Digital modulation mode identification method based on constellation diagram characteristics

Abstract

The invention discloses a digital modulation mode identification method based on constellation diagram characteristics. Firstly, the method uses a clustering algorithm to obtain amplitude correlation characteristics of constellation points, does not depend on prior information of signals, uses multiplication operation to replace exponential operation, and reduces calculation complexity. And secondly, the constellation points of the outermost circle are extracted for subsequent processing, so that the robustness of the clustering algorithm is improved, and meanwhile, the calculation complexity is further reduced. As the high-power spectrum is more robust for carrier frequency offset and non-ideal characteristics of a channel, the features are extracted from the high-power spectrum of the constellation points of the outermost ring. Therefore, the technical scheme provided by the invention improves the robustness and accuracy of the digital modulation mode identification method, reduces the calculation complexity, and has good practical value.

Description

Technical Field

[0001] The present invention relates to the field of display technology, and in particular to a digital modulation mode recognition method based on constellation diagram features. Background Art

[0002] Automatic modulation recognition (AMR) is an important technology in the field of non-cooperative communications. It aims to determine the modulation mode of a signal in the presence of noise interference and other unknown signal parameters, and provide prior information for subsequent signal processing. This plays a vital role in military, national security and civil fields. Therefore, AMR methods with high accuracy and strong robustness are of great significance for practical applications.

[0003] At present, AMR research is mainly divided into two categories: likelihood-based and feature-based. Among them, likelihood-based AMR has better recognition performance, but defects such as high computational complexity and large delay limit the application of this method in practical engineering. Feature-based methods are mainly divided into implicit feature extraction and explicit feature extraction. Implicit feature extraction is usually implemented based on neural networks, but due to insufficient generalization ability, most existing neural network methods are not suitable for non-cooperative communication scenarios. Explicit features are usually extracted from the time domain and frequency domain, among which the most representative is the high-order cumulant feature of the signal, but this feature is easily affected by carrier frequency deviation and channel non-ideal characteristics. Compared with high-order cumulant features, spectral features are more robust to carrier frequency deviation, but still face challenges such as channel sensitivity and low recognition accuracy at low signal-to-noise ratio. In order to ensure the robustness of the recognition method under changing channel conditions, the existing technology uses clustering to obtain amplitude features. However, the k-Medoids method in the existing technology requires prior information on the number of clusters, and the recognition accuracy is low. In addition, the subtractive clustering method in the existing technology requires exponential operations, which has high computational complexity. Summary of the invention

[0004] In order to solve the limitations and defects of the prior art, the present invention provides a digital modulation mode identification method based on constellation diagram features, comprising:

[0005] Step S1: Calculate the amplitude of each sampling point of the received signal y = [y(1), y(2), ..., y(n)] to obtain the amplitude vector x = [x1, x2, ..., x N ], where x i represents the amplitude of the i-th sampling point y(i), and N represents the total number of sampling points of the received signal y;

[0006] Step S2: Get element x i The difference index with other elements is expressed as follows:

[0007]

[0008] Among them, r a is a constant;

[0009] Step S3: In the kth (k=1, 2, ...) iteration, obtain the vector d=[D1, D2, ..., D N The maximum value in Select Corresponding elements As the cluster center, update the difference index of each element, the expression is as follows:

[0010]

[0011] Among them, r b is a constant, Represents the cluster center The difference index, represents D in the kth (k=1,2,...) iteration process i , Then it means D of the k+1th iteration i ;

[0012] Step S4: Before the nth iteration, for the vector d=[D1,D2,...,D N The maximum value in Compare and the size of the preset threshold δ;

[0013] Step S5: If the comparison result is is greater than the preset threshold δ, the nth round of iteration is performed; if the comparison result is When the value is less than or equal to the preset threshold δ, the iteration stops and n-1 cluster centers are obtained.

[0014] Step S6, classifying all elements in the amplitude vector x according to the distance from the element to the cluster center;

[0015] Step S7: Divide the amplitude vector x into a preset number of classes to obtain the received signal amplitude related feature N c ;

[0016] Step S8: Obtain the outermost constellation point feature I m ;

[0017] Step S9: according to the received signal amplitude related feature N c and the outermost constellation point feature I m Identify digital modulation methods;

[0018] Step S10: output the recognition result.

