CN103490863A - Space-time-code mode blind identification method based on partial sequence parameter detection - Google Patents

Abstract

The invention discloses a space-time-code mode blind identification method based on partial sequence parameter detection. The method mainly solves the problems that in the prior art, calculation complexity is high and the code pattern identification range is narrow. The method includes the steps of (1) extracting a code pattern set to obtain a characteristic quantity set, (2) calculating a parallel-connection matrix and calculating a characteristic quantity function value vector by means of the matrix, (3) pre-estimating a characteristic quantity by means of the characteristic quantity function value vector to obtain a new space-time-code set, (4) writing symbolic number vectors of code patterns in the new space-time-code set, (5) solving a parameter estimation vector, (6) solving a distance decision value vector by means of the step (4) and the step (5), and (7) taking the code pattern corresponding to a minimum element in the distance decision value vector as a decision code pattern. The method overcomes the defect that in the prior art, due to the fact that time-lag relative field norms are intensive in computation, system complexity is high, increases the identification range of an exiting space-time-code blind identification method and can be used for space-time-code mode identification in communication countermeasure.

Description

Pattern blind-identification method during based on partial sequence parameter detecting empty

Technical field

The invention belongs to communication technical field, the Space-Time Block Coding coding mode blind-identification method in signal detection technique field while further relating to sky, can be used for, in the multiple-input and multiple-output mimo system, Space-Time Block Coding being carried out to blind identification.

Background technology

Mimo system is the key technology of next generation wireless communication, and empty time-code is the important component part of mimo system.The blind identification of empty time-code is the field of communication countermeasures field in the urgent need to research, and it can have important theory significance and using value for the mimo system countermeasure techniques provides basis and technical support, has caused the concern of academia.

The blind identification of empty time-code is an emerging problem, and existing algorithm is divided into maximum likelihood algorithm and time lag related algorithm.The maximum likelihood algorithm construction of function is simple, and computation complexity is low, but it for block length the pattern None-identified identical with the class symbol number; The discernible pattern of time lag related algorithm is more, but its computation complexity is difficult to be applied to real-time detection in practice along with the sampling number exponentially increases.

Document [1V.Choqueuse, M.Marazin et al., Blind recognition of linear space time block codes:A likelihoodbased approach.IEEE Trans.Signal Processing, 58 (3), 2010,1290-1299] in the code parameters detection algorithm that proposes belong to maximum likelihood algorithm.To all Candidate Set patterns, construct only relevant with coding parameter likelihood function according to maximum-likelihood criterion, by comparing the likelihood function of different coding pattern, coding parameter is made to judgement, and then judge pattern.The method decision function simple structure, computation complexity is low, in the blind recognition system of MIMO, is widely used., when but the method is carried out the MIMO detection in engineering practice, the deficiency of existence is: to the multiple coding mode None-identified with same packets length and every grouping internal symbol number.

Document [2V.Choqueuse, K.Yao et al., Blind recognition of linear space time block codes.IEEE Int.Conf.Acoust.Speech Signal Process, 2008,2833-2836] in the Decision Classfication detection algorithm that proposes belong to the time lag related algorithm.It adopts contrast step by step according to the otherness of the Frobenius norm of correlation matrix under different delay of different Space-Time Block Codings, realizes the blind identification to Space-Time Block Coding.Because the discernible pattern of the method is wider, and very superior to the detection performance of orthogonal space time packet, therefore in detecting, MIMO also obtained certain application.But also there is a lot of deficiencies in the method in the blind recognition system of MIMO: be mainly manifested in and can't distinguish the multiple pattern with identical F Norm Solution, computation complexity increases along with sampling time length becomes how much multiples.

Summary of the invention

The object of the invention is to the deficiency for above-mentioned prior art, propose a kind of pattern blind-identification method during based on partial sequence parameter detecting empty, to improve empty time-code identification of code type scope, reduce the complexity of calculating.

