A signal clustering decision method and apparatus

Abstract

The application discloses a signal clustering decision method and system, and relates to the technical field of digital signal processing. The signal clustering decision method comprises the following steps: according to the signal modulation format and the nonlinear noise size, adjusting the proportion of the modulus difference value and the angle difference value of the constellation point and the current clustering center when calculating the distance, so as to calculate the distance between the constellation point and the current clustering center; according to the distance between the constellation point and the current clustering center, determining the final clustering center and the constellation points included in each final clustering center through iterative calculation; and based on the symbol decision attempt, taking the symbol relationship corresponding to the minimum bit error rate as the decision symbol to which each final clustering center and the included constellation point belong. The application can perform clustering on the constellation diagram with large linear noise and nonlinear noise.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital signal processing, and in particular to a signal clustering decision method and device. BACKGROUND

[0002] With the rise of machine learning technology, in the field of optical communication digital signal processing, people use machine learning algorithms on the basis of traditional technology to further reduce the influence of noise and seek the optimal bit error rate and decoding efficiency. The main idea is to use supervised algorithms to learn the optimal classification decision surface of the constellation diagram, or use unsupervised algorithms to learn the center of the constellation point.

[0003] Supervised algorithms require a large amount of labeled data for training, usually by sending some determined symbol data through the channel, and the receiving end models and trains the constellation diagram of these known symbols to obtain an optimal classification decision surface. However, such a scheme needs to occupy the channel to send some training data of determined symbols, and when the channel noise changes, it needs to resend the data and retrain.

[0004] Unsupervised clustering algorithms usually use K-means algorithm. Since the K-means algorithm based on the existing distance function can only cluster data distributed in a convex set, such algorithms can usually only be added after the phase recovery of traditional DSP algorithms. Assuming that the signal subset is an elliptical distribution around the center. When the linear noise and phase noise on the channel are both large, the constellation point distribution without phase recovery obviously does not meet the convex set condition, so such algorithms cannot replace the traditional phase recovery algorithm. SUMMARY

[0005] The present application provides a signal clustering decision method and device which can cluster on a constellation diagram with large linear noise and nonlinear noise.

[0006] In a first aspect, the embodiments of the present application provide a signal clustering decision method, which comprises the following steps:

[0007] According to the signal modulation format and the size of the nonlinear noise, the ratio of the modulus difference value and the angle difference value of the constellation point and the current clustering center in distance calculation is adjusted to calculate the distance between the constellation point and the current clustering center;

[0008] According to the distance between the constellation point and the current clustering center, the final clustering center and the constellation points included in each final clustering center are determined through iterative calculation;

[0009] Based on the symbol decision attempt, the symbol relationship corresponding to the minimum bit error rate is taken as the decision symbol to which each final clustering center and the included constellation point belongs.

[0010] In combination with the first aspect, in an implementation, before the step of adjusting the ratio of the modulus difference value and the angle difference value of the constellation point and the current cluster center to calculate the distance between the constellation point and the current cluster center according to the signal modulation format and the non-linear noise size, the method further comprises:

[0011] normalizing the coordinates of the constellation point after the frequency offset compensation, and normalizing the coordinates of the current cluster center according to the constellation diagram of different modulation formats.

[0012] In combination with the first aspect, in an implementation, the step of adjusting the ratio of the modulus difference value and the angle difference value of the constellation point and the current cluster center to calculate the distance between the constellation point and the current cluster center according to the signal modulation format and the non-linear noise size comprises:

[0013] calculating the distance between the constellation point and the current cluster center according to the formula:

[0014] are the polar coordinate vectors of the constellation point and the current cluster center, respectively, (a-b) 2 is the square of the modulus size difference of and , and [1-cos(θ a -θ b )] is the normalized vector angle size, and γ is a proportional coefficient set according to different signal modulation formats and noise characteristics.

[0015] In combination with the first aspect, in an implementation, the value range of γ is (0, 2).