[0019] Optionally, the step S8 includes:

[0020] Get the cluster center with the largest amplitude value from all cluster centers Determine the attribution The sampling points are the outermost constellation points;

[0021] Assume that there are L outer circle constellation points in total, obtain the outer circle constellation point z = [z(1), z(2), ..., z(L)], and calculate the Fourier transform result S of the outer circle constellation point z to the power of m m (k), the expression is as follows:

[0022]

[0023] Get binary variable I m , used to record S m The existence of the spectral lines in (k) is expressed as follows:

[0024]

[0025] Among them, the binary variable I m The outermost constellation point feature I m .

[0026] Optionally, step S6 includes:

[0027] Traverse all elements in the magnitude vector x and calculate the element x i The distance from each cluster center is obtained to obtain the cluster center corresponding to the minimum distance, and the element x i Divide into the class where the corresponding cluster center is located.

[0028] The present invention has the following beneficial effects:

[0029] The present invention provides a method for identifying a digital modulation mode based on constellation diagram features. First, the present invention uses a clustering algorithm to obtain amplitude-related features of constellation points, does not rely on prior information of the signal, and uses multiplication operations instead of exponential operations, thereby reducing the computational complexity. Secondly, the present invention extracts the outermost circle constellation points for subsequent processing, which improves the robustness of the clustering algorithm while further reducing the computational complexity. Since high-order spectra are more robust to carrier frequency deviation and channel non-ideal characteristics, the present invention extracts features from the high-order spectra of the outermost circle constellation points. Therefore, the technical solution provided by the present invention improves the robustness and accuracy of the digital modulation mode identification method, reduces the computational complexity, and has good practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1This is a flow chart of a method for identifying a digital modulation mode based on constellation diagram features provided in Embodiment 1 of the present invention.

[0031] Figure 2 This is a 16QAM signal constellation diagram provided in Example 1 of the present invention.

[0032] Figure 3 This is a 16APSK signal constellation diagram provided in Example 1 of the present invention.

[0033] Figure 4 This is a schematic diagram of the situation where the spectral line does not exist provided in the first embodiment of the present invention.

[0034] Figure 5 This is a schematic diagram of the existence of spectral lines provided in Example 1 of the present invention.

[0035] Figure 6 This is a modulation mode identification flow chart provided in Example 1 of the present invention.

[0036] Figure 7 A schematic diagram of the modulation mode recognition accuracy results in a multipath Rayleigh channel provided in Example 1 of the present invention.

[0037] Figure 8 A schematic diagram of the modulation mode recognition accuracy results under a nonlinear channel provided in Example 1 of the present invention.

[0038] Figure 9a A schematic diagram of the modulation mode recognition accuracy results in a multipath Rayleigh channel environment with frequency offset provided in the first embodiment of the present invention.

[0039] Figure 9b A schematic diagram of the modulation mode recognition accuracy results in a multipath Rayleigh channel environment without frequency offset provided in the first embodiment of the present invention.

[0040] Fig.10a A schematic diagram of the modulation mode recognition accuracy results in a nonlinear channel environment with frequency deviation provided in the first embodiment of the present invention.

[0041] Fig.10b A schematic diagram of the modulation mode recognition accuracy results in a nonlinear channel environment with no frequency offset provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to enable those skilled in the art to better understand the technical solution of the present invention, the digital modulation mode identification method based on constellation diagram features provided by the present invention is described in detail below with reference to the accompanying drawings.

