Signal modulation mode identification method based on high-order cumulant and constellation diagram clustering characteristics

Abstract

A signal modulation mode recognition method based on high-order cumulant and constellation diagram clustering features comprises the following steps: (1) extracting second-order cumulant features, fourth-order cumulant features and sixth-order cumulant features of input signals, and extracting clustering center number features of input signal constellation diagrams by using a subtraction clustering algorithm; (2) selecting identification parameters and setting a threshold value; and (3) sending the extracted characteristic parameters into a classifier, and judging a modulation mode. The method is suitable for non-cooperative communication, and the modulation mode of the digital signal is identified under the condition of no prior information. According to the method, the high-order cumulant characteristics of the signals are extracted, and the clustering center number characteristics of the constellation diagram are extracted according to the discrimination of the high-order quadrature modulation signals. According to the characteristics of the two modes, identification parameters are optimally selected and threshold values are set, so that the identification effect of the algorithm is better than that of a method only using high-order cumulant, and the correct identification rate of six common modulation modes under the condition of low signal-to-noise ratio is improved.

Description

Technical Field

[0001] The invention discloses a signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering characteristics, and belongs to the field of signal processing. Background Art

[0002] Compared with cooperative communication, which has sufficient prior information for signal demodulation, the various parameters of signals in non-cooperative communication fields such as military reconnaissance, electronic countermeasures, and information security are opaque and rely on the blind demodulation technology of the receiver. Among them, the recognition of signal modulation mode is the basis for demodulating unknown signals.

[0003] The commonly used recognition method is mainly pattern recognition. Traditional pattern recognition methods are often based on the extraction of a single feature parameter and then directly feed it into a classifier for signal classification. The advantages of this method are simple theoretical analysis, convenient feature parameter extraction, and concise and clear classifier. However, the traditional pattern recognition method also has obvious shortcomings: when identifying the same type of modulation modes with similar orders, especially when it comes to high-order orthogonal modulation modes, it is difficult to find a suitable threshold for classification due to the small absolute value difference of a single feature parameter; secondly, when the signal-to-noise ratio of the input signal is small, the single feature parameter is too affected by the noise, resulting in a significant decrease in the recognition efficiency of the algorithm. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features, which solves the problem of low recognition success rate of existing recognition methods under low signal-to-noise ratio conditions and further improves the recognition accuracy.

[0005] The technical solution of the present invention is as follows:

[0006] A signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features comprises the following steps:

[0007] (1) Extract the second, fourth, and sixth order cumulative features of the input signal, and use the subtractive clustering algorithm to extract the number of cluster centers of the input signal constellation diagram;

[0008] (2) Select recognition parameters and set thresholds;

[0009] (3) Send the extracted feature parameters to the classifier to determine the modulation mode;

[0010] Preferably, extracting the high-order cumulants and the number of cluster centers of the constellation diagram of the input signal in the above step (1) comprises:

[0011] (1-1) Extract the second, fourth, and sixth order cumulant characteristics of the input signal. The calculation steps are as follows:

[0012] Step a: Let the input signal be X(t), t be the time parameter, let X1(t) = X(t), X2(t) = X(t+τ1),…,X k (t) = X(t + τ) k-1 ), τ is the time constant, k is the order, then the K-order cumulant of the input signal is:

[0013] C kx (τ1,τ2...τ k-1 )=Cum{X1(t)=X(t),X2(t)=X(t+τ1),...,X k (t) = X(t + τ) k-1 )}.;

[0014] Step b: Calculate the K-th moment M of the signal km =E[X(t) k-m X * (t) m ], where X * (t) and X(t) are in conjugate relationship;

[0015] Step c: According to Calculate the second, fourth, and sixth order cumulants of the signal:

[0016] C 20 =Cum(X,X)=M 20 =E[X(t)X(t)],

[0017]

[0018] Where ∑ represents a set of all disconnected ordered partitions X = (X1, X2...X k ), q represents the number of subsets, U p Represents the subscript set of elements in the pth subset among its q subsets.

