fourth-order cyclic moment m 40 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 The training dataset is generated centered around the peak data. Receive the input signal and represent the input signal as: r(t) = r1(t) + r2(t) + n(t) Where r1(t) and r2(t) represent the respective modulation signals of the two mixed signals, and n(t) represents the noise; Calculate the m of the received signal 21 Cyclic moments are used to determine the code rate of each of the two signals. Find m 21 The maximum value and its index [m1, i1], and the second largest value and its index [m2, i2] in the cyclic moments, where i1 and i2 are the cyclic frequencies of the two mixed signals, denoted as... f m2 ; Calculate the m of the signal 20 and m 40 The cyclic moment, so as to be in m 20 The 2f of the calculation result c , 2f c +f m2 place and m 40 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 M data points are selected to the left and right of the center to generate the input dataset. A neural network is used to identify whether there is a peak in the training dataset. A 1×8 matrix is ​​generated based on the output of the neural network. Based on the mixed signal target matrix, the maximum likelihood probability density of the matrix is ​​calculated based on the generated 1×8 matrix. If the target matrix with the highest probability is [1 1 1 1 1 0 1 0], then calculate the signal C. 20 C 21 Higher-order cumulants, according to C 20 / C 21 Determine the modulation method.

2. The satellite co-frequency mixed signal modulation identification method according to claim 1, characterized in that, The second-order cyclic moment m is applied to any two combinations of the input signal binary phase shift keying (bpsk), quadrature phase shift keying (qpsk), π / 4 quadrature phase shift keying (π / 4dqpsk), unbalanced four-phase phase shift keying (uqpsk), and eight-phase phase shift keying (8psk). 20 and fourth-order cyclic moment m 40 The calculations include: m 20 (f,τ)= <r(t)r(t+τ)e -j2πft > t m 40 (f,τ1,τ2,τ3,)= <r(t)r(t+τ1)r(t+τ2)r(t+τ3)e -j2πft > t Where r(t) represents the input signal, τ, τ1, τ2, τ3 represent the time delay, and f represents the cycle frequency. t Indicates time average.

3. The satellite co-frequency mixed signal modulation identification method according to claim 1, characterized in that, The second-order cyclic moment m under each different code rate modulation scheme combination 20 The 2f of the calculation result c , 2f c +f m2 The fourth cyclic moment m 40 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 At this point, determine whether a peak generation target matrix exists, including: In the cyclic moment m 20 A signal with characteristics will be at m 20 2f in the cyclic spectrum c , 2f c +f m2 The peak value appears at [location]. In the cyclic moment m 40 A signal with characteristics will be at m 40 4f in the cyclic spectrum c , 4f c +f m2 , 4f c +0.5f m2 The peak value appears at [location]. according to Does the target matrix exist at a certain point? If there is a peak, the position within [] is 1; if there is no peak, it is 0.

4. The satellite co-frequency mixed signal modulation identification method according to claim 1, characterized in that, The second-order cyclic moment m under each different code rate modulation scheme combination 20 The 2f of the calculation result c , 2f c +f m2 and fourth-order cyclic moment m 40 4f of the calculation result c , 4f c +f m2 , 4f c +0.5f m2 The training dataset is generated centered around the peak data and includes: With 2f c , 2f c +f m2 4f c , 4f c +f m2 , 4f c +0.5f m2 For each of the M points to the left and right of the center, a training dataset is generated, and a training dataset is generated at each peak. Each generated training dataset is represented as: [C(kM)...C(k)...C(k+M)], where k represents the index of the peak cyclic frequency of the cyclic moment, C(k) represents the amplitude at k, and M represents half of the set sampling data point window, with a total number of points of 2M+1.

5. The satellite co-frequency mixed signal modulation identification method according to claim 1, characterized in that, The calculation of the received signal m 21 Cyclic moments are used to determine the code rate of each of the two signals. The calculation expression is as follows: m 21 (f,τ)= <r(t)r * (t+τ)e -j2πft > t Where, r * (t) represents the conjugate of the input signal.

6. The satellite co-frequency mixed signal modulation identification method according to claim 1, characterized in that, The search for m 21 The maximum value and its index [m1, i1], and the second largest value and its index [m2, i2] in the cyclic moments, where i1 and i2 are the cyclic frequencies of the two mixed signals, denoted as... f m2 ,include: m 21 The cyclic spectrum of the cyclic moment contains two peaks. The horizontal axis corresponding to each peak represents the code rate of the two signals. Find the two peaks and find the code rate of the two signals according to the horizontal axis index of the two peaks.

7. The satellite co-frequency mixed signal modulation identification method according to claim 1, characterized in that, Based on the mixed signal target matrix, and using the generated 1×8 matrix, the maximum likelihood probability density of the matrix is ​​calculated. The calculation expression is as follows: Where A represents the target matrix, B represents the matrix output by the neural network, and n is the number of elements in the matrix.

8. The satellite co-frequency mixed signal modulation identification method according to claim 1, characterized in that, If the target matrix with the highest probability is [1 1 1 1 1 0 1 0], then the signal C is calculated. 20 C 21 Higher-order cumulants, according to C 20 / C 21 Determining the modulation method includes: If the target matrix with the highest probability is [1 1 1 1 1 0 1 0], then calculate the signal C. 20 C 21 Higher-order cumulants are as follows: C 20 =E[r(t)r(t)] C 21 =E[r(t)r * (t)] Where E[·] represents the mean; According to C 20 / C 21 Determine the modulation method: if it is greater than or equal to 0.9, it is a mixed signal of bpsk and bpsk; if it is greater than or equal to 0.5 and less than 0.9, it is a mixed signal of bpsk and uqpsk; if it is less than 0.5, it is a mixed signal of uqpsk and uqpsk.

Images