[0016] In combination with the first aspect, in an implementation, the step of determining the final cluster center and the constellation points included in each final cluster center through iterative calculation according to the distance between the constellation point and the current cluster center comprises:

[0017] dividing each constellation point to the current cluster center closest to it according to the distance between the constellation point and the current cluster center;

[0018] After completing the nearest neighbor classification of all constellation points, recalculating the updated cluster center based on the coordinates of all constellation points of each current cluster center;

[0019] repeating the calculation of the distance between the constellation point and the current cluster center and the nearest neighbor classification of the constellation point to iteratively update the cluster center, and outputting the final cluster center and the constellation points included in each final cluster center when the proportion of constellation points that need to be re-divided is less than a threshold value or the number of iterations reaches a threshold value.

[0020] ​​​With reference to the first aspect, in an implementation form, after the nearest neighbor classification of all constellation points is completed, the updated cluster center is recalculated based on the coordinates of all constellation points of each current cluster center, including:

[0021] Converting all constellation points belonging to a current cluster center into a polar coordinate system, calculating the average of the amplitude and argument of all constellation points to determine the polar coordinates of the updated cluster center, and then converting into a plane rectangular coordinate system to determine the updated cluster center.

[0022] With reference to the first aspect, in an implementation form, the symbol decision attempt is based on the symbol relationship corresponding to the minimum bit error rate, as the decision symbol to which each final cluster center and the included constellation point belongs.

[0023] Final cluster centers with the same vector module size are grouped into a group.

[0024] Each final cluster center is assigned a symbol, and rotation attempts are performed in each group as an independent unit to obtain all combination modes of the symbol, and the corresponding bit error rate under each combination is calculated.

[0025] The minimum bit error rate under all combinations is determined, and the symbol relationship corresponding to the minimum bit error rate is taken as the decision symbol to which each final cluster center and the included constellation point belongs.

[0026] The second aspect, the embodiment of the application provides a signal clustering decision device, the signal clustering decision device includes:

[0027] The computing module adjusts the proportion of the modulus difference value and the angle difference value of the constellation point and the current cluster center when calculating the distance, according to the signal modulation format and the nonlinear noise size, to calculate the distance between the constellation point and the current cluster center.

[0028] The computing module further determines the final cluster center and the constellation points included in each final cluster center through iterative calculation according to the distance between the constellation point and the current cluster center.

[0029] The decision module attempts symbol decision based on the symbol relationship corresponding to the minimum bit error rate, as the decision symbol to which each final cluster center and the included constellation point belongs.

[0030] With reference to the second aspect, in an implementation form, further comprising:

[0031] an initialization module, configured to normalize coordinates of the constellation point after frequency offset compensation before the computing module adjusts a ratio of a modulus difference value and an angle difference value of the constellation point and the current cluster center according to a signal modulation format and a nonlinear noise size to calculate a distance between the constellation point and the current cluster center, and normalize coordinates of the current cluster center according to constellation diagrams of different modulation formats.

[0032] With reference to the second aspect, in an implementation form of the second aspect, the computing module adjusts the ratio of the modulus difference value and the angle difference value of the constellation point and the current cluster center to calculate the distance between the constellation point and the current cluster center according to the signal modulation format and the nonlinear noise size, including:

[0033] According to the formula: the distance between the constellation point and the current cluster center is calculated.

[0034] wherein, and are polar coordinate vectors of the constellation point and the current cluster center respectively, (a-b) 2 is and is a square of a modulus size difference of a and b is a normalized vector angle size, and γ is a proportional coefficient set according to different signal modulation formats and noise characteristics.

[0035] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects:

[0036] The signal clustering decision method in the present application adjusts the ratio of the modulus difference value and the angle difference value of the constellation point and the current cluster center to calculate the distance between the constellation point and the current cluster center according to the signal modulation format and the nonlinear noise size; the final cluster center and the constellation points included in each final cluster center are determined through iterative calculation according to the distance between the constellation point and the current cluster center; and the symbol relationship corresponding to the minimum bit error rate is taken as the decision symbol to which each final cluster center and the included constellation points belong based on symbol decision attempts. Thus, the K-means distance calculation method can be used to cluster on constellation diagrams with large linear noise and nonlinear noise. In addition, the algorithm can replace the phase recovery and symbol decision modules in the traditional DSP algorithm, effectively reduce the complexity of the DSP algorithm, and has strong practicability. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is a flowchart of an embodiment of the signal clustering decision method of the present application;

[0038] Figure 2 is a schematic diagram of the correspondence between the cluster center and the symbol of the present application.