[0019] (1-2) Use the subtractive clustering algorithm to extract the number of cluster centers of the signal constellation diagram. The calculation steps are as follows:

[0020] Step a: Assume that the constellation diagram of the signal has N points (x1, x2, ..., x n ), calculate the i-th point x i Density value Among them, take γ α =0.25;

[0021] Step b: Calculate the density values ​​of N points on the constellation diagram (D1, D2, ..., D N ), select the point x with the largest density value c1As the cluster center, the density value is D c1 , correct the density values ​​of all sample points Among them, γ β =1.5γ α ;

[0022] Step c: After one correction is completed, select the point with the largest density value among the remaining sample points as the next clustering center, and iterate step b M-1 times;

[0023] Step d: Compare D cM and D c(M-1) , when D cM Significantly smaller than D c(M-1) When , the iteration stops and the number of cluster centers is M-1.

[0024] Preferably, selecting identification parameters and setting thresholds in the above step (2) includes:

[0025] (2-1), the absolute values ​​of the five cumulative values ​​calculated according to (1-1) are |C 20 |,|C 40 |,|C 42 |,|C 60 |,|C 63 |, select two direct parameters f1=|C 20 |, f2=|C 40 | and two combined parameters In addition, the number of cluster centers f5 of the constellation diagram of the input signal is selected as the fifth identification parameter;

[0026] (2-2) Set the identification signal set to six modulation signals {BPSK (Binary Phase Shift Keying), QPSK (Quadrature Phase Shift Keying), 8PSK (8PhaseShift Keying), 8QAM (8Quadrature Amplitude Modulation), 16QAM (16Quadrature Amplitude Modulation), 64QAM (64Quadrature Amplitude Modulation)}, calculate the theoretical value of the identification parameter involved according to the definition of each modulation mode, and set the threshold according to the theoretical value, the steps are as follows:

[0027] Step a: The theoretical value of the characteristic parameter f1 of the signal set {BPSK, QPSK, 8PSK, 8QAM, 16QAM, 64QAM} is {1, 0, 0, 0, 0, 0}, and the threshold is set to 0.4;

[0028] Step b: The theoretical value of the characteristic parameter f2 of the signal set {QPSK, 8PSK, 8QAM, 16QAM, 64QAM} is {1, 0, 0.68, 0.68, 0.62}, and the threshold is set to 0.1;

[0029] Step c: The theoretical values ​​of the characteristic parameters of the signal set {QPSK, 8QAM, 16QAM, 64QAM} are {16, 15.8, 13.8, 13.6}, and the threshold is set to 15;

[0030] Step d: The theoretical value of the characteristic parameter f4 of the signal set {QPSK, 8QAM} is {4, 7.2}, and the threshold is set to 6;

[0031] Step e: The theoretical value of the characteristic parameter f5 of the signal set {16QAM, 64QAM} is {3, 4}, and the threshold is set to 3.5;

[0032] Preferably, in the above step (3), sending the extracted characteristic parameters to a classifier to determine the modulation mode comprises:

[0033] (3-1) Compare f1 with the threshold value 0.4. If f1 is less than or equal to 0.4, it is BPSK modulation and the recognition is complete. If f1 is greater than 0.4, go to step (3-2).

[0034] (3-2) Compare f2 with the threshold value 0.1. If it is less than or equal to 0.1, it is 8PSK modulation and the recognition is complete. If it is greater than 0.1, go to step (3-3).

[0035] (3-3) Compare f3 with the threshold value 15. If f3 is greater than 15, it is QPSK or 8QAM modulation, and go to step (4-4); if f3 is less than or equal to 15, it is 16QAM or 64QAM modulation, and go to step (4-5).

[0036] (3-4) Compare f4 with the threshold value 6. If it is less than or equal to 6, it is QPSK modulation and the recognition is complete. If it is greater than 6, it is 8QAM modulation and the recognition is complete.

[0037] (3-5) Compare f5 with the threshold value 3.5. If f5 is less than or equal to 3.5, it is 16QAM modulation and the recognition is complete. If f5 is greater than 3.5, it is 64QAM modulation and the recognition is complete.

