CN111083079A - Constellation Diagram Based Quadrature Modulation Format Identification Method - Google Patents

Abstract

The invention discloses a method for identifying a quadrature modulation format based on a constellation diagram. After equalizing a received signal, a small amount of equalized and normalized symbols are taken as the input of the identification method to match a normalized M quadrature amplitude modulation QAM signal. The template is clustered, and two-dimensional Gaussian model analysis is performed on each clustered constellation to obtain the probability value of each symbol in the cluster, and the fit degree of the possible MQAM normalized signal templates is calculated separately. Finally, according to the preset The fit range of the template determines the modulation format order of the signal to be identified. The identification method proposed by the invention uses the cluster analysis method in machine learning to establish a two-dimensional Gaussian model for analysis. The method has simple principle, low complexity and wide application range.

Description

Constellation Diagram Based Quadrature Modulation Format Identification Method

(1) Technical field

The present invention relates to the technical field of communications, in particular to a method for identifying an orthogonal modulation format based on a constellation diagram.

(2) Background technology

With the development of communication technology, the system and modulation patterns of communication signals have become more complex and diverse, and the automatic modulation format identification method has been paid more and more attention by people. Modulation format identification has been developed for decades, but there is not a unified and complete theoretical system to solve this problem, and it is difficult to find a unified method to identify all modulation formats. Because modulation format recognition is not an isolated problem, it integrates various aspects such as detection, estimation, feature selection and classification and recognition.

The identification methods of orthogonal modulation formats that have been proposed so far mainly include: (1) Using the high-order cumulant (High-Order Component HOC) of the modulated signal as a feature, using a supervised machine learning algorithm as a classifier to classify the modulated signal The modulation format of [M.W.Aslam, Z.Zhu and A.K.Nandi, “Automatic modulationclassification using combination of genetic programming and KNN,” IEEE T WIRELCOMMUN, vol.11, no.8, pp.2742-2750, AUG.2012]. (2) By calculating the power distribution characteristic value of the power normalized modulation signal, and comparing the power distribution characteristic value with a plurality of preset thresholds, the modulation format of the quadrature modulation signal is determined [CN 104756456B]. (3) Use unsupervised machine learning algorithms, such as K-means clustering algorithm, to perform cluster analysis on the constellation diagram, and estimate the order of the modulation format according to the obtained number of centroids [G.Jajoo, Y.Kumar, S.K.Yadav, B . Adhikari, and A. Kumar, "Blind signal modulation recognition throughclustering analysis of constellation signature," EXPERT SYST APPL, vol. 90, pp. 13-22, Aug. 2017].

Among the above methods, the use of supervised learning algorithm requires a large training overhead; the use of K-means to calculate the number of centroids requires multiple iterations, and the selection of the initial centroids will affect the K-means clustering results; the method of calculating the eigenvalues of power distribution is used for optical information. Noise ratio requirements are high, and its scope of application is limited.

(3) Contents of the invention

According to the problems existing in the prior art, the present invention provides a method for identifying an orthogonal modulation format based on a constellation diagram. After equalizing the received signal, a small number of normalized symbol sequences are taken, and the Normalize the constellation diagram for clustering; then analyze each cluster formed by a two-dimensional Gaussian model to obtain the degree of aggregation of each cluster; then calculate the average degree of aggregation of all clusters as the fit of the matching template , and determine the modulation format of the signal according to the preset fit range distribution. Because the clustering degree is different under different signal-to-noise ratio (SNR), the fit degree will change with the change of signal-to-noise ratio. The lower the signal-to-noise ratio, the smaller the fit degree. The invention has low computational complexity, wide application range of optical signal-to-noise ratio, and innovative practical value.