[0039] Figure 3 An example diagram of initial cluster centers in an embodiment of the present application;

[0040] Figure 4 An example diagram of the relationship between noise variation and constellation variation in an embodiment of the present application;

[0041] Figure 5 An example diagram of setting a scaling factor in an embodiment of the present application;

[0042] Figure 6 A structural block diagram of an embodiment of a signal clustering decision device of the present application; DETAILED DESCRIPTION

[0043] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative labor fall within the scope of protection of the present application.

[0044] The terms “include”, “has” and “has” and any variations thereof in the specification and claims of the present application and the above-described drawings are intended to cover the non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or device. The terms “first”, “second” and “third” and the like descriptions are used to distinguish different objects, and do not represent the order or limit the types of “first”, “second” and “third”.

[0045] In the description of the embodiments of the present application, “exemplary”, “for example” or “for instance” is used to mean as an example, illustration or description. Any embodiment or design scheme described as “exemplary”, “for example” or “for instance” in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words “exemplary”, “for example” or “for instance” are intended to present the relevant concept in a specific manner.

[0046] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.

[0047] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.

[0048] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0049] In a first aspect, embodiments of this application provide a signal clustering decision method.

[0050] In one embodiment, reference is made to Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the signal clustering decision method of this application. Figure 1 As shown, signal clustering decision methods include:

[0051] S1. Based on the signal modulation format and the magnitude of nonlinear noise, adjust the ratio of the modulus difference and the angle difference between the constellation point and the current cluster center when calculating the distance, so as to calculate the distance between the constellation point and the current cluster center.

[0052] It is worth noting that the input constellation diagram data is normalized before clustering. The normalized result Xn can be represented as {(x1,y1),(x2,y2),…,(x...} n ,y n )}, where n is the number of points in the constellation diagram sample data.

[0053] In this embodiment, the coordinate range of the constellation points after frequency offset compensation needs to be normalized to [-1,1]. The original data distribution is as follows: Figure 2 As shown in figure a, the coordinate range is -128 to +128. The original coordinates are (x0, y0), then the normalized coordinates are...

[0054] Then, the cluster centers (C1, C2, ..., C6) need to be initialized according to the modulation format.k ), the number of clustering centers is related to the signal modulation format, for QPSK (Quadrature Phase Shift Keying): the number of clustering centers k = 4; for 16QAM (Quadrature Amplitude Modulation): k = 16. That is, the QPSK modulation format has 4 initial clustering centers, and the 16QAM modulation format has 16 initial clustering centers.

[0055] For the selection of clustering centers, the initial clustering centers of the Kmeans algorithm can usually be randomly selected k points, but if the selection is appropriate, the algorithm convergence speed can be accelerated, so the preferred way is to give the initial value according to the constellation characteristics of different modulation formats. The method is: taking the constellation point center without any noise on the channel as the initial center. After data normalization, the QPSK constellation Figure Four The initial values of the 4 centers and the 16QAM constellation are shown in Figure 3 , and each clustering center is given a code (binary representation) according to the modulation format.

[0056] Specifically, after data normalization, the QPSK constellation Figure 4 The initial clustering centers are respectively:

[0057] (0.5, 0.5), (-0.5, 0.5), (-0.5, -0.5), (0.5, -0.5);

[0058] The initial clustering centers of the 16QAM constellation are respectively:

[0059] (-0.75, 0.75), (-0.25, 0.75), (0.25, 0.75), (0.75, 0.75),

[0060] (-0.75, 0.25), (-0.25, 0.25), (0.25, 0.25), (0.75, 0.25),

[0061] (-0.75, -0.25), (-0.25, -0.25), (0.25, -0.25), (0.75, -0.25),

[0062] (-0.75, -0.75), (-0.25, -0.75), (0.25, -0.75), (0.75, -0.75).

