CN102065056B - The method realizing the MQAM signal Modulation Mode Recognition of any planisphere based on cluster - Google Patents

Abstract

A kind of method that the invention discloses MQAM signal Modulation Mode Recognition realizing any planisphere based on cluster, for MQAM signal, order of modulation can be the biggest in theory, the pattern of modulation constellation can have the most multiple, therefore, often it is difficult to determine the possible range of signal modulation system to be identified, the most likely faces the modulation system of the unknown.The present invention proposes two step clustering algorithms and realizes the reconstruct of planisphere, is then based on reconstructing planisphere and completes Modulation Mode Recognition；Self adaptation subtractive clustering based on signal to noise ratio provides initial clustering result, namely completes preliminary star map reconstruction, is using fuzzy C-means clustering to complete final modulated signal constellation point reconstruct on this basis；After constellation point has reconstructed, the order of modulation of signal can be judged according to constellation point number, for the modulation system set that exponent number is identical, mode identification can be modulated to re-use Generalized Likelihood Ratio；The method using the present invention can realize the Modulation Mode Recognition of the MQAM signal of any planisphere by reconstruct planisphere.

Description

The method realizing the MQAM signal Modulation Mode Recognition of any planisphere based on cluster

Technical field

The present invention is a kind of Digital Communication Signal Modulation recognition methods, belongs to signal of communication and processes and detection technique neck Territory.

Background technology

Automatic recognition is a technology between signal detection and signal demodulate, and main task is to realize The intelligence of modulated signal receives, processes.At civilian aspect, government is in order to implement effective radio spectrum management, it is often necessary to Supervision civil signal is transmitted, in order to they are kept control or finds and monitor the transmitter of unregistered registration.Military and National security aspect, the application of Automatic modulation classification technology is the most extensive.In order to obtain communication intelligence, first have to judge the tune of signal Mode processed, could implement correctly to demodulate and information processing subsequently and analysis afterwards；In electronic warfare, for implementing electronics pair Anti-, electronic counter-countermeasures, threat detection, alarm, target acquistion and search etc., be required for being understood fully by Modulation identification technology relevant logical Letter or the parameter of electronic signal and character.In addition to multicarrier modulated signal MFSK, remaining modulation system MPSK, MASK and MQAM Signal can obtain the baseband complex signal of sign synchronization by all purpose communication signal processing algorithm, utilize this baseband complex signal I Patent of the present invention ground method can be utilized to complete the identification of modulation system.

Summary of the invention

A kind of method that the invention provides MQAM signal Modulation Mode Recognition realizing any planisphere based on cluster.

The object of the present invention is achieved like this:

For MQAM signal, order of modulation can be very big in theory, and the pattern of modulation constellation can have the most multiple, because of This Modulation Identification is highly difficult；Modulation constellation represents different modulation systems, and the present invention proposes two step clustering algorithms and realizes star The reconstruct of seat figure, is then based on reconstructing planisphere and completes Modulation Mode Recognition, concretely comprise the following steps:

Step one: Signal Pretreatment: mainly by universal demodulation algorithm, demodulate sign synchronization from intermediate-freuqncy signal Complex baseband signal；

Step 2: use self adaptation subtractive clustering based on signal-to-noise ratio (SNR) estimation, this is first step clustering algorithm, main purpose It is initial number and the initial position providing cluster centre,

The complex baseband signal r of sign synchronization_kFor

$r_k=R_k{e}^{j{\mathrm{\θ}}_k}+n_k,$ K=1,2 ..., N. (1)

Wherein, whereinThe modulation constellation points of representation signal, n_kRepresenting Gaussian noise, N represents that sample sequence is long Degree；Signal to noise ratio is SNR, and this value is obtained by blind signal-to-noise ratio (SNR) estimation；By SNR, calculate average noise power P_n, enter And calculate " density " index D of each sampled point_k:

$D_k=\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\exp (-\frac{{\left|\right|r_k-r_j\left|\right|}^2}{K_a*P_n}).---\left(2\right)$

Wherein, K_aIt it is a regulation coefficient.According to the result of formula (2), select D_kMaximum D_{C, 1}Corresponding sampled point r_{C, 1} It is first cluster centre；Then, to " density " index correction above；