In order to achieve the above object, the technical scheme adopted in the present invention is as follows, including the following content:

(T₂,T₄,T₅,T₆)。切合度值大小可以代表匹配该模板的程度；例如：T₂越大，代表这H个符号为 4QAM调制格式的可能性越大。最后根据预设的四种模板的切合度范围判断待 测信号调制格式阶数。在SNR较高时，正确的模板匹配得到的切合度值会较大， 因此可以通过MQAM匹配模板切合度值就能准确区分M的值；当SNR较小 时，由于用来识别的H符号数较少(H<＝1000)，使得每一个星座团的信号数量大 约为1000/M，且由于受噪声影响较大，凝聚度不高，得到的某个切合度值不能 准确判断，通过多个切合度值组合进行判定，可以提高识别准确度。First, the modulated signal is received, and the equalized signal is obtained through digital signal processing. In order to reduce the computational complexity, 1000 symbols are taken from the equalized signal for identification, and the range of the real part and the imaginary part is limited to (-1.3～1.3) Within (the symbols that are too much affected by noise are eliminated), the number of symbols finally obtained for identification is H. Then, the H symbols taken for identification are matched with possible standard normalized MQAM signal templates respectively to perform clustering, and M clustered constellations are obtained respectively. After clustering, a two-dimensional Gaussian model analysis is performed on each cluster constellation to obtain the probability value of each symbol in the cluster. Then, the average value of the symbol probability values in the cluster constellation group is calculated separately to obtain the agglomeration degree of the cluster constellation group. Finally, the degree of fit of the corresponding MQAM normalized template is obtained through the degree of aggregation of the M constellation clusters

(T ₂ , T ₄ , T ₅ , T ₆ ). The size of the fit value can represent the degree of matching the template; for example, the larger T ₂ is, the higher the probability that the H symbols are in the 4QAM modulation format is. Finally, the order of the modulation format of the signal to be tested is determined according to the fit range of the preset four templates. When the SNR is high, the fit value obtained by correct template matching will be large, so the value of M can be accurately distinguished by matching the template fit value through MQAM; when the SNR is small, the number of H symbols used for identification is relatively large. less (H<=1000), so that the number of signals in each constellation group is about 1000/M, and due to the large influence of noise, the degree of cohesion is not high, and a certain fit value obtained cannot be accurately judged. The combination of degree values can be judged, which can improve the recognition accuracy.

(4) Description of drawings

1 is an identification process of a modulation format provided by an embodiment of the present invention;

2 is a constellation diagram of multiple normalized QAM modulation schemes adopted in the embodiment of the present invention;

Fig. 3 is the normalized 4QAM signal constellation diagram after the equalization received in the embodiment of the present invention, the constellation diagram for identifying the three symbol sequences of modulation format, and the result schematic diagram of matching 4QAM normalization template clustering;

4A is a schematic diagram of the first-order cohesion degree ε ₁ and the second-level cohesion degree β ₁ of the first cluster constellation group in which the normalized 4QAM signal received matches the normalized 4QAM signal template in the embodiment of the present invention;

Fig. 4B(a) is a 3D view of the 2D Gaussian model distribution of the first clustered constellation in Fig. 4A, Fig. 4B(b) is the part of Fig. 4B(a) where the probability value is greater than ε ₁ , Fig. 4B(c) ) is the part where the probability value is greater than β ₁ in Fig. 4B(a);

Fig. 5 is the 16QAM signal when the signal received in the embodiment of the present invention is SNR=20dB, and sampling H symbols is used to identify the step-by-step schematic diagram of the modulation format;

FIG. 6 is a flow chart of determining a suitability range preset in an embodiment of the present invention.

(5) Specific implementation methods

Below in conjunction with specific embodiment and accompanying drawing, the present invention is described in detail, it should be understood that the embodiment described below is only used to illustrate and explain the present invention, and is not intended to limit the present invention:

1 is an identification process of a modulation format provided by an embodiment of the present invention, which specifically includes the following steps:

S1: Receive a modulated signal, and obtain an equalized signal through digital signal processing. Taking FIG. 3(a) as an example, it shows an equalized constellation diagram obtained by performing digital signal processing on the received signal in the present invention. The constellation diagram When SNR=10dB, for the transmitted 4QAM modulated signal, the abscissa is the real part of the signal, and the ordinate is the imaginary part of the signal. In order to reduce the computational complexity, 1000 symbols are taken from the equalized signal for identification, and the range of the real and imaginary parts is limited within (-1.3 to 1.3) (symbols that are too affected by noise are eliminated). The number of symbols used for identification is H, and the constellation diagram is shown in Figure 3(b).