[0063] It is worth noting that when the nonlinear noise and the linear noise are both large, the constellation point distribution is no longer a convex set, such as Figure 4As shown, the Kmeans algorithm cannot complete constellation point clustering when both linear and nonlinear noise are large.

[0064] To address this issue, this invention proposes a distance calculation function that improves the distance calculation of K-means, enabling it to solve the aforementioned problem.

[0065] Specifically, in this embodiment, according to the formula: Calculate the distance between constellation points and the current cluster center;

[0066] in, and Let (ab) be the polar coordinate vectors of the constellation point and the current cluster center, respectively. 2 For a jθ a and The square of the difference in magnitude, [1-cos(θ) a -θ b )] represents the normalized vector angle, and γ is a scaling factor set according to different signal modulation formats and noise characteristics, with a value range of (0, 2).

[0067] The scaling factor γ is used to adjust the ratio between the magnitude difference and the angle difference in the distance calculation formula. Within the range of (0, 2), as γ increases, the distance calculation result will be more affected by the magnitude difference, thus allowing for flexible adjustment based on the signal modulation format and the magnitude of nonlinear noise.

[0068] The following is based on Figure 5 This explains the effect of γ on the 16QAM constellation diagram under different values. When γ = 0, the formula can be expressed as follows: At this point, the distance calculation result is mainly determined by the angle (θ) between the constellation points and the cluster center. a -θ b The decision is made to divide the area of ​​each cluster center as follows: Figure 5 As shown in Figure a, the division range of cluster centers is determined by the angle between the constellation points and the cluster centers. As γ increases, the proportion of the magnitude difference gradually increases. When γ = 1, the formula can be simplified to... The formula here represents the Euclidean distance between constellation points and cluster centers, which is the traditional distance calculation method. The corresponding cluster center partitioning results are as follows: Figure 5 As shown in c, constellation points will be assigned to the nearest cluster centers. As γ continues to increase, the weight of the magnitude difference further increases, as... Figure 5 As shown in d, the partition range of each cluster center is mainly determined by the modulus difference between the constellation points and the cluster centers. When γ = 2, the formula can be expressed as follows: like Figure 5As shown in FIG. 6, at this time, the division of the constellation points is completely determined by the modulus of the constellation points.

[0069] Therefore, in 16QAM, γ can be set according to the size of the nonlinear noise, when the nonlinear noise is small, γ can be set as 1, when the nonlinear noise is large, 1 < γ < 2 can be set, so as to tolerate larger phase noise and achieve lower bit error rate. In the standard QPSK constellation, the modulus of the four cluster centers is equal, only the amplitude angle of the constellation points is concerned, so γ can be set as 0. In the embodiment, it can be known from the constellation that the signal is subjected to large nonlinear noise, therefore, γ is set as 1.5.

[0070] S2, according to the distance between the constellation points and the current cluster center, the final cluster center and the constellation points included in each final cluster center are determined through iterative calculation;

[0071] It is worth noting that, before the cluster center is stable, a part of data points is re-divided into the cluster center in each iteration, the stop condition is defined as that the proportion of the re-divided constellation points to the total points after one iteration is less than a threshold t, or the number of iterations reaches a threshold. t is set as 0.01, which means that the iteration is stopped when less than 1% of the points are re-divided, at this time, it is considered that the cluster center is relatively stable. In another way, the number of iteration threshold can be set as 100.

[0072] In the specific implementation, step S2 includes:

[0073] S21, for all the constellation points in the constellation, the distance to the k cluster centers is calculated by using the distance calculation formula, the constellation points are classified into the cluster center with the minimum distance, until all the constellation points are classified, and the proportion of the re-classified constellation points is counted (the proportion of the first re-classification is set as 1).

[0074] S22, according to the classification result, the k cluster centers are re-calculated. The calculation method is that all the constellation points belonging to a cluster center are converted to the polar coordinate system, the average of the amplitude and the amplitude angle of all the constellation points is calculated, that is, the polar coordinate representation of the cluster center, and then converted to the plane rectangular coordinate system.