$D_k=D_k-D_{c,1}\exp (-\frac{{\left|\right|r_k-r_{c,1}\left|\right|}^2}{K_b*P_n}).---\left(3\right)$

Wherein, K_bIt it is a regulation coefficient.Second cluster centre r is selected by same method_{C, 2}；

From the 3rd cluster centre r_{C, 3}Start, select new cluster centre r every time_{C, l+1}After, it is judged that cluster centre is the most complete Portion selects:

min{‖r_{C, k}-r_{C, l+1}‖²＞ K_c*P_n, k=1,2 ..., l. (4)

Wherein, K_bIt is a regulation coefficient, if formula (4) is set up, then proceeds subtractive clustering, otherwise terminate；

Step 3: use C means clustering algorithm or Fuzzy C-Means Cluster Algorithm reconstruct planisphere.This is second step cluster Algorithm, main purpose is to complete the reconstruct of planisphere based on first step cluster result；

R={r_k, k=1,2 ..., N} is the sample set of N number of baseband complex signal sample composition, and C is predetermined classification number Mesh, m_i, i=1,2 ... C is the center of each cluster, μ_i(r_k) it is the kth sample membership function for ith cluster；

According to known initial cluster center, calculate the sample degree of membership size to all initial clustering points, degree of membership letter Number is defined as follows

${\mathrm{\μ}}_i\left(r_k\right)=\frac{{(1/{\left|\right|r_k-m_i\left|\right|}^2)}^{1/(b-1)}}{\underset{l=1}{\overset{C}{\mathrm{\Σ}}}{(1/{\left|\right|r_k-m_l\left|\right|}^2)}^{1/(b-1)}},$ K=1,2 ..., N, i=1,2 ..., C. (5)

After degree of membership has calculated, the value of calculating cluster centre under this degree of membership

$m_i=\frac{\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{\left[{\mathrm{\μ}}_i\left(r_k\right)\right]}^b{r}_k}{\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{\left[{\mathrm{\μ}}_i\left(r_k\right)\right]}^b},$ I=1,2 ..., C. (6)

After having had cluster centre and degree of membership, the cluster cost function that calculating defines with membership function

$J_f=\underset{i=1}{\overset{C}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{\left[{\mathrm{\μ}}_i\left(r_k\right)\right]}^b{\left|\right|r_k-m_i\left|\right|}^2.---\left(7\right)$

Wherein, b ＞ 1 is the constant of a fog-level that can control cluster result；By above formula (5), (6), (7) iterate so that cluster centre gradually converges to cluster the point that cost function is minimum, is now subordinate to angle value and tends to steady Fixed, the termination condition of iteration is

J_f(k+1)≈J_f(k). (8)

Star map reconstruction can be completed by above two clustering algorithms.

Step 4: use Generalized Likelihood Ratio test (GLRT:General LikelihoodRate based on reconstruct planisphere Test), the identification of modulation system is completed: first according to the constellation point number reconstructed, it is judged that modulation system exponent number；If waiting to know The exponent number having two kinds of modulation systems in other modulation system set is the same, then be continuing with GLRT classification；

By reconstruct constellation point m_iAnd modulation constellation pointsBetween Euclidean distance nearest, to the pole of first modulation constellation points Maximum-likelihood is estimatedSubscript c represents different modulation types；Then GLRT classification same order modulation type is used.Expression formula As follows

Wherein,Represent the order of modulation estimated,Represent and make l_{GLR, m}Obtain minima Modulation system c；Final Modulation Mode Recognition is completed by formula (10).

This method has an advantage that the identification of the modulation system of the MQAM signal that can complete any planisphere.MQAM signal Order of modulation can be from 4 rank to 1024 rank (the highest), and modulation constellation can also arrange change, traditional tune according to both sides Recognition methods processed, the MQAM range of signal that can identify is the most limited.The method using the present invention to propose completes Modulation Mode Recognition Have: the advantages such as MQAM signal constellation (in digital modulation) figure scope is unrestricted, discrimination is high, anti-Gaussian noise ability is strong.