[0043] Embodiment 1

[0044] 1. General Introduction

[0045] In order to improve the robustness and accuracy of the modulation mode recognition method under harsh environments, such as carrier frequency deviation, time-varying channels, etc., and reduce the computational complexity, this embodiment proposes a modulation mode recognition method based on constellation diagram features. First, this embodiment designs an adaptive low-complexity clustering algorithm to obtain the amplitude characteristics of the constellation points. This method does not rely on the prior information of the signal, and uses multiplication operations to replace the exponential operations in the prior art, thereby reducing the computational complexity. Secondly, in order to improve the robustness of the algorithm and further reduce the computational complexity, this embodiment extracts the outermost circle points of the constellation diagram for subsequent processing. Furthermore, considering that the high-order spectrum is more robust to carrier frequency deviation and channel non-ideal characteristics, this embodiment extracts features from the high-order spectrum of the outermost circle points. Finally, the above features are used to realize modulation classification. Compared with the prior art, the digital modulation mode recognition method provided by this embodiment has lower complexity, higher robustness to carrier frequency deviation and channel non-ideal characteristics, and has good practical value.

[0046] 2. Modulation mode identification method

[0047] This embodiment proposes a modulation mode identification method based on the characteristics of the constellation diagram of the received signal. Specifically, this embodiment considers seven commonly used digital modulation methods: QPSK, 8PSK, 16QAM, 64QAM, 16APSK, 32APSK, and 64APSK. Among them, 16APSK includes two constellation structures: 16APSK (4+12) and 16APSK (8+8), and 64APSK includes two constellation structures: 64APSK (16+16+16+16) and 64APSK (8+16+20+20). The core idea of ​​the technical solution provided in this embodiment is to extract the amplitude characteristics of the constellation points of the received signal and the characteristics of the outermost points of the constellation diagram, and then identify various modulation modes according to the difference in characteristics. Figure 1 This is a flow chart of a method for identifying a digital modulation mode based on constellation diagram features provided in Embodiment 1 of the present invention.

[0048] Wherein, y represents the received signal, x represents the amplitude value of the received signal, and z represents the outermost point of the received signal on the constellation diagram. Figure 1 The blue box in the middle is the clustering related step, which is used to obtain the amplitude related feature N c ; The orange box is the high-order spectrum correlation step, which is used to obtain the outermost constellation point feature I m Next, this section describes the feature extraction method and modulation recognition process in detail.

[0049] 2.1 Clustering to obtain amplitude-related features N c

[0050] Figure 2 This is a 16QAM signal constellation diagram provided in Example 1 of the present invention. Figure 3This is a 16APSK signal constellation diagram provided in Example 1 of the present invention. Figure 2 and Figure 3 The 16QAM and 16APSK (4+12) constellation diagrams shown in FIG. 1 are taken as examples, where the red dots represent the outermost constellation points and the blue dots represent the remaining constellation points. In order to extract the amplitude-related features in the constellation diagram, this embodiment first calculates the amplitude of each sampling point of the received signal y=[y(1), y(2), ..., y(n)] to obtain the amplitude vector x=[x1, x2, ..., x N ], where x i represents the amplitude of the i-th sampling point y(i), and N represents the total number of sampling points of the received signal y. Then, this embodiment designs a clustering algorithm, which divides the elements in the vector x into several categories according to the difference in the amplitude of each sampling point of the signal. The number of obtained categories N c Represents the number of amplitudes present in the ideal signal constellation diagram. The process of the clustering algorithm is as follows:

[0051] (1) Define an element x i Difference indicators from other elements:

[0052]

[0053] Among them, r a is a constant. The difference index is negatively correlated with the difference between elements, that is, the larger the difference index, the smaller the difference between elements. Specifically, if the element x j With x i The difference of |x i -x j The smaller |, the D ij The larger the value, the larger the i When there are more elements with smaller difference, D i In addition, with x i The pair of elements with larger difference D i The impact is small, in other words, D i The value of is not affected by outliers in vector x.

[0054] (2) In the kth (k = 1, 2, ...) iteration, find the vector d = [D1, D2, ..., D N The maximum value in and select The corresponding element As the cluster center, the difference index of each element is updated as follows:

[0055]

[0056] Among them, r b is another constant, Represents the cluster center The difference index, represents D in the kth iteration process i , Then it means D of the k+1th iteration i . After the above calculation, close to the cluster center The element difference index is significantly reduced, far away from the cluster center There was no significant change in the element difference index.