S2: The H symbols taken for identification are respectively matched with possible standard normalized MQAM signal templates for clustering, and the cluster centroids are marked as c _j ={c ₁ ^(j) ,c ₂ ^(j) } ,j∈{1,2,…,M}, then calculate the Euclidean between each initial centroid and each symbol x _i ={x ₁ ,x ₂ },i∈{1,2,…,H} Shi's distance:

则判定x_i属于质心c_j形成的聚类星座团，最后得到M个聚类星座团(匹配4QAM/16QAM/32QAM/64QAM模板进行聚类，分 别会得到4/16/32/64个星座团)。如图3(c)所示，形成的四个聚类星座团分 别用四种颜色表示。假设第j个星座团的符号个数为m_j,有：if any

Then determine that x _i belongs to the cluster constellation cluster formed by the centroid c _j , and finally get M cluster constellation clusters (matching the 4QAM/16QAM/32QAM/64QAM template for clustering, you will get 4/16/32/64 constellation clusters respectively. ). As shown in Fig. 3(c), the formed four cluster constellations are represented by four colors respectively. Assuming that the number of symbols in the jth constellation group is m _j , there are:

S3: After clustering, if m _j < 2, no need for S4 and S5 to directly determine that the matching degree of the template is 0. If m _j >= 2, perform two-dimensional Gaussian model analysis on each cluster constellation to obtain the probability value of each symbol in the cluster, taking the jth cluster constellation as an example:

n为维数：S301: Calculate the parameters of the two-dimensional Gaussian model of the jth cluster constellation cluster

n is the dimension:

为第i个信号的横坐标x₁，

为第i个信号的纵坐 标x₂，建立一个二维高斯模型；where i∈{1,2,…,m _j },

is the abscissa x ₁ of the i-th signal,

Establish a two-dimensional Gaussian model for the ordinate x ₂ of the i-th signal;

S302: Calculate the probability value p _ji of each signal in the cluster constellation to which it belongs:

i∈{1,2,…,m_j}，T表示转置。in,

i∈{1,2,…,m _j }, T stands for transpose.

S4: Calculate the average value of the symbol probability values p _ji in the clustered constellation group respectively to obtain the degree of agglomeration of the clustered constellation group: ε _j =mean{p _ji }.

(T₂,T₄,T₅, T₆)。切合度值大小可以代表匹配该模板的程度；例如：T₂越大，代表这H个符 号为4QAM调制格式的可能性越大。S5: Obtain the fit degree of the corresponding template through the degree of aggregation of the M constellation clusters

(T ₂ , T ₄ , T ₅ , T ₆ ). The size of the fit value can represent the degree of matching the template; for example, the larger T ₂ is, the higher the probability that the H symbols are in the 4QAM modulation format is.

S501 calculates T ₂ : Since each cluster constellation cluster of 4QAM has a large number of symbols, we calculate the second-order agglomeration degree β _j to calculate the fit degree of the 4QAM template. Take the symbol of p _ji >ε _j , calculate the mean probability of the remaining part of the symbol β _j =mean{p _ji }, p _ji ∈{p _ji >ε _j } to obtain the second-order cohesion degree. In Fig. 4A, the black circles indicate the size of the second degree of cohesion β ₁ of the first cluster constellation group, where p _ji of red symbols is less than ε _j , p _ji of black symbols is greater than ε _j and less than β _j , blue symbols p _ji is greater than β _j . Fig. 4B shows a 3D view of the 2D Gaussian model of the first clustered constellation cluster in 4QAM, where the bottom center in Fig. 4B(a) is (μ ₁₁ , μ ₁₂ ), and Fig. 4B(b) shows p _ji > The part of ε _j , Fig. 4B(c) shows the part of p _ji > β _j . Calculate the average of the second-order cohesion degrees of the j cluster constellations as the fit degree of the matching 4QAM template: T ₂ =mean(β).

个ε_j作为β_k,k∈{1,2,…,12}。计算β_k的平均值作为匹配16QAM模板切合度： T₄＝mean(β)。S502 Calculate T ₄ : after calculating the degree of agglomeration ε _j of the jth cluster constellation group, we take the larger

ε _j as β _k , k∈{1,2,…,12}. The mean value of βk was calculated as matching 16QAM template fit: T _{4 =} mean(β).

个ε_j作为β_k,k∈{1,2,…,24},计算β_k的平均值作为匹配32QAM模板切合度： T₅＝mean(β)。S503 calculates T ₅ : after calculating the degree of agglomeration ε _j of the jth cluster constellation group, we take the larger

Each ε _j is taken as β _k , k∈{1,2,...,24}, and the average value of β _k is calculated as the fit degree of the matching 32QAM template: T ₅ =mean(β).