Accompanying drawing explanation

Fig. 1 is present invention digital modulation signals based on star map reconstruction Modulation Mode Recognition method handling process；

Fig. 2 is the design sketch using the present invention to carry out the reconstruct of 16QAM signal constellation (in digital modulation) figure；

Fig. 3 is the design sketch using the present invention to carry out V.29/16QAM signal constellation (in digital modulation) figure reconstruct；

Detailed description of the invention

Below in conjunction with the accompanying drawings, the present invention is further illustrated:

For MQAM signal, order of modulation can be very big in theory, and the pattern of modulation constellation can have the most multiple, because of This Modulation Identification is highly difficult；Modulation constellation represents different modulation systems, and the present invention proposes two step clustering algorithms and realizes star The reconstruct of seat figure, is then based on reconstructing planisphere and completes Modulation Mode Recognition, concretely comprise the following steps:

Step one: Signal Pretreatment: mainly by universal demodulation algorithm, demodulate sign synchronization from intermediate-freuqncy signal Complex baseband signal；

Step 2: use self adaptation subtractive clustering based on signal-to-noise ratio (SNR) estimation, this is first step clustering algorithm, main purpose It is initial number and the initial position providing cluster centre,

The complex baseband signal r of sign synchronization_kFor

$r_k=R_k{e}^{j{\mathrm{\θ}}_k}+n_k,$ K=1,2 ..., N. (11)

Wherein, whereinThe modulation constellation points of representation signal, n_kRepresenting Gaussian noise, N represents that sample sequence is long Degree；Signal to noise ratio is SNR, and this value is obtained by blind signal-to-noise ratio (SNR) estimation；By SNR, calculate average noise power P_n, And then calculate " density " index D of each sampled point_k:

$D_k=\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\exp (-\frac{{\left|\right|r_k-r_j\left|\right|}^2}{K_a*P_n}).---\left(12\right)$

Wherein, K_aIt it is a regulation coefficient.According to the result of formula (2), select D_kMaximum D_{C, 1}Corresponding sampled point r_{C, 1} It is first cluster centre；Then, to " density " index correction above；

$D_k=D_k-D_{c,1}\exp (-\frac{{\left|\right|r_k-r_{c,1}\left|\right|}^2}{K_b*P_n}).---\left(13\right)$

Wherein, K_bIt it is a regulation coefficient.Second cluster centre r is selected by same method_{C, 2}；

From the 3rd cluster centre r_{C, 3}Start, select new cluster centre r every time_{C, l+1}After, it is judged that cluster centre is the most complete Portion selects:

min{‖r_{C, k}-r_{C, l+1}‖²＞ K_c*P_n, k=1,2 ..., l. (14)

Wherein, K_bIt is a regulation coefficient, if formula (4) is set up, then proceeds subtractive clustering, otherwise terminate；

Step 3: use C means clustering algorithm or Fuzzy C-Means Cluster Algorithm reconstruct planisphere.This is second step cluster Algorithm, main purpose is to complete the reconstruct of planisphere based on first step cluster result；

R={r_k, k=1,2 ..., N} is the sample set of N number of baseband complex signal sample composition, and C is predetermined classification number Mesh, m_i, i=1,2 ... C is the center of each cluster, μ_i(_rK) it is the kth sample membership function for ith cluster；

According to known initial cluster center, calculate the sample degree of membership size to all initial clustering points, degree of membership letter Number is defined as follows

${\mathrm{\μ}}_i\left(r_k\right)=\frac{{(1/{\left|\right|r_k-m_i\left|\right|}^2)}^{1/(b-1)}}{\underset{l=1}{\overset{C}{\mathrm{\Σ}}}{(1/{\left|\right|r_k-m_l\left|\right|}^2)}^{1/(b-1)}},$ K=1,2 ..., N, i=1,2 ..., C. (15)

After degree of membership has calculated, the value of calculating cluster centre under this degree of membership

$m_i=\frac{\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{\left[{\mathrm{\μ}}_i\left(r_k\right)\right]}^b{r}_k}{\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{\left[{\mathrm{\μ}}_i\left(r_k\right)\right]}^b},i=\mathrm{1,2},...,C.---\left(16\right)$

After having had cluster centre and degree of membership, the cluster cost function that calculating defines with membership function

$J_f=\underset{i=1}{\overset{C}{\mathrm{\Σ}}}\underset{k=1}{\overset{N}{\mathrm{\Σ}}}{\left[{\mathrm{\μ}}_i\left(r_k\right)\right]}^b{\left|\right|r_k-m_i\left|\right|}^2.---\left(17\right)$

Wherein, b ＞ 1 is the constant of a fog-level that can control cluster result；By above formula (5), (6), (7) iterate so that cluster centre gradually converges to cluster the point that cost function is minimum, is now subordinate to angle value and tends to steady Fixed, the termination condition of iteration is

J_f(k+1)≈J_f(k). (18)

Star map reconstruction can be completed by above two clustering algorithms.