[0057] (3) Before the nth iteration, for the maximum value in vector d If satisfied (δ is the preset threshold), then the nth round of iteration is performed, otherwise the iteration is stopped and n-1 cluster centers are obtained. Taking 16APSK as an example, its constellation diagram is composed of two circles of points. Theoretically, after clustering, two cluster centers can be obtained, corresponding to the two amplitudes of the 16APSK signal.

[0058] (4) Classify all elements in vector x according to their distance to the cluster center. First, traverse all elements in vector x. For a specific element x i , calculate x i The distance between each cluster center, find the cluster center corresponding to the minimum distance, and convert x i Divide into the class where the cluster center is located.

[0059] Through the above clustering algorithm, the received signal amplitude x can be divided into several categories, and the characteristics N of the received signal amplitude can be obtained. c .

[0060] 2.2 High-order spectrum to obtain the characteristics of the outermost point I m

[0061] First, according to the clustering results, select the cluster center with the largest amplitude value from all cluster centers Belong to The sampling point is the outermost point. Assuming there are L outer points, the outer points are recorded as z = [z(1), z(2), ..., z(L)]. Next, calculate the Fourier transform result S of z raised to the power of m. m (k):

[0062]

[0063] Finally, define a binary variable I m , used to record S m Existence of spectral lines in (k):

[0064]

[0065] by Figure 3 Taking the 16APSK signal in as an example, when m=8, there is no spectrum line, that is, I8=0, such as Figure 4 As shown; when m = 12, there is a spectral line, namely I 12 =1, such as Figure 5 shown.

[0066] 2.3 Modulation Mode Identification

[0067] Figure 6 The modulation mode identification flow chart provided in the first embodiment of the present invention. First, the QAM signal is distinguished based on I4, and then the number of cluster centers N is c Perform QAM intra-class identification and finally construct the feature vector k=[I8,I 12 ,I 16 ,I 20 ], and combined with N c Complete further intra-class identification.

[0068] The present embodiment provides a method for identifying digital modulation modes based on constellation diagram features. First, the present embodiment uses a clustering algorithm to obtain amplitude-related features of constellation points, does not rely on prior information of the signal, and uses multiplication operations instead of exponential operations, thereby reducing computational complexity. Secondly, the present embodiment extracts the outermost constellation points for subsequent processing, thereby improving the robustness of the clustering algorithm and further reducing computational complexity. Since high-order spectra are more robust to carrier frequency deviation and channel non-ideal characteristics, the present embodiment extracts features from the high-order spectra of the outermost constellation points.

[0069] 3. Simulation Results

[0070] This section gives the simulation results of the proposed method. Specifically, two channels, a multipath channel and a nonlinear channel, are set in the experiment. Gaussian white noise is added to the transmitted signal after passing through the channel, and then the recognition algorithm proposed in this embodiment is used to obtain the modulation mode recognition result. The setting of the experimental parameters is shown in Table 1.

[0071] Table 1 Experimental parameter settings

[0072] parameter Value Symbol rate 100kHz Carrier frequency deviation (relative to symbol rate) <![CDATA[10 -4 ]]> Multipath Rayleigh channel path number 6 Delay spread of multipath Rayleigh channel 1μs Doppler spread (relative to symbol rate) <![CDATA[3×10 -5 ]]> Nonlinear channel order 3 <![CDATA[Signal-to-noise ratio (E b / N0)]]> 0~15dB

[0073] Figure 7 A schematic diagram of the modulation mode recognition accuracy results in a multipath Rayleigh channel provided in Example 1 of the present invention. Figure 8 Schematic diagram of the modulation recognition accuracy under nonlinear channels provided by the first embodiment of the present invention. In the two channel environments, the relationship between the modulation recognition accuracy and the signal-to-noise ratio is as follows: Figure 7 and Figure 8As shown in the figure, in a multipath Rayleigh channel, the recognition accuracy of QPSK, 8PSK, 16QAM, 64QAM, 16APSK (8+8), 16APSK (4+12), 32APSK, 64APSK (16+16+16+16), and 64APSK (8+16+20+20) is more than 99% when the signal-to-noise ratio is 2dB, 5dB, 10dB, 14dB, 3dB, 6dB, 9dB, 9dB, and 10dB respectively; in a nonlinear channel, QPSK, When the signal-to-noise ratios of 8PSK, 16QAM, 64QAM, 16APSK(8+8), 16APSK(4+12), 32APSK, 64APSK(16+16+16+16), and 64APSK(8+16+20+20) are 1dB, 4dB, 9dB, 13dB, 3dB, 6dB, 8dB, 8dB, and 8dB respectively, the recognition accuracy rate reaches over 99%. It can be seen that the modulation mode recognition method proposed in this embodiment is robust to non-ideal channels.