个ε_j作为β_k,k∈{1,2,…,48},计算β_k的 平均值作为匹配64QAM模板切合度：T₆＝mean(β)。图5显示了接收的信号为 SNR＝20dB时的16QAM信号，采样H个符号用于识别调制格式的分步示意图； 其中黑色符号的p_ji大于β_j，红色符号的p_ji大于ε_j，蓝色符号的p_ji小于ε_j。从图 5中可以看出，匹配相同阶数的调制模板得到的各个聚类的符号分布较均匀，切 合度较高。匹配不同阶数的调制模板得到的某些聚类的符号个数可能为0，切合 度小。S504 calculates T ₆ : in order to improve the recognition accuracy of the 64QAM and 32QAM signals, calculate the sum of the symbols of the cluster constellation clusters matching the four corners of the 64QAM template, that is, in the 1st, 5th, 33rd and 37th cluster constellation clusters The sum of the number of symbols: sum(m ₁ +m ₅ +m ₃₃ +m ₃₇ ), if the sum is less than H/32, it is judged that the matching degree of the matching 64QAM template is 0, that is, T ₆ =0, if the sum is greater than or equal to H/32, after calculating the agglomeration degree εj of the _jth cluster constellation group, we take the larger

Each ε _j is taken as β _k , k∈{1,2,...,48}, and the average value of β _k is calculated as the matching degree of 64QAM template: T ₆ =mean(β). Fig. 5 shows the step-by-step schematic diagram of sampling H symbols for identifying the modulation format when the received signal is a 16QAM signal with SNR=20dB; the p _ji of the black symbol is greater than β _j , the p _ji of the red symbol is greater than ε _j , and the blue symbol p ji is greater than ε j . p _ji of color symbols is smaller than ε _j . It can be seen from Fig. 5 that the symbols of each cluster obtained by matching modulation templates of the same order are relatively uniform and have a high degree of fit. The number of symbols of some clusters obtained by matching modulation templates of different orders may be 0, and the degree of fit is small.

S6: Determine the modulation format order of the signal to be tested according to the fit ranges of the preset four templates. When the SNR is high, the fit value obtained by correct template matching will be large, so the value of M can be accurately distinguished by MQAM matching the template fit value; when the SNR is small, the number of H symbols used for identification is relatively large. less (H<=1000), so that the number of signals in each constellation group is about 1000/M, and due to the large influence of noise, the degree of cohesion is not high, and a certain fit value obtained cannot be accurately judged. The combination of degree values can be judged, which can improve the recognition accuracy. The detailed threshold judgment process is shown in Figure 6, and the sequence numbers A to S in Figure 6 correspond to the sequence numbers in Table 1:

S601: Identify 4QAM and 32QAM respectively;

S602: Identify 16QAM and 64QAM under higher SNR;

S603: The recognition accuracy of 16QAM and 64QAM under lower SNR;

S604: further improve the recognition accuracy of 16QAM and 64QAM under lower signal-to-noise ratio;

S605: Ensure the recognition accuracy of 16QAM under higher signal-to-noise ratio;

S606: Ensure the recognition accuracy of 32QAM under a higher signal-to-noise ratio;

S607: Ensure the recognition accuracy of 64QAM under a higher signal-to-noise ratio.