[0038] The beneficial effects of the present invention are:

[0039] The present invention is applicable to the recognition of the modulation mode of digital signals in non-cooperative communication without prior information. The present invention not only extracts the high-order cumulative amount characteristics of the signal, but also extracts the number characteristics of the cluster centers of the constellation diagram for the discrimination of high-order orthogonal modulation signals. According to the characteristics of the two, the recognition parameters are selected and the thresholds are set, so that the recognition effect of the algorithm is better than that of the method using only high-order cumulative amounts, and the correct recognition rate of six commonly used modulation modes under low signal-to-noise ratio conditions is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 is a flow chart of the present invention;

[0041] Figure 2 is the modulation mode decided by the classifier;

[0042] Figure 3 It is a data table of the simulation experiment of the embodiment. DETAILED DESCRIPTION

[0043] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0044] Example:

[0045] like Figure 1 As shown, a signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features specifically includes the following steps:

[0046] (1) Extract the second, fourth, and sixth order cumulative features of the input signal, and use the subtractive clustering algorithm to extract the number of cluster centers of the input signal constellation diagram;

[0047] (1-1) Extract the second, fourth, and sixth order cumulant characteristics of the input signal. The calculation steps are as follows:

[0048] Step a: Let the input signal be X(t), t be the time parameter, let X1(t) = X(t), X2(t) = X(t+τ1),…,X k (t) = X(t + τ) k-1 ), τ is the time constant, k is the order, then the K-order cumulant of the input signal is:

[0049] C kx (τ1,τ2...τ k-1 )=Cum{X1(t)=X(t),X2(t)=X(t+τ1),...,X k (t) = X(t + τ) k-1 )}.;

[0050] Step b: Calculate the K-th moment M of the signal km=E[X(t) k-m X * (t) m ], where X * (t) and X(t) are in conjugate relationship;

[0051] Step c: According to Calculate the second, fourth, and sixth order cumulants of the signal:

[0052] C 20 =Cum(X,X)=M 20 =E[X(t)X(t)],

[0053]

[0054] Where ∑ represents a set of all disconnected ordered partitions X = (X1, X2...X k ), q represents the number of subsets, U p Represents the subscript set of elements in the pth subset among its q subsets.

[0055] (1-2) Use the subtractive clustering algorithm to extract the number of cluster centers of the signal constellation diagram. The calculation steps are as follows:

[0056] Step a: Assume that the constellation diagram of the signal has N points (x1, x2, ..., x n ), calculate the i-th point x i Density value Among them, take γ α =0.25;

[0057] Step b: Calculate the density values ​​of N points on the constellation diagram (D1, D2, ..., D N ), select the point x with the largest density value c1 As the cluster center, the density value is D c1 , correct the density values ​​of all sample points Among them, γ β =1.5γ α ;

[0058] Step c: After one correction is completed, select the point with the largest density value among the remaining sample points as the next clustering center, and iterate step b M-1 times;

[0059] Step d: Compare D cM and D c(M-1) , when D cM Significantly smaller than D c(M-1) When , the iteration stops and the number of cluster centers is M-1.

[0060] (2) Select recognition parameters and set thresholds;

[0061] (2-1), the absolute values ​​of the five cumulative values ​​calculated according to (1-1) are |C 20 |,|C 40 |,|C 42 |,|C 60 |,|C 63 |, select two direct parameters f1=|C 20 |, f2=|C 40 | and two combined parameters In addition, the number of cluster centers f5 of the constellation diagram of the input signal is selected as the fifth identification parameter;

[0062] (2-2) Set the identification signal set to six modulation signals {BPSK (Binary Phase Shift Keying), QPSK (Quadrature Phase Shift Keying), 8PSK (8PhaseShift Keying), 8QAM (8Quadrature Amplitude Modulation), 16QAM (16Quadrature Amplitude Modulation), 64QAM (64Quadrature Amplitude Modulation)}, calculate the theoretical value of the identification parameter involved according to the definition of each modulation mode, and set the threshold according to the theoretical value, the steps are as follows:

[0063] Step a: The theoretical value of the characteristic parameter f1 of the signal set {BPSK, QPSK, 8PSK, 8QAM, 16QAM, 64QAM} is {1, 0, 0, 0, 0, 0}, and the threshold is set to 0.4;