[0075] S23, the above steps S21 and S22 are repeated until the proportion of the re-classified constellation points in a certain iteration is less than t or the number of iterations exceeds the threshold, the iteration is ended, and the k cluster centers and the constellation points included in each cluster center are output.

[0076] For example, after 73 iterations, the proportion of the re-classified points is 0.68%, which is less than the set threshold of 1%, and the iteration is stopped. The coordinate points of the cluster centers are respectively:

[0077] (0.19, 0.28), (0.29, -0.15), (-0.17, -0.27), (-0.22, 0.16);

[0078] (0.13, 0.72), (0.57, 0.44), (0.73, -0.09), (0.41, -0.49);

[0079] (-0.16, -0.68), (-0.57, -0.36), (-0.62, 0.11), (-0.34, 0.57);

[0080] (0.46, 0.85), (0.77, -0.55), (-0.42, -0.81), (-0.73, 0.58).

[0081] The constellation point clustering results are shown in FIG. 2. Figure 2

[0082] S3, based on symbol decision attempt, the symbol relationship corresponding to the minimum bit error rate as the final clustering center and the included constellation point belongs to the decision symbol.

[0083] After K-means clustering in step S2, the clustering center and the constellation points contained in each center are obtained, but at this time, the symbol represented by each clustering center is not determined. For symbol decision of the clustering center, the decision method is to first perform symbol decision attempt, and then take the symbol decision with the minimum bit error rate BER as the final symbol decision result. The following takes 16QAM as an example to illustrate. Figure 2

[0084] After clustering, A-P 16 groups of data are formed, which are to be corresponded with 0000-1111 16 symbols. The vector modules of the four clustering centers A, B, C and D are equal, which are respectively corresponded with 0000, 1000, 1100 and 0100, and there are four possible corresponding relationships:

[0085] A-0000, B-1000, C-1100, D-0100;

[0086] B-0000, C-1000, D-1100, A-0100;

[0087] C-0000, D-1000, A-1100, B-0100;

[0088] D-0000, A-1000, B-1100, C-0100.

[0089] ​​Similarly, the clustering centers of 0010, 0001, 1001, 1010, 1110, 1101, 0101, 0110, and E, F, G, H, I, J, K, L, which have equal modulus, have 8 possible corresponding relations; 0011, 1011, 1111, 0111, and M, N, O, P have 4 possible corresponding relations.

[0090] Therefore, a total of 4*8*4=128 attempts are required. The binary stream decoded by the 128 attempts is sent to the subsequent decoding module of the DSP, and the bit error rate output by the decoding module is obtained. By comparing the BER sizes of the 128 decodings, the smallest BER is 1.6e-4, and the symbol relationship corresponding to this BER is the optimal symbol decision result of the clustering center. Finally, the final clustering center and the decision symbol to which each constellation point belongs can be output, and the final symbol decision result is as shown in FIG. 8c. Figure 2 c, the clustering of the constellation points and the symbol decision are completed.

[0091] In summary, the signal clustering decision method in the present application adjusts the proportion of the modulus difference value and the angle difference value of the constellation point and the current clustering center when calculating the distance, according to the signal modulation format and the size of the nonlinear noise, to calculate the distance between the constellation point and the current clustering center; determines the final clustering center and the constellation points included in each final clustering center through iterative calculation according to the distance between the constellation point and the current clustering center; and takes the symbol relationship corresponding to the minimum bit error rate as the decision symbol to which each final clustering center and the included constellation point belongs based on the symbol decision attempt. Therefore, the K-means distance calculation method can be used to cluster on the constellation diagram in which the linear noise and the nonlinear noise are both large. In addition, the algorithm can replace the phase recovery and symbol decision module in the traditional DSP algorithm, effectively reduce the complexity of the DSP algorithm, and has strong practicability.

[0092] In a second aspect, the embodiments of the present application also provide a signal clustering decision device.