Table 1

serial number fit range M A T₂&gt;=0.95 4 B T₅&gt;5&amp;T₆=0&amp;T₄&lt;2.4 32 C T₆&gt;12&amp;T₄&lt;2.6 64 D T₆&gt;13&amp;T₄&lt;2.65 64 E T₄&gt;=2.65 16 F T₆&lt;10.4&amp;T₆&gt;9.7&amp;T₄&lt;2.36&amp;T<sub>5</sub >>4.5 64 G T₆&gt;10.4&amp;T₄&gt;2.36&amp;T₄&lt;2.38&amp;T<sub>5</sub >>4.5 64 H T₆&lt;=10.2&amp;T₆&gt;=9&amp;T₄&gt;2.3 16 I T₆&gt;9.5&amp;T₆&lt;10&amp;T₄&lt;2.6&amp;T₅ &gt;4.5&amp;T₅&lt;5.2 16 J T₆&gt;10&amp;T₄&lt;2.3&amp;T₅&gt;4.5 64 K T₆&gt;11&amp;T₄&lt;2.45 64 L T₆&gt;12&amp;T₄&lt;2.5 64 M T₆&lt;11&amp;T₄&gt;2.2&amp;T₅&lt;4.5 16 N T₆&lt;10.3&amp;T₄&gt;2.3 16 O T₆&lt;11.2&amp;T₄&gt;2.4 16 P T₄&gt;4&amp;T₅&lt;30&amp;T₂&lt;0.95 16 Q T₅&gt;4.9&amp;T₆=0&amp;T₄&lt;2.5&amp;T₂&lt ;0.95 32 R T₅&gt;6&amp;T₆=0 32 S T₆&gt;14&amp;&amp;T₄&lt;4 64

The optical signal-to-noise ratio range applicable to the above-mentioned identification methods for 4QAM format, 16QAM format, 32QAM format, and 64QAM format provided by the embodiment of the present invention is shown in Table 2, and the identification accuracy reaches 100% in the applicable range.

Table 2

The above description is only a preferred embodiment of the present invention, but the protection scope of the present invention is not limited to this. The equivalent replacement or change of the inventive concept thereof shall be included within the protection scope of the present invention.

(6) Main technical advantages

Existing automatic modulation format identification is devoted to finding features or feature combinations that are sufficient to distinguish different modulation formats. The identification method proposed by the present invention adopts a two-dimensional Gaussian model. The method is simple in principle and low in complexity. When the signal-to-noise ratio is high, The template obtained by the Gaussian model analysis has a high degree of fit and is easy to identify.

Claims (5)

1. a method for identifying an orthogonal modulation format based on a constellation diagram, is characterized in that, comprising: Normalize the symbol sequence of the received quadrature modulation signal, and obtain an equalized signal through digital signal processing; take 1000 symbols from the equalized signal for identification, and limit the real part and imaginary part range to (- 1.3～1.3) (the symbols that are too much affected by noise will be eliminated), the number of symbols finally obtained for identification is H; the H symbols taken for identification are respectively matched with possible standard normalized MQAM The signal templates are clustered to obtain M cluster constellations respectively; after clustering, a two-dimensional Gaussian model analysis is performed on each cluster constellation to obtain the probability value of each symbol in the cluster. Then calculate the average value of the symbol probability values in the cluster constellation group to obtain the cohesion degree of the cluster constellation group; calculate the degree of fit of the corresponding MQAM normalization template through the aggregation degree of the M constellation groups

Determine the modulation format order of the signal to be tested according to the fit range of the preset four templates. 2. the method according to claim 1, the H symbols that are taken for identification are matched with possible standard normalized MQAM signal templates respectively and carry out clustering, respectively, obtain M cluster constellations, specifically include: A normalized MQAM signal is generated as the initial centroids c _j ={c ₁ ^(j) ,c ₂ ^(j) },j∈{1,2,...,M}. Calculate the Euclidean distance between each initial centroid and each signal x _i ={x ₁ ,x ₂ },i∈{1,2,…,H}:

if any

Then it is determined that x _i belongs to the cluster constellation group formed by the centroid c _j . 3. method according to claim 1, carries out two-dimensional Gaussian model analysis to each cluster constellation group and obtains the probability value of each symbol in this cluster, calculates the mean value of symbol probability value in the cluster constellation group to obtain this. The degree of cohesion of the clustered constellation, including: Calculate the parameters of the 2D Gaussian model of the j,j∈{1,2,…,M}th cluster constellation

where n is the dimension, i∈{1,2,…,m _j },

is the abscissa x ₁ of the i-th signal in the j-th cluster constellation group,

is the ordinate x ₂ of the i-th signal; then calculate the probability value p _ji of each signal in the cluster constellation to which it belongs:

in,

n is the dimension, i∈{1,2,…,m _j }, and T denotes the transpose. The degree of agglomeration of the cluster constellation ε _j =mean{p _ji }. 4. method according to claim 1, calculates the fit degree of corresponding MQAM normalization template by the degree of agglomeration of these M constellation clusters

