MFSK type modulation identification method based on instantaneous frequency two-step clustering

Abstract

The invention discloses an MFSK class modulation identification method based on instantaneous frequency two-step clustering, which realizes intra-class identification of an MFSK signal in a severe communication environment with high dynamic Doppler frequency offset and nonlinear distortion, and has good practical value. Aiming at three common MFSK signals of 2FSK, 4FSK and 8FSK, the technical scheme provided by the invention can perform effective intra-class identification on the MFSK signal when the MFSK signal passes through a non-linear and high-dynamic Doppler frequency offset channel. Besides, according to the technical scheme provided by the invention, each group of data blocks are averaged, so that the influence caused by factors such as noise, frequency hopping and frequency jitter can be reduced, the instantaneous frequency characteristic difference of the 4FSK signal and the 8FSK signal is highlighted, and the identification of the 4FSK signal and the 8FSK signal is realized.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to an MFSK type modulation recognition method based on instantaneous frequency two-step clustering. Background Art

[0002] Modulation recognition technology is a technology that uses signal prior information or statistical information to identify the modulation mode of a signal. Correctly identifying the modulation mode of a signal is a prerequisite for signal demodulation. Frequency shift keying (FSK) signals are widely used in sea, land, air transportation and military communications due to their relatively good anti-interference performance, and play a vital role in wireless communications. However, under non-cooperative communication conditions, the complex communication environment makes the channel worse. Channel characteristics such as Doppler frequency deviation and nonlinear effects make it difficult to identify multiple frequency shift keying (MFSK) signals within a class. In order to solve the above problems, it is particularly important to design an MFSK intra-class identification scheme with good robustness under complex communication conditions.

[0003] Traditional MFSK intra-class identification schemes are usually implemented by extracting statistical features of the signal, such as instantaneous frequency mean, high-order cumulants, power spectrum entropy, wavelet transform entropy, etc. The algorithm principle adopted by this type of scheme is relatively simple, and the threshold is easy to divide. However, these methods have obvious limitations. For example, when the modulation index of two FSK signals differs greatly, the scheme based on the instantaneous frequency mean is difficult to make a correct judgment on the order of the FSK signal; the scheme based on high-order cumulants and the scheme based on transform domain entropy cannot effectively combat channel distortion. Therefore, the universality of this type of scheme still has room for improvement.

[0004] On the other hand, with the development of artificial intelligence technology, MFSK intra-class recognition algorithms based on machine learning have gradually become an emerging research topic. Such algorithms include those based on support vector machines, residual neural networks, and multi-feature fusion networks. However, algorithms based on artificial intelligence technology usually require a large amount of labeled training data, which is extremely difficult to obtain in non-cooperative communication scenarios. At the same time, such algorithms perform poorly when facing signals with abnormal distortion and are not suitable for communication systems with poor channel conditions. Summary of the invention

[0005] In order to solve the limitations and defects of the prior art, the present invention provides an MFSK modulation recognition method based on instantaneous frequency two-step clustering, comprising:

[0006] Step S1, calculating the linear phase of the MFSK signal, wherein the MFSK signal includes a 2FSK signal, a 4FSK signal and an 8FSK signal;

[0007] Step S2, calculating the zero center instantaneous frequency of the MFSK signal according to the linear phase;

[0008] Step S3, using a subtractive clustering method to perform a coarse clustering operation according to the absolute value of the zero-center instantaneous frequency, and distinguishing the 2FSK signal according to the number of cluster centers;

[0009] Step S4, calculating the Euclidean distance of each zero-center instantaneous frequency absolute value relative to each cluster center, and regrouping the zero-center instantaneous frequency absolute value according to the Euclidean distance to obtain regrouped data;

[0010] Step S5, dividing the data of the same group in the regrouped data into blocks, and averaging each block of data to obtain reorganized data;

[0011] Step S6: Use a subtractive clustering method to perform fine clustering operations on the reorganized data, and distinguish the 4FSK signal from the 8FSK signal according to the number of cluster centers.

[0012] Optionally, the step S1 includes:

[0013] Performing amplitude normalization processing on the MFSK signal;

[0014] The real part I(i) and the imaginary part Q(i) of the MFSK signal are set, and the instantaneous phase of the MFSK signal is calculated. The expression is as follows:

[0015]

[0016] The instantaneous phase is corrected to a linear phase using a phase correction sequence, and the expression is as follows:

[0017]

[0018] Where, C(0)=0;

[0019] Define the linear phase sequence as Among them, θ(i) is the instantaneous phase and C(i) is the linear phase.