[0093] In an embodiment, refer to FIG. 8a, Figure 6 , Figure 6 FIG. 8b is a functional module schematic diagram of an embodiment of the signal clustering decision device of the present application. As shown in FIG. 8b, the signal clustering decision device comprises: Figure 6

[0094] a calculation module, which adjusts the proportion of the modulus difference value and the angle difference value of the constellation point and the current clustering center when calculating the distance, according to the signal modulation format and the size of the nonlinear noise, to calculate the distance between the constellation point and the current clustering center;

[0095] The calculation module also determines the final clustering center and the constellation points included in each final clustering center through iterative calculation according to the distance between the constellation point and the current clustering center;

[0096] ​a decision module configured to determine, based on a symbol decision attempt, a symbol relationship corresponding to a minimum bit error rate as a decision symbol to which each final cluster center and included constellation point belongs.

[0097] Further, in an embodiment, the method further comprises:

[0098] an initialization module configured to normalize coordinates of the constellation point after frequency offset compensation before the calculation module adjusts a ratio of a modulus difference value and an angle difference value of the constellation point and the current cluster center in distance calculation according to a signal modulation format and a non-linear noise size to calculate a distance between the constellation point and the current cluster center, and normalize coordinates of the current cluster center according to constellation diagrams of different modulation formats.

[0099] Further, in an embodiment, the calculation module adjusts the ratio of the modulus difference value and the angle difference value of the constellation point and the current cluster center in distance calculation according to the signal modulation format and the non-linear noise size to calculate the distance between the constellation point and the current cluster center, comprising:

[0100] According to the formula: the distance between the constellation point and the current cluster center is calculated.

[0101] wherein, and are polar coordinate vectors of the constellation point and the current cluster center, respectively, (a-b) 2 is and is a square of a modulus size difference of a and b is a normalized vector angle size, and γ is a proportional coefficient set according to different signal modulation formats and noise characteristics.

[0102] Further, in an embodiment, the value range of the γ is (0, 2).

[0103] Further, in an embodiment, the calculation module determines the final cluster center and the constellation points included in each final cluster center through iterative calculation according to the distance between the constellation point and the current cluster center, comprising:

[0104] each constellation point is divided into the current cluster center closest to the constellation point according to the distance between the constellation point and the current cluster center;

[0105] after the nearest neighbor classification of all constellation points is completed, the updated cluster center is recalculated based on coordinates of all constellation points of each current cluster center;

[0106] the cluster center is iteratively updated until the proportion of constellation points that need to be re-divided is less than a threshold value or the number of iterations reaches a threshold value, and the final cluster center and the constellation points included in each final cluster center are output.

[0107] Further, in an embodiment, after the computing module completes the nearest neighbor classification for all constellation points, the updated cluster centers are recalculated based on the coordinates of all constellation points of each current cluster center, including:

[0108] Converting all constellation points belonging to a current cluster center to the polar coordinate system, calculating the average of the amplitude and argument of all constellation points to determine the polar coordinates of the updated cluster center, and then converting to the plane rectangular coordinate system to determine the updated cluster center.

[0109] Further, in an embodiment, the decision module based on the symbol decision attempt takes the symbol relationship corresponding to the minimum bit error rate as the decision symbol to which each final cluster center and the included constellation point belongs, including:

[0110] Final cluster centers of the same vector module size are divided into a group;

[0111] A symbol is assigned to each final cluster center, and rotation attempts are made for each group as an independent unit to obtain all combination modes of the symbol and calculate the corresponding bit error rate under each combination;

[0112] The minimum bit error rate under all combinations is determined, and the symbol relationship corresponding to the minimum bit error rate is taken as the decision symbol to which each final cluster center and the included constellation point belongs.