[0074] Figure 9a A schematic diagram of the modulation mode recognition accuracy results in a multipath Rayleigh channel environment with frequency offset provided in the first embodiment of the present invention. Figure 9b A schematic diagram of the modulation mode recognition accuracy results in a multipath Rayleigh channel environment without frequency offset provided in the first embodiment of the present invention. Fig.10a A schematic diagram of the modulation mode recognition accuracy results in a nonlinear channel environment with frequency deviation provided in the first embodiment of the present invention. Fig.10b A schematic diagram of the modulation mode recognition accuracy results in a nonlinear channel environment with no frequency offset provided in the first embodiment of the present invention. Figure 9a-9b and Figure 10a-10b The recognition confusion matrix of the method proposed in this embodiment in two cases, with and without carrier frequency offset, is given. It can be seen from the figure that the carrier frequency offset does not affect the recognition accuracy, that is, the method proposed in this embodiment is robust to the carrier frequency offset.

[0075] In summary, the algorithm proposed in this embodiment realizes automatic recognition of common digital modulation modes, has good performance in complex channel environments, and can resist carrier frequency deviation to a certain extent, which has practical engineering significance.

[0076] 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 method for identifying digital modulation modes based on constellation features, characterized in that: include: Step S1: Calculate the amplitude of each sampling point of the received signal y = [y(1), y(2), ..., y(n)] to obtain the amplitude vector x = [x1, x2, ..., x N ], where x i represents the amplitude of the i-th sampling point y(i), and N represents the total number of sampling points of the received signal y; Step S2: Get element x i The difference index with other elements is expressed as follows: Among them, r a is a constant; Step S3: In the kth (k=1, 2, ...) iteration, obtain the vector d=[D1, D2, ..., D N The maximum value in Select Corresponding elements As the cluster center, update the difference index of each element, the expression is as follows: Among them, r b is a constant, Represents the cluster center The difference index, represents D in the kth (k=1,2,...) iteration process i , Then it means D of the k+1th iteration i ; Step S4: Before the nth iteration, for the vector d=[D1, D2, ..., D N The maximum value in Compare and the size of the preset threshold δ; Step S5: If the comparison result is is greater than the preset threshold δ, the nth round of iteration is performed; if the comparison result is When the value is less than or equal to the preset threshold δ, the iteration stops and n-1 cluster centers are obtained. Step S6, classifying all elements in the amplitude vector x according to the distance from the element to the cluster center; Step S7: Divide the amplitude vector x into a preset number of classes to obtain the received signal amplitude related feature N c ; Step S8: Obtain the outermost constellation point feature I m ; Step S9: according to the received signal amplitude related feature N c and the outermost constellation point feature I m Identify digital modulation methods; Step S10: output the recognition result.

2. The method for identifying digital modulation modes based on constellation features according to claim 1, characterized in that: The step S8 comprises: Get the cluster center with the largest amplitude value from all cluster centers Determine the attribution The sampling points are the outermost constellation points; Assume that there are L outer circle constellation points in total, obtain the outer circle constellation point z = [z(1), z(2), ..., z(L)], and calculate the Fourier transform result S of the outer circle constellation point z to the power of m m (k), the expression is as follows: Get binary variable I m , used to record S m The existence of the spectral lines in (k) is expressed as follows: Among them, the binary variable I m The outermost constellation point feature I m .

3. The method for identifying digital modulation modes based on constellation features according to claim 2, characterized in that: The step S6 comprises: Traverse all elements in the magnitude vector x and calculate the element x i The distance from each cluster center is obtained to obtain the cluster center corresponding to the minimum distance, and the element x i Divide into the class where the corresponding cluster center is located.

Images