Realize that the object of the invention ground technical thought is: by adopting characteristic quantity pre-estimation technology, the multipath reception signal is carried out to the pre-estimation of grouping feature amount, utilize the grouping feature amount estimated, dwindle the set of sky time-code; Then utilize partial sequence code parameters detection technique, carry out the detection of partial sequence code parameters, utilize detected partial sequence code parameters to find the judgement pattern in the empty time-code set of dwindling.Concrete scheme comprises the steps:

1) receiving terminal receives by r root reception antenna the burst that length that transmitting terminal sends is N, obtains the reception signal matrix R' of r * N, N >=64 wherein, r >=2;

2) utilize all empty time-codes that need identification, form the pattern set omega, get combination (s, k) the constitutive characteristic duration set (U, V) that the class symbol of every kind of pattern in Ω is counted s and block length k, remember that i is combined as (s _i, k _i),

I=1,2... Ζ, Ζ is number of combinations in characteristic quantity set (U, V);

3) will receive real part and the imaginary part parallel connection of signal matrix R', obtain incidence matrix R,

$R=\left[\begin{array}{c}\mathrm{Re}\left(R^{\′}\right)\\ \mathrm{Im}\left(R^{\′}\right)\end{array}\right]$

Wherein Re () means to get the real part computing, and Im () means to get imaginary-part operation;

4) calculated characteristics flow function value:

4a) for i characteristic quantity combination (s in characteristic quantity set (U, V) _i, k _i), structure grouping correlation matrix R _i:

$R_i=\left[\begin{array}{cccc}\tilde{R}\left(1\right)& \tilde{R}(k_i+1)& \·\·\·& \tilde{R}((\frac{N}{k_i}-1)k_i+1)\\ \tilde{R}\left(2\right)& \tilde{R}(k_i+2)& \·\·\·& \tilde{R}((\frac{N}{k_i}-1)k_i+2)\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \tilde{R}\left(k_i\right)& \tilde{R}\left({2k}_i\right)& \·\·\·& \tilde{R}\left(\left(\frac{N}{k_i}\right)k_i\right)\end{array}\right]$

Wherein R (τ) means the τ row of incidence matrix R, τ=1,2...N;

4b) calculate grouping correlation matrix R _ithe grouping covariance matrix: C _i=E[R _ir _i ^t], E[wherein] mean to ask expectation computing, () ^tmean the transposition computing;

4c) to grouping covariance matrix C _ido Eigenvalues Decomposition, by the characteristic value that obtains by descending, constitutive characteristic value vector

wherein, ρ _ηfor grouping covariance matrix C _icharacteristic value, η=1,2 ... 2rk _i, get feature value vector

front 2s _iindividual characteristic value forms validity feature value vector form noise characteristic value vector by remaining characteristic value

?

${\stackrel{\→}{\mathrm{\λ}}}_{i1}=[{\mathrm{\ρ}}_1,{\mathrm{\ρ}}_2,...{\mathrm{\ρ}}_{2s_i}],{\stackrel{\→}{\mathrm{\λ}}}_{i2}=[{\mathrm{\ρ}}_{2s_i+1},{\mathrm{\ρ}}_{2s_i+2},...,{\mathrm{\ρ}}_{2r{k}_i}];$

4d) according to step 4c), obtain characteristic quantity combination (s _i, k _i) characteristic of correspondence flow function value M (s _i, k _i):

$M(s_i,k_i)=\frac{N}{k_i}\left(\log \left(\frac{\sqrt{\∏{\stackrel{\→}{\mathrm{\λ}}}_{i1}}}{\∏{\stackrel{\→}{\mathrm{\λ}}}_{i2}}\right)\right)$ Wherein ∏ () means that vector element connects multiplication;

5) pre-estimation characteristic quantity, obtain new empty time-code set omega ':

5a) to every kind in characteristic quantity set (U, V) combination repeating step 4, obtain every kind of combination characteristic of correspondence flow function value, composition characteristic flow function value vector: Φ=[M (s ₁, k ₁), M (s ₂, k ₂) ... M (s _m, k _m...)], wherein, m=1,2 ... Ζ;

5b) find out the element characteristic of correspondence amount combination of numerical value minimum in characteristic quantity functional value vector Φ

obtain new empty time-code set omega ';