[0113] The above is only a preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent flow transformation using the content of the specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A signal clustering decision method, characterized in that, The signal clustering decision method includes the following steps: Based on the signal modulation format and the magnitude of nonlinear noise, adjust the ratio of the modulus difference and the angle difference between the constellation point and the current cluster center when calculating the distance; Based on the distance between constellation points and the current cluster center, the final cluster center and the constellation points included in each final cluster center are determined through iterative calculation; Based on symbol decision attempts, the symbol relationship corresponding to the minimum bit error rate is used as the decision symbol to which each final cluster center and the constellation points it includes belong; Before adjusting the ratio of the modulus difference and the angle difference between the constellation point and the current cluster center according to the signal modulation format and the magnitude of nonlinear noise to calculate the distance between the constellation point and the current cluster center, the method further includes: The coordinates of the constellation points after frequency offset compensation are normalized, and the coordinates of the current cluster center are normalized according to the constellation diagram of different modulation formats. The step of adjusting the ratio of the modulus difference and the angle difference between the constellation point and the current cluster center during distance calculation based on the signal modulation format and the magnitude of nonlinear noise, to calculate the distance between the constellation point and the current cluster center, includes: According to the formula: Calculate the distance between constellation points and the current cluster center; in, and These are the polar coordinate vectors of the constellation point and the current cluster center, respectively. for and The square of the difference in magnitude, The angle between the normalized vectors. The scaling factor is set according to different signal modulation formats and noise characteristics. The range of values ​​for is (0, 2).

2. The signal clustering decision method as described in claim 1, characterized in that, The step of determining the final cluster center and the constellation points included in each final cluster center through iterative calculation based on the distance between the constellation points and the current cluster center includes: Based on the distance between each constellation point and the current cluster center, assign each constellation point to the nearest current cluster center; After completing the nearest neighbor classification for all constellation points, the updated cluster centers are recalculated based on the coordinates of all constellation points for each current cluster center. Repeatedly calculate the distance between constellation points and the current cluster center, and repeatedly perform nearest neighbor classification on constellation points to iteratively update the cluster centers until the proportion of constellation points that need to be re-divided is less than the threshold or the number of iterations reaches the threshold. Then, output the final cluster center and the constellation points included in each final cluster center.

3. The signal clustering decision method as described in claim 2, characterized in that, After completing the nearest neighbor classification for all constellation points, the updated cluster centers are recalculated based on the coordinates of all constellation points for each current cluster center, including: Transform all constellation points belonging to a current cluster center to a polar coordinate system, calculate the average of the amplitude and argument of all constellation points to determine the polar coordinates of the updated cluster center, and then transform them to a Cartesian coordinate system to determine the updated cluster center.

4. The signal clustering decision method as described in claim 1, characterized in that, The symbol-based decision attempt, using the symbol relation corresponding to the minimum bit error rate as the decision symbol to which each final cluster center and its included constellation points belong, includes: Group the final cluster centers with the same vector magnitude into one group; Assign a symbol to each final cluster center, perform rotation attempts on each group as an independent unit to obtain all possible combinations of symbols, and calculate the corresponding bit error rate for each combination. Determine the minimum bit error rate for all combinations, and use the sign relationship corresponding to the minimum bit error rate as the decision sign to which each final cluster center and its included constellation points belong.

5. A signal clustering decision device, characterized in that, The signal clustering decision device includes: The calculation module adjusts the ratio of the modulus difference and the angle difference between the constellation point and the current cluster center during distance calculation based on the signal modulation format and the magnitude of nonlinear noise. The calculation module also determines the final cluster center and the constellation points included in each final cluster center through iterative calculation based on the distance between the constellation points and the current cluster center. The decision module is based on symbol decision attempts, using the symbol relationship corresponding to the minimum bit error rate as the decision symbol to which each final cluster center and the constellation points it includes belong. An initialization module is used to normalize the coordinates of the constellation points after frequency offset compensation and normalize the coordinates of the current cluster center according to the constellation diagram of different modulation formats before the calculation module adjusts the ratio of the modulus difference and the included angle difference between the constellation points and the current cluster center during distance calculation based on the signal modulation format and the magnitude of nonlinear noise. The calculation module adjusts the ratio of the modulus difference and the angle difference between the constellation point and the current cluster center during distance calculation based on the signal modulation format and the magnitude of nonlinear noise, in order to calculate the distance between the constellation point and the current cluster center, including: According to the formula: Calculate the distance between constellation points and the current cluster center; in, and These are the polar coordinate vectors of the constellation point and the current cluster center, respectively. for and The square of the difference in magnitude, The angle between the normalized vectors. The scaling factor is set according to different signal modulation formats and noise characteristics. The range of values ​​for is (0, 2).

Images