[0020] Optionally, step S2 includes:

[0021] The instantaneous frequency of the MFSK signal is calculated according to the linear phase sequence difference, and the expression is as follows:

[0022]

[0023] Among them, f s is the sampling rate of the signal;

[0024] The instantaneous frequency is subtracted from the mean value of the instantaneous frequency to obtain the zero-center instantaneous frequency of the MFSK signal.

[0025] Optionally, step S3 includes:

[0026] Calculate element x i With other elements x j The density index is as follows:

[0027]

[0028] Among them, the radius r a is a positive number, used to define the element x i Neighborhood of;

[0029] In the first round of iteration, we first select the element x with the largest density index. m1 As the first cluster center, element x m1 The corresponding density index is D m1 , and then correct each element x according to the following expression i Density index:

[0030]

[0031] Among them, the constant r b It is used to define the neighborhood where the decrease of density index is greater than the preset threshold, r b Greater than r a ;

[0032] When performing the kth iteration, the density index is corrected according to the following expression:

[0033]

[0034] Set the preset threshold δ, when D mk / D m1 When ≤δ, the iteration process is stopped and k-1 cluster centers are obtained;

[0035] The 2FSK signal is distinguished according to the number of cluster centers.

[0036] Optionally, step S4 includes:

[0037] Calculate the Euclidean distance of the absolute value of each zero-centered instantaneous frequency relative to each cluster center;

[0038] Select the cluster center with the smallest Euclidean distance as the judgment result;

[0039] According to the judgment result, each zero-center instantaneous frequency absolute value is divided into two groups to complete the regrouping of the data. The expression is as follows:

[0040]

[0041] Among them, G1 is the first group, b n is the absolute value of the instantaneous frequency at the nth zero center, d 1n for b n The Euclidean distance to the cluster center C1, d 2n for b n The Euclidean distance to the cluster center C2, G2 is the second group, b m is the absolute value of the instantaneous frequency at the mth zero center, d 1m for b m The Euclidean distance to the cluster center C1, d 2m for b m The Euclidean distance to cluster center C2.

[0042] Optionally, step S5 includes:

[0043] The regrouped data is averaged according to the group, and the data of the same group is divided into blocks, and the average of each block of data is obtained to obtain the reorganized data. The expression is as follows:

[0044]

[0045] Where L represents the data block length, r1(i) represents the reorganized data of group G1, represents the kth data in the data belonging to group G1, r2(i) represents the reorganized data of group G2, Indicates the k-th data among the data belonging to group G2.

[0046] Optionally, step S6 includes:

[0047] The reshaped data are subjected to fine clustering using the subtractive clustering method, and the constant r b The sum threshold δ is set to a preset value to achieve fine clustering of signals;

[0048] The number of cluster centers obtained for the 4FSK signal is 2, and the number of cluster centers obtained for the 8FSK signal is greater than or equal to 4. The 4FSK signal and the 8FSK signal are distinguished according to the number of cluster centers.

[0049] The present invention has the following beneficial effects:

[0050] The MFSK modulation recognition method based on instantaneous frequency two-step clustering provided by the present invention realizes the intra-class recognition of MFSK signals in a harsh communication environment with high dynamic Doppler frequency deviation and nonlinear distortion, and has good practical value. For three common MFSK signals of 2FSK, 4FSK and 8FSK, the technical solution provided by the present invention can effectively identify the MFSK signal within the class when the MFSK signal passes through a channel with nonlinearity and high dynamic Doppler frequency deviation. In addition, the technical solution provided by the present invention averages each group of data in blocks, which can reduce the influence of factors such as noise, frequency hopping, and frequency jitter, highlight the difference in instantaneous frequency characteristics between 4FSK signals and 8FSK signals, and realize the recognition of 4FSK signals and 8FSK signals. In summary, the technical solution provided by the present invention realizes the intra-class recognition of MFSK signals under channel conditions with large Doppler frequency deviation and obvious nonlinear effect, has reliable recognition accuracy, and has good performance, and has good practical value in actual communication systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flow chart of the MFSK-type modulation recognition method based on two-step clustering of instantaneous frequency provided in the first embodiment of the present invention.

[0052] Figure 2 A schematic diagram of a curve showing a variation of the signal-to-noise ratio for MFSK-type modulation recognition accuracy based on two-step clustering of instantaneous frequency in a linear channel provided by the first embodiment of the present invention.

[0053] Figure 3 This is a schematic diagram of a curve showing a variation of the signal-to-noise ratio for MFSK-type modulation recognition accuracy based on two-step clustering of instantaneous frequencies under a nonlinear channel provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to enable those skilled in the art to better understand the technical solution of the present invention, the MFSK-type modulation recognition method based on two-step clustering of instantaneous frequency provided by the present invention is described in detail below with reference to the accompanying drawings.