6) write the sky time-code set omega that makes new advances ' in the symbolic number of each row of encoder matrix of j kind pattern, form symbolic number vector P _j, j=1 wherein, 2...T, T be new empty time-code set omega ' the pattern number;

7) estimating part symbolic number:

7a) utilize R in step (3) and the block length k in step (5b), structure part correlation matrix V:

$\begin{array}{c}\begin{array}{cccc}V=[\tilde{R}\left(\mathrm{\β}\right)& \tilde{R}(k+\mathrm{\β})& \·\·\·& \tilde{R}\end{array}\end{array},((\frac{N}{k}-1)k+\mathrm{\β})],$

Wherein, β be new empty time-code set omega ' in the row mark of encoder matrix of pattern, β=1,2 ... k;

7b) the part covariance matrix of calculating section correlation matrix V: D=E[VV ^t], to the part covariance matrix, D does Eigenvalues Decomposition, by the characteristic value that obtains by descending, the component part feature value vector ε wherein _γfor the characteristic value of part covariance matrix D, γ=1,2 ... 2rk;

7c) respectively will $n=1,n=2...n=\hat{s}$ Bring following formula into and calculate the likelihood function value

$L\left(n\right)=\begin{array}{c}-n(4r-2n+1)-\frac{N}{2k}\underset{\mathrm{\α}=1}{\overset{2n}{\mathrm{\Σ}}}\log \left({\mathrm{\ϵ}}_{\mathrm{\α}}\right)\\ -\frac{N}{k}(r-n)\log \left(\frac{1}{2(r-n)}\underset{\mathrm{\α}=2n+1}{\overset{2r}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{\mathrm{\α}}\right)\end{array}$

And then obtain likelihood function value vector

7d) getting n corresponding to numerical value greatest member in likelihood function value vector L is the part symbolic number;

8) calculate apart from the decision value vector:

8a) repeating step (7), estimate the part symbolic number of each row of encoder matrix, obtains the parameter Estimation vector

Ψ=[n (1), n (2) ... n (β) ...], the part symbolic number that wherein n (β) is encoder matrix β row;

8b) utilize the symbolic number vector P in step (6) _j, compute sign number vector P _jwith parameter Estimation vector Ψ apart from decision value θ _j=(P _j-Ψ) ²;

8c) repeating step (8b), obtain new empty time-code set omega ' in every kind of pattern and parameter Estimation vector Ψ apart from decision value, form apart from decision value vector Π=[θ ₁, θ ₂... θ _t];

9) getting the pattern corresponding apart from the element of numerical value minimum in decision value vector Π is judgement pattern, the blind identification of pattern while completing sky.

The present invention compared with prior art has the following advantages:

First, the present invention is owing to having adopted partial sequence parameter detecting technology, can estimate the coding symbol number of each row of clearancen time-code encoder matrix, and identify the transmission pattern by the symbolic number of each row, overcome time lag related algorithm in the prior art and can not identify the pattern with the relevant norm of identical time lag, the maximum likelihood function algorithm can not be identified the deficiency with the digital type of same packets length and class symbol, the identification range of empty time-code that made the present invention increase.

Second, the present invention is owing to having adopted partial sequence parameter detecting technology, directly adjudicate pattern therefore can utilize the partial sequence parameter of estimation, thereby do not need calculating to receive the relevant norm of time lag of signal, overcome in the time lag related algorithm the relevant norm calculation amount of time lag of empty time-code large, the high shortcoming of system implementation complexity caused, obviously reduce the complexity that the present invention realizes.

The accompanying drawing explanation

Fig. 1 is the system block diagram that the present invention uses;

Fig. 2 is realization flow figure of the present invention;

Fig. 3 is with the recognition correct rate figure of the present invention under different parameters;

Fig. 4 is the present invention and the recognition correct rate comparison diagram that has two kinds of blind-identification methods now.