[0064] Step b: The theoretical value of the characteristic parameter f2 of the signal set {QPSK, 8PSK, 8QAM, 16QAM, 64QAM} is {1, 0, 0.68, 0.68, 0.62}, and the threshold is set to 0.1;

[0065] Step c: The theoretical values ​​of the characteristic parameters of the signal set {QPSK, 8QAM, 16QAM, 64QAM} are {16, 15.8, 13.8, 13.6}, and the threshold is set to 15;

[0066] Step d: The theoretical value of the characteristic parameter f4 of the signal set {QPSK, 8QAM} is {4, 7.2}, and the threshold is set to 6;

[0067] Step e: The theoretical value of the characteristic parameter f5 of the signal set {16QAM, 64QAM} is {3, 4}, and the threshold is set to 3.5;

[0068] (3) Send the extracted feature parameters into Figure 2 The classifier shown decides the modulation mode;

[0069] (3-1) Compare f1 with the threshold value 0.4. If f1 is less than or equal to 0.4, it is BPSK modulation and the recognition is complete. If f1 is greater than 0.4, go to step (3-2).

[0070] (3-2) Compare f2 with the threshold value 0.1. If it is less than or equal to 0.1, it is 8PSK modulation and the recognition is complete. If it is greater than 0.1, go to step (3-3).

[0071] (3-3) Compare f3 with the threshold value 15. If f3 is greater than 15, it is QPSK or 8QAM modulation, and go to step (3-4). If f3 is less than or equal to 15, it is 16QAM or 64QAM modulation, and go to step (3-5).

[0072] (3-4) Compare f4 with the threshold value 6. If it is less than or equal to 6, it is QPSK modulation and the recognition is complete. If it is greater than 6, it is 8QAM modulation and the recognition is complete.

[0073] (3-5) Compare f5 with the threshold value 3.5. If f5 is less than or equal to 3.5, it is 16QAM modulation and the recognition is complete. If f5 is greater than 3.5, it is 64QAM modulation and the recognition is complete.

[0074] The simulation experiment was conducted using MATLAB software, simulating the signal-to-noise ratio range of 0-30dB, the step length of 1dB, 100 groups of each type of modulation mode, and a total of 600 groups of modulation signals were input into the algorithm of the present invention. The results are as follows Figure 3 As shown in the figure, starting from a signal-to-noise ratio of 4dB, the recognition accuracy of the five modulation modes of BPSK, QPSK, 8PSK, 8QAM, and 16QAM can all reach 90%. In this sample, the recognition accuracy of the BPSK and 8PSK modulation modes can reach nearly 100% starting from 0dB; in addition, starting from a signal-to-noise ratio of 15dB, the recognition rate of the 64QAM modulation mode can also reach 90%. The results show the excellent performance of the algorithm.

[0075] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features, characterized in that: The steps include: (1) Extract the second, fourth, and sixth order cumulative features of the input signal, and use the subtractive clustering algorithm to extract the number of cluster centers of the input signal constellation diagram; (2) Select recognition parameters and set thresholds; (3) Send the extracted feature parameters to the classifier to determine the modulation mode.

2. The signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features according to claim 1 is characterized in that: In step (1), the second, fourth, and sixth order cumulative feature of the input signal are extracted, and the calculation steps are as follows: Step 2a: Let the signal be X(t), t be the time parameter, let X1(t) = X(t), X2(t) = X(t+τ1), …, X k (t) = X(t + τ) k-1 )τ is the time constant, k is the order, then the K-order cumulant of the signal is: C kx (τ1,τ2...τ k-1 )=Cum{X1(t)=X(t),X2(t)=X(t+τ1),...,X k (t)=X(t+τ k-1 )}.; Step 2b: Calculate the K-th moment M of the signal km =E[X(t) k-m X * (t) m ], where X * (t) and X(t) are in conjugate relationship; Step 2c: Based on Calculate the second, fourth, and sixth order cumulants of the signal: Where ∑ represents a set of all disconnected ordered partitions X = (X1, X2...X k ), q represents the number of subsets, U p Represents the subscript set of elements in the pth subset among its q subsets.