[0055] Embodiment 1

[0056] 1. General Introduction

[0057] This embodiment studies the instantaneous frequency characteristics of MFSK, combines the instantaneous frequency characteristics of MFSK signals with clustering algorithms, and proposes an MFSK modulation recognition method based on two-step clustering of instantaneous frequency on the basis of traditional recognition algorithms. This method can realize intra-class recognition of MFSK signals in harsh communication environments with high Doppler frequency deviation, high dynamics, and nonlinear distortion, and has good practical value.

[0058] 2. MFSK modulation recognition scheme

[0059] This embodiment proposes a MFSK modulation recognition method based on instantaneous frequency two-step clustering. This method is targeted at three common MFSK signals: 2FSK, 4FSK, and 8FSK. It can effectively identify MFSK signals within a class when the signals pass through a nonlinear channel with high dynamic Doppler frequency deviation. The overall flow chart of this method is shown in the figure. Figure 1 shown.

[0060] Step 1: Calculate the linear phase of the signal

[0061] Take the real part I(i) and imaginary part Q(i) of the amplitude-normalized signal, and then calculate the instantaneous phase of the signal according to the following formula:

[0062]

[0063] Since the instantaneous phase range is only [-π,π], when the true phase value exceeds this range, a jump of 2π will occur. Therefore, the phase correction sequence is calculated to correct the instantaneous phase to a linear phase. The phase correction sequence is as follows:

[0064]

[0065] Where C(0) = 0. The linear phase sequence is defined as:

[0066]

[0067] Step 2: Calculate the zero-centered instantaneous frequency

[0068] The instantaneous frequency of the signal can be calculated by using the linear phase sequence difference, which is the following formula:

[0069]

[0070] where f s is the sampling rate of the signal.

[0071] The zero-center instantaneous frequency is obtained by subtracting the average of the above instantaneous frequencies from the above instantaneous frequencies.

[0072] Step 3: Coarse clustering to distinguish 2FSK

[0073] The absolute value of the zero-center instantaneous frequency obtained in step 2 is clustered. This embodiment uses a subtractive clustering method for clustering, and one cluster center can be obtained for a 2FSK signal; and two cluster centers can be obtained for 4FSK and 8FSK. Therefore, 2FSK can be distinguished first based on the number of cluster centers.

[0074] The process of obtaining the cluster center using the subtractive clustering algorithm is as follows:

[0075] (1) Calculate an element xi Density indicators with other elements:

[0076]

[0077] The radius r a is a positive number that defines a neighborhood of the element.

[0078] (2) In the first round of iteration, the element with the largest density index is first selected as the first cluster center, denoted by x m1 , and its corresponding density index is marked as D m1 . Then, each element x is modified according to the following formula i Density index:

[0079]

[0080] The constant r b Defines a neighborhood where the density index is significantly reduced, usually r b Greater than r a , to avoid cluster centers that are very close to each other.

[0081] (3) Similarly, when performing the kth iteration, the density index is corrected according to the following formula:

[0082]

[0083] (4) Set a suitable threshold δ, when D mk / D m1 When ≤δ, the iteration process stops and k-1 cluster centers are obtained.

[0084] In the above algorithm, r b If δ is set to a larger value and the threshold δ is set to a larger value, coarse clustering of the signal can be achieved.

[0085] Step 4: Regrouping of the absolute values ​​of the instantaneous frequencies centered at zero

[0086] For a 4FSK signal, the absolute value of its zero-center instantaneous frequency ideally includes only 2 values; whereas for an 8FSK signal, the absolute value of its zero-center instantaneous frequency ideally includes 4 values.

[0087] Taking the cluster center obtained in step 3 as the basis for judgment, calculate the Euclidean distance of each zero-center instantaneous frequency absolute value relative to each cluster center. For each zero-center instantaneous frequency absolute value, select the cluster center with the smallest Euclidean distance as its judgment result. Then, divide each zero-center instantaneous frequency absolute value into two groups according to its judgment result, and complete the regrouping of the data. That is:

[0088]

[0089] Among them, b n is the absolute value of the instantaneous frequency at the nth zero center, d 1n , d 2n b n Euclidean distance to cluster centers C1, C2.

[0090] Step 5: Regroup the data into blocks and average them to get the reshaped data

[0091] The regrouped data in step 4 are averaged by group, that is, for the data of the same group, every several data blocks are averaged to obtain the reorganized data. That is:

[0092]

[0093] Where L represents the length of the data block. Indicates the k-th data among the data belonging to group G1.