Embodiment

With reference to Fig. 1, the system that the present invention uses comprises: the t transmit antennas, and r root reception antenna, modulation system is 4QAM.At transmitting terminal, be converted to the parallel sequence that sends after serial transmission sequence is space-time encoded, then will send after the Parallel Sequence modulation.At receiving terminal, the reception signal matrix is R':R'=HX+B, wherein, t >=2, r>t, H is the channel matrix that element is independently obeyed multiple Gaussian Profile, and X is the information sequence of emission, and B is the white Gaussian noise matrix, t=3 in this example.

The present invention is exactly according to receiving signal matrix R', the blind Space Time Coding pattern that identifies the transmitting terminal use.

With reference to Fig. 2, specific implementation step of the present invention is as follows:

Step 1, receiving terminal receives by r root reception antenna the burst that length that transmitting terminal sends is N, obtains the reception signal matrix R' of r * N, N >=64, N=1024 or N=512 in this example, r=8 or 6.

Step 2, obtain characteristic quantity set (U, V):

2a) utilize all empty time-codes that need identification, form the pattern set omega, Ω comprises orthogonal space time packet, quasi-orthogonal space time block code and non-orthogonal space-time block.In this example, the pattern set omega is: Ω=BALST (3,1),

Tarokh-OSTBC(4,8),Ganesan1-OSTBC(3,4),Ganesan2-OSTBC(3,4),Tarokh-OSTBC(3,4),Tarokh-OSTBC(4,4)}，

Wherein, BLAST is hierarchical space-time code, and OSTBC is orthogonal space time packet, and the bracket content after empty time-code represents that the class symbol of pattern counts the combination of s and block length k;

2b) get characteristic quantity combination (s, k) constitutive characteristic duration sets (U, V) all in Ω, in this example, set (U, V) is: (U, V)={ (3,1), (4,8), (3,4), (4,4) }, in the note set, i is combined as (s _i, k _i), i=1,2... Ζ, Ζ is number of combinations in characteristic quantity set (U, V), Z=4 in this example, each combination (s _i, k _i) a kind of pattern or multiple pattern in corresponding empty time-code set omega.

Step 3, will receive real part and the imaginary part parallel connection of signal matrix R', obtains incidence matrix R:

$R=\left[\begin{array}{c}\mathrm{Re}\left(R^{\′}\right)\\ \mathrm{Im}\left(R^{\′}\right)\end{array}\right]$

Wherein Re () means to get the real part computing, and Im () means to get imaginary-part operation.

Step 4, calculated characteristics flow function value:

4a) for i characteristic quantity combination (s in characteristic quantity set (U, V) _i, k _i), structure grouping correlation matrix R _i:

$R_i=\left[\begin{array}{cccc}\tilde{R}\left(1\right)& \tilde{R}(k_i+1)& \·\·\·& \tilde{R}((\frac{N}{k_i}-1)k_i+1)\\ \tilde{R}\left(2\right)& \tilde{R}(k_i+2)& \·\·\·& \tilde{R}((\frac{N}{k_i}-1)k_i+2)\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \tilde{R}\left(k_i\right)& \tilde{R}\left({2k}_i\right)& \·\·\·& \tilde{R}\left(\left(\frac{N}{k_i}\right)k_i\right)\end{array}\right]$

Wherein R (τ) means the τ row of incidence matrix R, τ=1,2...N;

4b) calculate grouping correlation matrix R _ithe grouping covariance matrix: C _i=E[R _ir _i ^t], E[wherein] mean to ask expectation computing, () ^tmean the transposition computing;

4c) adopt the orthogonal diagonal factorization method to grouping covariance matrix C _ido Eigenvalues Decomposition, first at grouping covariance matrix C _iorthogonal matrix Q and transposed matrix thereof are multiplied by respectively in both sides, obtain characteristic value diagonal matrix Δ=Q ^tc _iq; Extract again grouping covariance matrix C from characteristic value diagonal matrix Δ _icharacteristic value;

4d) by the characteristic value that obtains by descending, constitutive characteristic value vector

wherein, ρ _ηfor grouping covariance matrix C _icharacteristic value, η=1,2 ... 2rk _i, get feature value vector

front 2s _iindividual characteristic value forms validity feature value vector

form noise characteristic value vector by remaining characteristic value

that is:

${\stackrel{\→}{\mathrm{\λ}}}_{i1}=[{\mathrm{\ρ}}_1,{\mathrm{\ρ}}_2,...{\mathrm{\ρ}}_{2s_i}],$

${\stackrel{\→}{\mathrm{\λ}}}_{i2}=[{\mathrm{\ρ}}_{2s_i+1},{\mathrm{\ρ}}_{2s_i+2},...,{\mathrm{\ρ}}_{2r{k}_i}];$

4e) according to step 4d), obtain characteristic quantity combination (s _i, k _i) characteristic of correspondence flow function value M (s _i, k _i):

$M(s_i,k_i)=\frac{N}{k_i}\left(\log \left(\frac{\sqrt{\∏{\stackrel{\→}{\mathrm{\λ}}}_{i1}}}{\∏{\stackrel{\→}{\mathrm{\λ}}}_{i2}}\right)\right),$

Wherein ∏ () means that vector element connects multiplication.

Step 5, the pre-estimation characteristic quantity, obtain new empty time-code set omega '.

5a) to every kind in characteristic quantity set (U, V) combination repeating step 4, obtain every kind of combination characteristic of correspondence flow function value, composition characteristic flow function value vector: Φ=[M (s ₁, k ₁), M (s ₂, k ₂) ... M (s _m, k _m) ...], wherein, m=1,2 ...;

5b) find out the element characteristic of correspondence amount combination of numerical value minimum in characteristic quantity functional value vector Φ

obtain new empty time-code set omega ', its all class symbol number in empty time-code set omega is

and the set that the pattern that block length is k forms, this set comprises one or more patterns, wherein

Step 6, write the sky time-code set omega that makes new advances ' in the symbolic number of each row of encoder matrix of j kind pattern, form symbolic number vector P _j, the encoder matrix of this j kind pattern and new empty time-code set omega ' in the encoder matrix of other patterns there is identical columns, but the coding symbol number that each row contains is not quite similar, j=1 wherein, 2...T, T be new empty time-code set omega ' the pattern number.

Step 7, the estimating part symbolic number.

7a) utilize incidence matrix R in step (3) and the block length k in step (5b), structure part correlation matrix V:

$\begin{array}{c}\begin{array}{cccc}V=[\tilde{R}\left(\mathrm{\β}\right)& \tilde{R}(k+\mathrm{\β})& \·\·\·& \tilde{R}\end{array}\end{array},((\frac{N}{k}-1)k+\mathrm{\β})],$

Wherein, β be new empty time-code set omega ' in the row mark of encoder matrix of pattern, β=1,2 ... k;

7b) the part covariance matrix of calculating section correlation matrix V: D=E[VV ^t], adopt the orthogonal diagonal factorization method to do Eigenvalues Decomposition to part covariance matrix D, by the characteristic value that obtains by descending, the component part feature value vector

ε wherein _γfor the characteristic value of part covariance matrix D, γ=1,2 ... 2rk;

7c) respectively will $n=1,n=2...n=\hat{s}$ Bring following formula into and calculate the likelihood function value

$L\left(n\right)=\begin{array}{c}-n(4r-2n+1)-\frac{N}{2k}\underset{\mathrm{\α}=1}{\overset{2n}{\mathrm{\Σ}}}\log \left({\mathrm{\ϵ}}_{\mathrm{\α}}\right)\\ -\frac{N}{k}(r-n)\log \left(\frac{1}{2(r-n)}\underset{\mathrm{\α}=2n+1}{\overset{2r}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{\mathrm{\α}}\right)\end{array},$

And then obtain likelihood function value vector

7d) get n corresponding to numerical value greatest member in likelihood function value vector L, as the part symbolic number.

Step 8, calculate apart from the decision value vector.