Step 4: use Generalized Likelihood Ratio test (GLRT:General LikelihoodRate based on reconstruct planisphere Test), the identification of modulation system is completed: first according to the constellation point number reconstructed, it is judged that modulation system exponent number；If waiting to know The exponent number having two kinds of modulation systems in other modulation system set is the same, then be continuing with GLRT classification；

By reconstruct constellation point m_iAnd modulation constellation pointsBetween Euclidean distance nearest, to the pole of first modulation constellation points Maximum-likelihood is estimatedSubscript c represents different modulation types；Then GLRT classification same order modulation type is used.Expression formula As follows

Wherein,Represent the order of modulation estimated,Represent and make l_{GLR, m}Obtain minima Modulation system c；Final Modulation Mode Recognition is completed by formula (10).

Fig. 1 is present invention digital modulation signals based on star map reconstruction Modulation Mode Recognition method handling process, therefrom Frequently signal uses universal demodulation algorithm to draw required sign synchronization baseband signal, then uses self adaptation subtractive clustering to complete Initial clustering, provides the initial cluster center value required for fuzzy C-means clustering.It is then used by Fuzzy C-Means Cluster Algorithm, The signal modulation constellation that deviation is the least can be reconstructed.Finally utilize reconstruct planisphere, use GLRT grader, identify tune Mode processed.

Illustrate, the 16QAM (square) that bit signal to noise ratio (Eb/No) is 5dB, V.29/16QAM signal are carried out respectively Star map reconstruction, now the bit error rate of two modulated signals is all 16%.Fig. 2 is to use the present invention to carry out 16QAM signal constellation (in digital modulation) figure The design sketch of reconstruct.Fig. 3 is the design sketch using the present invention to carry out V.29/16QAM signal constellation (in digital modulation) figure reconstruct.Can from two figures Going out, received signal points is because containing noise, dispersion is on a complex plane.Initial cluster center is given by self adaptation subtractive clustering, Make to reconstruct planisphere by fuzzy C-means clustering and actual modulated constellation point is sufficiently close to.Therefore we can use reconstruct Planisphere completes Modulation Identification.

This method has an advantage that the identification of the modulation system of the MQAM signal that can complete any planisphere.MQAM signal Order of modulation can be from 4 rank to 1024 rank (the highest), and modulation constellation can also arrange change, traditional tune according to both sides Recognition methods processed, the MQAM range of signal that can identify is the most limited.The method using the present invention to propose completes Modulation Mode Recognition Have: the advantages such as MQAM signal constellation (in digital modulation) figure scope is unrestricted, discrimination is high, anti-Gaussian noise ability is strong.

Above example is merely to illustrate the preferred embodiment of the present invention, but the present invention is not limited to above-mentioned embodiment party Formula, in the ken that described field those of ordinary skill is possessed, that is made within the spirit and principles in the present invention is any Amendment, equivalent replacement and improvement etc., it all should be contained within the scope of the technical scheme that the present invention is claimed.

Claims (1)

1. the method realizing the MQAM signal Modulation Mode Recognition of any planisphere based on cluster, it is characterised in that include walking as follows Rapid:

Step one: Signal Pretreatment: mainly by universal demodulation algorithm, demodulate the complex radical of sign synchronization from intermediate-freuqncy signal Band signal；

Step 2: self adaptation subtractive clustering reconstruct planisphere, this is first step clustering algorithm, and main purpose is to provide cluster centre Initial number and initial position,

The complex baseband signal r of sign synchronization_kFor

$r_k=R_k{e}^{{\mathrm{j\θ}}_k}+n_k,k=1,2,L,N.---\left(1\right)$

Wherein, whereinThe modulation constellation points of representation signal, n_kRepresenting Gaussian noise, N represents sample sequence length；Noise Ratio is SNR, and this value is obtained by blind signal-to-noise ratio (SNR) estimation；By SNR, calculate average noise power P_n, and then calculate " density " index D of each sampled point_k；