[0094] In this step, for the 4FSK signal, it has 4 frequency points and the absolute value of the zero-center instantaneous frequency has only 2 values. Under ideal conditions, the same set of data only contains 1 absolute value of the zero-center instantaneous frequency, so its corresponding reshaped data only contains 2 values; and for the 8FSK signal, it has 8 frequency points and the absolute value of the zero-center instantaneous frequency has 4 values. Under ideal conditions, the same set of data contains 2 absolute values ​​of the zero-center instantaneous frequency, so its corresponding reshaped data contains more than 4 values.

[0095] In this step, averaging each group of data blocks can reduce the impact of factors such as noise, frequency hopping, and frequency jitter, and highlight the difference in instantaneous frequency characteristics between 4FSK and 8FSK.

[0096] Step 6: Fine clustering to distinguish 4FSK from 8FSK

[0097] Based on the subtractive clustering method in step 3, a fine clustering operation is performed on the reorganized data, and r b And the threshold δ is set to a suitable value, which can achieve fine clustering of the signal. For 4FSK signal, 2 cluster centers can be obtained; and for 8FSK signal, 4 or even more cluster centers can be obtained. Therefore, 4FSK can be distinguished from 8FSK based on the number of cluster centers.

[0098] The MFSK modulation identification method based on two-step clustering of instantaneous frequency provided in this embodiment realizes intra-class identification of MFSK signals in harsh communication environments with high dynamic Doppler frequency deviation and nonlinear distortion, and has good practical value. For the three common MFSK signals of 2FSK, 4FSK and 8FSK, the technical solution provided in this embodiment can effectively identify the MFSK signals within the class when the MFSK signals pass through a channel with nonlinear and high dynamic Doppler frequency deviation. In addition, the technical solution provided in this embodiment averages each group of data in blocks, which can reduce the influence of factors such as noise, frequency hopping, and frequency jitter, highlight the difference in instantaneous frequency characteristics between 4FSK signals and 8FSK signals, and realize the identification of 4FSK signals and 8FSK signals.

[0099] 3. Simulation Results

[0100] This embodiment simulates and tests the performance of the MFSK modulation recognition method based on two-step clustering of instantaneous frequency under linear channels and nonlinear channels respectively. The linear channel adopts a single-path Gaussian white noise channel model with Doppler frequency deviation, and the nonlinear channel is generated by cascading a linear channel and a Volterra nonlinear channel. The test environment parameters of the linear channel are shown in Table 1, and the Volterra nonlinear channel parameters are shown in Table 2. The modulation mode set of the test signal is {2FSK, 4FSK, 8FSK}. Among them, the modulation index of the three modulation modes is 0.375, and the upsampling multiples are 6 times, 12 times, and 24 times, respectively. It should be noted that different upsampling multiples are adopted to make the ratio of the bandwidth occupied by the signal on the spectrum to the sampling frequency consistent, so as to simulate the situation of artificially screening signals (the proposed scheme is still applicable under the same upsampling multiple). Under the conditions of linear channels and nonlinear channels, the recognition accuracy curves of various modulation modes obtained by testing are as follows: Figure 2 , Figure 3 shown.

[0101] Table 1 Test environment parameters

[0102]

[0103] Table 2 Volterra nonlinear channel parameters

[0104]

[0105] from Figure 2 It can be seen that the scheme proposed in this embodiment can achieve relatively reliable intra-class recognition of MFSK under conventional linear channel conditions. When the signal-to-noise ratio is 8dB, the recognition accuracy of the three FSK signals reaches more than 97%.

[0106] from Figure 3It can be seen that the scheme proposed in this embodiment can still achieve relatively reliable MFSK intra-class recognition even under relatively poor communication conditions. In the test under nonlinear channel conditions, for the three FSK signals, when the signal-to-noise ratio is 8dB, the recognition accuracy of the three modulation modes of the proposed scheme reaches more than 95%.

[0107] In summary, the scheme proposed in this embodiment realizes MFSK intra-class recognition under channel conditions with large Doppler frequency deviation and obvious nonlinear effect, and has a relatively reliable recognition accuracy, good performance, and good practical value in actual communication systems.