8a) repeating step (7), estimate the part symbolic number of each row of encoder matrix, obtains the parameter Estimation vector:

Ψ=[n (1), n (2) ... n (β) ...], the part symbolic number that wherein n (β) is encoder matrix β row;

8b) utilize the symbolic number vector P in step (6) _j, compute sign number vector P _jwith parameter Estimation vector Ψ apart from decision value θ _j=(P _j-Ψ) ²;

8c) repeating step (8b), obtain new empty time-code set omega ' in every kind of pattern and parameter Estimation vector Ψ apart from decision value, form apart from decision value vector Π=[θ ₁, θ ₂... θ _t].

Step 9, get the pattern corresponding apart from the element of numerical value minimum in decision value vector Π as judgement pattern, the blind identification of pattern while completing sky.

Effect of the present invention can further describe by following emulation.

Emulation 1: under three groups of different parameters, the recognition correct rate of diplomatic copy invention to empty time-code set omega hollow time-code, simulation result is as Fig. 3.Wherein: the solid line in Fig. 3 represents that transmitting antenna is 3, reception antenna is 8, send the system identification accuracy of sequence length N=1024, zone circle solid line in Fig. 3 represents that transmitting antenna is 3, and reception antenna is 6, sends the system identification accuracy of sequence length N=1024, dotted line in Fig. 3 represents that transmitting antenna is 3, reception antenna is 8, sends the system identification accuracy of sequence length N=512

As can be seen from Figure 3: solid line, on the zone circle solid line, illustrates that the quantity by improving reception antenna can improve the recognition correct rate of the present invention to empty time-code set omega; Solid line, on dotted line, illustrates that increasing the transmission sequence length can improve the recognition correct rate of the present invention to empty time-code set omega.

Emulation 2: with existing two kinds of blind-identification methods, empty time-code set omega is identified with the present invention.

Existing two kinds of blind-identification methods are:

Document [1V.Choqueuse, M.Marazin et al., Blind recognition of linear space time block codes:A likelihoodbased approach.IEEE Trans.Signal Processing, 58 (3), 2010, 1290-1299] the middle code parameters detection algorithm proposed, it according to maximum-likelihood criterion to all Candidate Set patterns, construct only relevant with coding parameter likelihood function, by comparing the likelihood function of different coding pattern, coding parameter is judged, and then draw the judgement pattern, in this emulation, brief note is the code parameters detection method.

Document [2V.Choqueuse, K.Yao et al., Blind recognition of linear space time block codes.IEEE Int.Conf.Acoust.Speech Signal Process, 2008,2833-2836] the middle Decision Classfication detection algorithm proposed, it is according to the otherness of the Frobenius norm of correlation matrix under different delay of different Space-Time Block Codings, adopt contrast step by step, the blind identification of realization to Space-Time Block Coding, in this emulation, brief note is the Decision Classfication detection method.

If emulation signal to noise ratio scope is-5dB～10dB, every 1000 Monte Carlo Experiments of 1dB emulation, each Monte Carlo Experiment sends identification successively by the empty time-code in empty time-code set omega, record the correct identification number of times under each signal to noise ratio, and then obtain the recognition correct rate under each signal to noise ratio, can identify the ratio of pattern duty time-code set omega, simulation result is as Fig. 4.Wherein solid line represents recognition correct rate of the present invention, and the zone circle solid line represents the recognition correct rate of code parameters detection method, and dotted line represents the recognition correct rate of Decision Classfication detection method.

As can be seen from Figure 4: solid line on zone circle solid line and dotted line, illustrates under same signal to noise ratio far away, the present invention to the recognition correct rate of empty time-code set omega far above existing two kinds of blind-identification methods.

From Fig. 4, it can also be seen that: the zone circle solid line finally levels off to 4/6, dotted line finally levels off to 5/6, and solid line finally levels off to 1,4/6 of the pattern duty time-code set omega that the description code parameter detection method can be identified, 5/6 of the pattern duty time-code set omega that the Decision Classfication detection method can be identified, and the present invention can all identify the pattern in empty time-code set omega, i.e. the present invention can identify more pattern than code parameters detection method and Decision Classfication detection method.