Specifically include: Calculate T ₂ : Calculate the second degree of aggregation β _j to calculate the 4QAM template fit. Take the symbol of p _ij > _j ε, and calculate the symbol probability mean β _j =mean{p _ji },p _ji ∈{p _ji >ε _j } of the remaining part to obtain the second-order agglomeration degree. Calculate the average of the second-order agglomeration degrees of the j cluster constellations as the fit degree of the matching 4QAM template: T ₂ =mean(β). Calculate T ₄ : After calculating the agglomeration degree ε _j of the jth cluster constellation group, we take the larger

Each ε _j is taken as β _k , k∈{1,2,...,12}, and the average value of β _k is calculated as the fit degree of the matching 16QAM template: T ₄ =mean(β). Calculate T ₅ : After calculating the agglomeration degree ε _j of the jth cluster constellation group, we take the larger

Each ε _j is taken as β _k , k∈{1,2,...,24}, and the average value of β _k is calculated as the fit degree of the matching 32QAM template: T ₅ =mean(β). Calculate T ₆ : Calculate the sum of the number of symbols in the cluster constellation clusters matching the back four corners of the 64QAM template, that is, the sum of the symbols in the 1st, 5th, 33rd and 37th cluster constellation clusters: sum(m ₁ + m5+m ₃₃ +m ₃₇ ), if the sum is less than H/32, then judge that the degree of fit of the matching 64QAM template is 0, that is, T ₆ =0, if the sum is greater than or equal to H/32, calculate the jth cluster constellation After the degree of aggregation _εj of the cluster, we take the larger

Each ε _j is taken as β _k , k∈{1,2,...,48}, and the average value of β _k is calculated as the matching degree of 64QAM template: T ₆ =mean(β). 5. according to the fit degree scope of the preset four kinds of templates according to claim 1, judge the modulation format order of the signal to be tested, specifically comprising: The order of the modulation format is determined according to the pre-set fit range shown in Table 1. Table 1 serial number fit range M A T₂&gt;=0.95 4 B T₅&gt;5&amp;T₆=0&amp;T₄&lt;2.4 32 C T₆&gt;12&amp;T₄&lt;2.6 64 D T₆&gt;13&amp;T₄&lt;2.65 64 E T₄&gt;=2.65 16 F T₆&lt;10.4&amp;T₆&gt;9.7&amp;T₄&lt;2.36&amp;T<sub>5</sub >>4.5 64 G T₆&gt;10.4&amp;T₄&gt;2.36&amp;T₄&lt;2.38&amp;T<sub>5</sub >>4.5 64 H T₆&lt;=10.2&amp;T₆&gt;=9&amp;T₄&gt;2.3 16 I T₆&gt;9.5&amp;T₆&lt;10&amp;T₄&lt;2.6&amp;T₅ &gt;4.5&amp;T₅&lt;5.2 16 J T₆&gt;10&amp;T₄&lt;2.3&amp;T₅&gt;4.5 64 K T₆&gt;11&amp;T₄&lt;2.45 64 L T₆&gt;12&amp;T₄&lt;2.5 64 M T₆&lt;11&amp;T₄&gt;2.2&amp;T₅&lt;4.5 16 N T₆&lt;10.3&amp;T₄&gt;2.3 16 O T₆&lt;11.2&amp;T₄&gt;2.4 16 P T₄&gt;4&amp;T₅&lt;30&amp;T₂&lt;0.95 16 Q T₅&gt;4.9&amp;T₆=0&amp;T₄&lt;2.5&amp;T₂&lt ;0.95 32 R T₅&gt;6&amp;T₆=0 32 S T₆&gt;14&amp;&amp;T₄&lt;4 64 If condition A is met, M is assigned a value of 4; if condition A is not met but condition B is met, M is assigned a value of 32; if condition C or D is met, M is assigned a value of 64, and if condition E is met, M is assigned a value of 16 ; If conditions F or G are met, M is assigned a value of 64, if conditions H or I are met, then M is assigned a value of 16; if conditions J or K or L are met, then M is assigned a value of 64, if conditions M or N or O are met , then M is assigned a value of 16; if the condition P is satisfied, M is assigned a value of 16; if the condition Q or R is satisfied, M is assigned a value of 32; if the condition S is satisfied, M is assigned a value of 64. Finally, the value of the modulation order M output by the identification method is obtained.

Images