[0108] It is to be understood that the above embodiments are merely exemplary embodiments used to illustrate the principles of the present invention, but the present invention is not limited thereto. For those of ordinary skill in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A MFSK modulation recognition method based on two-step clustering of instantaneous frequency, characterized in that: include: Step S1, calculating the linear phase of the MFSK signal, wherein the MFSK signal includes a 2FSK signal, a 4FSK signal and an 8FSK signal; Step S2, calculating the zero center instantaneous frequency of the MFSK signal according to the linear phase; Step S3, using a subtractive clustering method to perform a coarse clustering operation according to the absolute value of the zero-center instantaneous frequency, and distinguishing the 2FSK signal according to the number of cluster centers; Step S4, calculating the Euclidean distance of each zero-center instantaneous frequency absolute value relative to each cluster center, and regrouping the zero-center instantaneous frequency absolute value according to the Euclidean distance to obtain regrouped data; Step S5, dividing the data of the same group in the regrouped data into blocks, and averaging each block of data to obtain reorganized data; Step S6: Use a subtractive clustering method to perform fine clustering operations on the reorganized data, and distinguish the 4FSK signal from the 8FSK signal according to the number of cluster centers.

2. The MFSK modulation recognition method based on instantaneous frequency two-step clustering according to claim 1 is characterized in that: The step S1 comprises: Performing amplitude normalization processing on the MFSK signal; The real part I(i) and the imaginary part Q(i) of the MFSK signal are set, and the instantaneous phase of the MFSK signal is calculated. The expression is as follows: The instantaneous phase is corrected to a linear phase using a phase correction sequence, and the expression is as follows: Where, C(0)=0; Define the linear phase sequence as Among them, θ(i) is the instantaneous phase and C(i) is the linear phase.

3. The MFSK modulation recognition method based on instantaneous frequency two-step clustering according to claim 2 is characterized in that: The step S2 comprises: The instantaneous frequency of the MFSK signal is calculated according to the linear phase sequence difference, and the expression is as follows: Among them, f s is the sampling rate of the signal; The instantaneous frequency is subtracted from the mean value of the instantaneous frequency to obtain the zero-center instantaneous frequency of the MFSK signal.

4. The MFSK modulation recognition method based on instantaneous frequency two-step clustering according to claim 3 is characterized in that: The step S3 comprises: Calculate element x i With other elements x j The density index is as follows: Among them, the radius r a is a positive number, used to define the element x i Neighborhood of; In the first round of iteration, we first select the element x with the largest density index. m1 As the first cluster center, element x m1 The corresponding density index is D m1 , and then correct each element x according to the following expression i Density index: Among them, the constant r b It is used to define the neighborhood where the decrease of density index is greater than the preset threshold, r b Greater than r a ; When performing the kth iteration, the density index is corrected according to the following expression: Set the preset threshold δ, when D mk / D m1 When ≤δ, the iteration process is stopped and k-1 cluster centers are obtained; The 2FSK signal is distinguished according to the number of cluster centers.

5. The MFSK modulation recognition method based on instantaneous frequency two-step clustering according to claim 4 is characterized in that: The step S4 comprises: Calculate the Euclidean distance of the absolute value of each zero-centered instantaneous frequency relative to each cluster center; Select the cluster center with the smallest Euclidean distance as the judgment result; According to the judgment result, each zero-center instantaneous frequency absolute value is divided into two groups to complete the regrouping of the data. The expression is as follows: Among them, G1 is the first group, b n is the absolute value of the instantaneous frequency at the nth zero center, d 1n for b n The Euclidean distance to the cluster center C1, d 2n for b n The Euclidean distance to the cluster center C2, G2 is the second group, b m is the absolute value of the instantaneous frequency at the mth zero center, d 1m for b m The Euclidean distance to the cluster center C1, d 2m for b m The Euclidean distance to cluster center C2.

6. The MFSK modulation recognition method based on instantaneous frequency two-step clustering according to claim 5 is characterized in that: The step S5 comprises: The regrouped data is averaged according to the group, and the data of the same group is divided into blocks, and the average of each block of data is obtained to obtain the reorganized data. The expression is as follows: Where L represents the data block length, r1(i) represents the reorganized data of group G1, represents the kth data in the data belonging to group G1, r2(i) represents the reorganized data of group G2, Indicates the k-th data among the data belonging to group G2.

7. The MFSK modulation recognition method based on instantaneous frequency two-step clustering according to claim 6 is characterized in that: The step S6 comprises: The reshaped data are subjected to fine clustering using the subtractive clustering method, and the constant r b The sum threshold δ is set to a preset value to achieve fine clustering of signals; The number of cluster centers obtained for the 4FSK signal is 2, and the number of cluster centers obtained for the 8FSK signal is greater than or equal to 4. The 4FSK signal and the 8FSK signal are distinguished according to the number of cluster centers.

Images