Claims (6)

1. the blind-identification method of pattern during based on partial sequence parameter detecting empty, comprise the steps:

1) receiving terminal receives by r root reception antenna the burst that length that transmitting terminal sends is N, obtains the reception signal matrix R' of r * N, N >=64 wherein, r >=2;

2) utilize all empty time-codes that need identification, form the pattern set omega, get combination (s, k) the constitutive characteristic duration set (U, V) that the class symbol of every kind of pattern in Ω is counted s and block length k, remember that i is combined as (s _i, k _i), i=1,2... Ζ, Ζ is number of combinations in characteristic quantity set (U, V);

3) will receive real part and the imaginary part parallel connection of signal matrix R', obtain incidence matrix R,

$R=\left[\begin{array}{c}\mathrm{Re}\left(R^{\′}\right)\\ \mathrm{Im}\left(R^{\′}\right)\end{array}\right]$

Wherein Re () means to get the real part computing, and Im () means to get imaginary-part operation;

4) calculated characteristics flow function value:

4a) for i characteristic quantity combination (s in characteristic quantity set (U, V) _i, k _i), structure grouping correlation matrix R _i:

$R_i=\left[\begin{array}{cccc}\tilde{R}\left(1\right)& \tilde{R}(k_i+2)& \·\·\·& \tilde{R}((\frac{N}{k_i}-1)k_i+1)\\ \tilde{R}\left(2\right)& \tilde{R}(k_i+2)& \·\·\·& \tilde{R}((\frac{N}{k_i}-1)k_i+2)\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \tilde{R}\left(k_i\right)& \tilde{R}\left({2k}_i\right)& \·\·\·& \tilde{R}\left(\left(\frac{N}{k_i}\right)k_i\right)\end{array}\right]$

Wherein R (τ) means the τ row of incidence matrix R, τ=1,2...N;

4b) calculate grouping correlation matrix R _ithe grouping covariance matrix: C _i=E[R _ir _i ^t], E[wherein] mean to ask expectation computing, () ^tmean the transposition computing;

4c) to grouping covariance matrix C _ido Eigenvalues Decomposition, by the characteristic value that obtains by descending, constitutive characteristic value vector wherein, ρ _ηfor grouping covariance matrix C _icharacteristic value, η=1,2 ... 2rk _i, get feature value vector front 2s _iindividual characteristic value forms validity feature value vector

form noise characteristic value vector by remaining characteristic value

?

${\stackrel{\→}{\mathrm{\λ}}}_{i1}=[{\mathrm{\ρ}}_1,{\mathrm{\ρ}}_2,...{\mathrm{\ρ}}_{2s_i}],{\stackrel{\→}{\mathrm{\λ}}}_{i2}=[{\mathrm{\ρ}}_{2s_i+1},{\mathrm{\ρ}}_{2s_i+2},...,{\mathrm{\ρ}}_{2r{k}_i}];$

4d) according to step 4c), obtain characteristic quantity combination (s _i, k _i) characteristic of correspondence flow function value M (s _i, k _i):

$M(s_i,k_i)=\frac{N}{k_i}\left(\log \left(\frac{\sqrt{\∏{\stackrel{\→}{\mathrm{\λ}}}_{i1}}}{\∏{\stackrel{\→}{\mathrm{\λ}}}_{i2}}\right)\right)$ Wherein ∏ () means that vector element connects multiplication;

5) pre-estimation characteristic quantity, obtain new empty time-code set omega ':

5a) to every kind in characteristic quantity set (U, V) combination repeating step 4, obtain every kind of combination characteristic of correspondence flow function value, composition characteristic flow function value vector: Φ=[M (s ₁, k ₁), M (s ₂, k ₂) ... M (s _m, k _m) ...], wherein, m=1,2 ... Ζ;

5b) find out the element characteristic of correspondence amount combination of numerical value minimum in characteristic quantity functional value vector Φ

obtain new empty time-code set omega ';

6) write the sky time-code set omega that makes new advances ' in the symbolic number of each row of encoder matrix of j kind pattern, form symbolic number vector P _j, j=1 wherein, 2...T, T be new empty time-code set omega ' the pattern number;