$D_k=\underset{j=1}{\overset{N}{\mathrm{\Σ}}}\exp (-\frac{\left|\right|r_k-r_j|{|}^2}{K_a*P_n}).---\left(2\right)$

Wherein, K_aIt it is a regulation coefficient；According to the result of formula (2), select D_kMaximum D_c,1Corresponding sampled point r_c,1It is One cluster centre；Then, to " density " index correction above；

$D_k=D_k-D_{c,1}\exp (-\frac{\left|\right|r_k-r_{c,1}|{|}^2}{K_b*P_n}).---\left(3\right)$

Wherein, K_bIt it is a regulation coefficient；Second cluster centre r is selected by same method_c,2；

From the 3rd cluster centre r_c,3Start, select new cluster centre r every time_c,l+1After, it is judged that cluster centre the most all selects Go out；

min{||r_c,k-r_c,l+1||²}>K_c*P_n, k=1,2, L, l. (4)

Wherein, K_cIt it is a regulation coefficient；If formula (4) is set up, then proceed subtractive clustering, otherwise terminate；

Step 3, use Fuzzy C-Means Cluster Algorithm or C means clustering algorithm reconstruct planisphere, this is second step clustering algorithm, Main purpose is to complete the reconstruct of planisphere based on first step cluster result；

{r_k, k=1,2, L, N} are the sample sets of N number of baseband complex signal sample composition,The cluster estimated for subtractive clustering Center Number, m_i, i=1,2, LFor the center of each cluster, μ_i(r_k) it is the kth sample degree of membership for ith cluster Function；

According to known initial cluster center, calculating the sample degree of membership size to all initial clustering points, membership function is fixed Justice is as follows

${\mathrm{\μ}}_i\left(r_k\right)=\frac{{\left(1/\left|\right|r_k-m_i|{|}^2\right)}^{1/\left(b-1\right)}}{\underset{l=1}{\overset{\hat{C}}{\mathrm{\Σ}}}{\left(1/\left|\right|r_k-m_l|{|}^2\right)}^{1/\left(b-1\right)}},k=1,2,L,N,i=1,2,L,\hat{C}.---\left(5\right)$

After degree of membership has calculated, calculate the cluster centre value under this degree of membership

$m_i=\frac{\underset{k=1}{\overset{N}{\Σ}}{\[{\mathrm{\μ}}_i\left(r_k\right)\]}^b{r}_k}{\underset{k=1}{\overset{N}{\Σ}}{\[{\mathrm{\μ}}_i\left(r_k\right)\]}^b},i=1,2,L,\hat{C}.---\left(6\right)$

After having had cluster centre and degree of membership, the cluster cost function that calculating defines with membership function

$J_f=\underset{i=1}{\overset{\hat{C}}{\Σ}}\underset{k=1}{\overset{N}{\Σ}}{\[{\mathrm{\μ}}_i\left(r_k\right)\]}^b\left|\right|r_k-m_i|{|}^2.---\left(7\right)$

Wherein, b > 1 is the constant of a fog-level that can control cluster result, anti-by above formula (5) (6) (7) Multiple iteration so that cluster centre gradually converges to cluster the point that cost function is minimum, is now subordinate to angle value and tends towards stability, iteration Termination condition is

J_f(k+1)≈J_f(k). (8)

Star map reconstruction can be completed by above two step clustering algorithms；

Step 4, Generalized Likelihood Ratio test (GLRT) based on reconstruct planisphere is used to complete Modulation Mode Recognition:

First according to the constellation point number reconstructed, it is judged that modulation system exponent number；If modulation system set to be identified has two The exponent number planting modulation system is the same, then be continuing with GLRT classification；

By reconstruct constellation point m_iAnd modulation constellation pointsBetween Euclidean distance nearest, provide the maximum likelihood of modulation constellation points EstimateSubscript c represents different modulation types；Then GLRT classification same order modulation type is used；Expression formula is as follows

Wherein,Represent the order of modulation estimated,Represent and make l_GLR,mObtain the modulation of minima Mode c, completes final Modulation Mode Recognition by formula (10).