3. The signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features according to claim 1 is characterized in that: In the step (1), a subtractive clustering algorithm is used to extract the cluster center number feature of the input signal constellation diagram, and the calculation steps are as follows: Step 3a: Assume that the constellation diagram of the input signal has N points (x1, x2, ..., x n ), calculate the i-th point x i Density value Among them, take γ α =0.25; Step 3b: Calculate the density values ​​of N points on the constellation diagram (D1, D2, ..., D N ), select the point x with the largest density value c1 As the cluster center, the density value is D c1 , correct the density values ​​of all sample points Among them, γ β =1.5γ α ; Step 3c: After one correction is completed, select the point with the largest density value among the remaining sample points as the next cluster center, and iterate step 3b M-1 times; Step 3d: Compare D cM and D c(M-1) , when D cM Significantly smaller than D c(M-1) When , the iteration stops and the number of cluster centers is M-1.

4. The signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features according to claim 2 is characterized in that: The step (2) of selecting identification parameters and setting thresholds includes: (4-1), the absolute values ​​of the five cumulative values ​​calculated according to step (1) are |C 20 |,|C 40 |,|C 42 |,|C 60 |,|C 63 |, select two direct parameters f1=|C 20 |, f2=|C 40 | and two combined parameters In addition, the number of cluster centers f5 of the constellation diagram of the input signal is selected as the fifth identification parameter; (4-2) Set the identification signal set to six modulation signals {binary phase shift keying, quadrature phase shift keying, octal phase shift keying, octal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation}, calculate the theoretical value of the identification parameter involved according to the definition of each modulation mode, and set the threshold value according to the theoretical value, the steps are as follows: Step 4a: The theoretical value of the characteristic parameter f1 of the signal set {binary phase shift keying, quadrature phase shift keying, octal phase shift keying, octal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation} is {1, 0, 0, 0, 0, 0}, and the threshold is set to 0.4; Step 4b: The theoretical value of the characteristic parameter f2 of the signal set {binary phase shift keying, quadrature phase shift keying, octal phase shift keying, octal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation} is {1, 0, 0.68, 0.68, 0.62}, and the threshold is set to 0.1; Step 4c: The theoretical values ​​of the characteristic parameters of the signal set {binary phase shift keying, octal phase shift keying, hexadecimal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation} are {16, 15.8, 13.8, 13.6}, and the threshold is set to 15; Step 4d: The theoretical value of the characteristic parameter f4 of the signal set {quadrature phase shift keying, octal quadrature amplitude modulation} is {4, 7.2}, and the threshold is set to 6; Step 4e: The theoretical value of the characteristic parameter f5 of the signal set {hexadecimal quadrature amplitude modulation, hexadecimal quadrature amplitude modulation} is {3, 4}, and the threshold is set to 3.

5.

5. The signal modulation pattern recognition method based on high-order cumulants and constellation diagram clustering features according to claim 4 is characterized in that: The step (3) of sending the extracted characteristic parameters to the classifier to determine the modulation mode includes: (5-1), compare f1 with the threshold value 0.4, if it is less than or equal to 0.4, it is binary phase shift keying modulation, and the recognition is completed; if it is greater than 0.4, go to step (5-2); (5-2), compare f2 with the threshold value 0.1, if it is less than or equal to 0.1, it is octal phase shift keying modulation, and the recognition is completed; if it is greater than 0.1, go to step (5-3); (5-3), compare f3 with the threshold value 15, if it is greater than 15, it is quadrature phase shift keying or octal quadrature amplitude modulation, and go to step (4-4); if it is less than or equal to 15, it is hexadecimal quadrature amplitude modulation or hexadecimal quadrature amplitude modulation, and go to step (4-5); (5-4), compare f4 with the threshold value 6. If it is less than or equal to 6, it is quadrature phase shift keying modulation and the recognition is completed; if it is greater than 6, it is octal quadrature amplitude modulation and the recognition is completed; (5-5) Compare f5 with the threshold value 3.

5. If it is less than or equal to 3.5, it is hexadecimal quadrature amplitude modulation and the recognition is complete. If it is greater than 3.5, it is hexadecimal quadrature amplitude modulation and the recognition is complete.

Images