7) estimating part symbolic number:

7a) utilize R in step (3) and the block length k in step (5b), structure part correlation matrix V:

$\begin{array}{c}\begin{array}{cccc}V=[\tilde{R}\left(\mathrm{\β}\right)& \tilde{R}(k+\mathrm{\β})& \·\·\·& \tilde{R}\end{array}\end{array},((\frac{N}{k}-1)k+\mathrm{\β})],$

Wherein, β be new empty time-code set omega ' in the row mark of encoder matrix of pattern, β=1,2 ... k;

7b) the part covariance matrix of calculating section correlation matrix V: D=E[VV ^t], to the part covariance matrix, D does Eigenvalues Decomposition, by the characteristic value that obtains by descending, the component part feature value vector

ε wherein _γfor the characteristic value of part covariance matrix D, γ=1,2 ... 2rk;

7c) respectively by n=1,

bring following formula into and calculate likelihood function value L (1),

$L\left(n\right)=\begin{array}{c}-n(4r-2n+1)-\frac{N}{2k}\underset{\mathrm{\α}=1}{\overset{2n}{\mathrm{\Σ}}}\log \left({\mathrm{\ϵ}}_{\mathrm{\α}}\right)\\ -\frac{N}{k}(r-n)\log \left(\frac{1}{2(r-n)}\underset{\mathrm{\α}=2n+1}{\overset{2r}{\mathrm{\Σ}}}{\mathrm{\ϵ}}_{\mathrm{\α}}\right)\end{array}$

And then obtain likelihood function value vector

7d) getting n corresponding to numerical value greatest member in likelihood function value vector L is the part symbolic number;

8) calculate apart from the decision value vector:

8a) repeating step (7), estimate the part symbolic number of each row of encoder matrix, obtains the parameter Estimation vector

Ψ=[n (1), n (2) ... n (β) ...], the part symbolic number that n (β) is encoder matrix β row;

8b) utilize the symbolic number vector P in step (6) _j, compute sign number vector P _jwith parameter Estimation vector Ψ apart from decision value θ _j=(P _j-Ψ) ²;

8c) repeating step (8b), obtain new empty time-code set omega ' in every kind of pattern and parameter Estimation vector Ψ apart from decision value, form apart from decision value vector Π=[θ ₁, θ ₂... θ _t];

9) getting the pattern corresponding apart from the element of numerical value minimum in decision value vector Π is judgement pattern, the blind identification of pattern while completing sky.

2. pattern blind-identification method during based on partial sequence parameter detecting empty according to claim 1, it is characterized in that, empty time-code set omega in described step (2), comprise orthogonal space time packet, quasi-orthogonal space time block code and non-orthogonal space-time block.

3. pattern blind-identification method during based on partial sequence parameter detecting empty according to claim 1, is characterized in that the combination (s in described step (2) _i, k _i), a kind of pattern or multiple pattern in the corresponding empty time-code set omega of its each combination.

4. pattern blind-identification method during based on partial sequence parameter detecting empty according to claim 1, is characterized in that described step 4c) in Eigenvalues Decomposition, adopt the orthogonal diagonal factorization method, first at grouping covariance matrix C _iorthogonal matrix Q and transposed matrix thereof are multiplied by respectively in both sides, obtain characteristic value diagonal matrix Δ=Q ^tc _iq; Extract again grouping covariance matrix C from characteristic value diagonal matrix Δ _icharacteristic value.

5. pattern blind-identification method during based on partial sequence parameter detecting empty according to claim 1, is characterized in that described step 5b) in new empty time-code set omega ', be that all class symbol numbers are in empty time-code set omega

and the set that the pattern that block length is k forms, this set comprises one or more patterns, wherein $\hat{s}\≥2,k\≥1.$

6. pattern blind-identification method during based on partial sequence parameter detecting empty according to claim 1, it is characterized in that, new empty time-code set omega in described step (6) ' in the encoder matrix of j kind pattern, its with new empty time-code set omega ' in the encoder matrix of other patterns there is identical columns, but the coding symbol number that each row contains is not quite similar.

Images