CN1207859C - Method of time space solution for estimating wave diretion of maultiple paths signals in correlative CDMA and its device - Google Patents

Abstract

The present invention relates to a method and a device for estimating the wave destination direction of space-time de-correlation code division multiple access (CDMA) multi-path signals. Firstly, a matched filter bank connected to the back of an array antenna array element is utilized to separate user multi-path signals and calculate the output of the matched filter bank under an asynchronous multi-path channel; secondly, the space-time united de-correlation treatment is carried out for the output of the matched filter bank; thirdly, the covariance matrixes of the user multi-path signals are estimated by user decoupling multi-path signals. Fourthly, the characteristic of the covariance matrixes is decomposed and calculated; finally, the wave destination direction estimation of the user multi-path signals based on principal character vectors is realized. The device comprises two parts: a space-time de-correlation processor (A) for the user multi-path signals, a wave arrival direction estimator (B) for the user multi-path signals. The method and the device are very suitable for accurately estimating the wave destination direction of user multi-paths signals in array antenna CDMA systems by a space-time de-correlation detection technology.

Description

Decorrelation CDMA multipath method for estimating signal wave direction and device when empty

Technical field

The invention belongs to code division multiple access cdma cellular field of wireless communications.

Background technology

The cdma cellular communication technology is simple with its frequency planning, power system capacity is big, ability of anti-multipath is strong, good communication quality, electromagnetic interference are little etc., and characteristics demonstrate huge development potentiality, are the mainstream technologys of future mobile communications.If particularly in cdma system, use array antenna can improve capacity, spectrum efficiency, communication quality and the coverage of system significantly and high-precision wireless location service is provided.The direction of arrival of user multipaths signal estimates that the user radio location technology that wave beam to down link forms, reduces to disturb and estimates based on angle has important function in the array antenna CDMA system, is one of key technology of array antenna CDMA system.

In the past few decades, proposed many signal direction of arrival DOA algorithm for estimating, be commonly referred to traditional DOA algorithm for estimating based on aerial array.Traditional DOA algorithm for estimating such as multiple signal classification method MUSIC and require the array number of aerial array more than signal number by invariable rotary technology estimated signal parameter ESPRIT algorithm, and signal source spatially can not be overlapping, otherwise aerial array can't effectively be estimated the DOA of signal source.

The up channel of cdma system is generally asynchronous multipath channel, and user job is in identical frequency range.Tens users are arranged in the typical cellular sub-district usually, and each user's signal can produce many single sub path again, makes traditional DOA method of estimation based on aerial array can not directly apply in the array antenna CDMA system.

Summary of the invention

Technical problem solved by the invention is, proposes a kind of decorrelation CDMA multipath method for estimating signal wave direction and device when empty.This method can effectively be differentiated and the direction of arrival of estimating user multipath signal accurately.The accurate estimation of the array antenna CDMA system user multipath signal direction of arrival of decorrelation detection technique when the device based on said method that the present invention proposes is very suitable for adopting sky.

Decorrelation CDMA multipath method for estimating signal wave direction when technical scheme of the present invention is empty:

The output that at first utilizes the matched filter banks that connects after the array antenna array element to carry out the separation of user multipaths signal and calculate matched filter banks under the asynchronous multipath channel, secondly the space-time joint decorrelation being carried out in the output of matched filter banks handles, utilize the user to separate the covariance matrix of lotus root multipath output signal estimating user multipath signal then, carry out the calculating of covariance matrix feature decomposition afterwards again, realize at last estimating based on the user multipaths signal direction of arrival of covariance matrix feature decomposition principal eigenvector.

Its process is as follows:

At first user multipaths signal is separated and the output of calculating matched filter banks: utilize matched filter banks user multipaths signal effectively to be separated and utilize the output of the property calculation matched filter banks that is associated of transmission signals signature waveform under the asynchronous multipath channel based on matched filter banks.

Decorrelation was handled when next carried out sky: decorrelation was handled when the desired user multipath signal of matched filter banks output was carried out sky, eliminate the interference of other multipath signals, and the user multipaths signal decorrelation disturbed of other multipaths of being eliminated or separate lotus root output, the space-time joint decorrelation handles that be based on the space-time joint correlation matrix contrary realize.

Estimate covariance matrix then: ask Estimation of covariance matrix to eliminating the desired user decoupled multipaths signal that other multipaths disturb, obtain desired user decoupled multipaths signal Estimation of covariance matrix, the calculating utilization of user's decoupled multipaths signal covariance matrix be long process window method.

Afterwards, calculating user decoupled multipaths signal covariance matrix is carried out feature decomposition, obtain the principal eigenvector of corresponding eigenvalue of maximum.

At last, realize the estimation of user multipaths signal direction of arrival: after obtaining principal eigenvector, utilize multiple signal classification method estimating user multipath signal direction of arrival based on principal eigenvector.

Decorrelation CDMA multipath signal direction of arrival estimation unit when empty.This device comprises: decorrelation processor A and user multipaths signal direction of arrival estimator B two parts during user multipaths signal empty, the input of the input termination array antenna baseband sampling signal of decorrelation processor A when empty, the input of the output termination user multipaths signal direction of arrival estimator B of decorrelation processor A when empty.

Below to the invention in each composition discussed respectively.

1. decorrelation is handled when empty

A. the user multipaths signal based on matched filter banks separates and calculating

1) received signal of array

Investigate one and go up line asynchronous multipath channel array antenna CDMA system, suppose that mobile channel is the frequency selectivity slow fading channel, the correlation time of channel is much larger than symbol period.Certain cellular cell has K user launching bpsk signal by multipath channel separately in the supposing the system.Like this, k the N bit signal that the user launched can be expressed as:

$x_k\left(t\right)=A_k\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}{b}_k\left(n\right)c_k(t-n{T}_b),--k=1,...,K$

[formula 1]

In the formula, A _kThe amplitude of representing k subscriber signal, b _k(n) { 1 ,+1} represents the n bit signal of general transmission such as k user, c to ∈ _k(t) k user's of expression frequency spreading wave has

$c_k\left(t\right)=\underset{g=0}{\overset{G-1}{\mathrm{\Σ}}}{c}_{\mathrm{kg}}p(t-g{T}_c)$

[formula 2]

In the formula, c _Kg∈ 1, and+1} (g=0 ..., G-1) be its spreading code, p (t) is that width is T _cCut general pulse, T _bBe the bit interval time, G is defined as G=T _b/ T _cSpreading gain, its frequency spreading wave has normalized energy, promptly ${\∫}_0^{T_b}{\left|c_k\left(t\right)\right|}^2=1.$ Equally, suppose that the information bit that each user launches is independently, the information bit of different user also is independently.

Suppose that base station array antenna has M array element, customer mobile terminal adopts single antenna.Like this, the baseband multi-path channel between k user transmitter and base station receiver can represent that its impulse response vector is with the many output of single input SIMO channels

$h_k\left(t\right)=\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\α}}_{k,l}\left(t\right)a_{k,l}\left({\mathrm{\θ}}_{k,l}\left(t\right)\right)\mathrm{\δ}(t-{\mathrm{\τ}}_{k,l})$

[formula 3]

Wherein, L is each user's a multipath number, α _{K, l}And τ _{K, l}Be respectively the multiple decay and the time delay of k user l footpath signal, $a_{k,l}\left({\mathrm{\θ}}_{k,l}\left(t\right)\right)={[a_{k,l,1}\left({\mathrm{\θ}}_{k,l}\left(t\right)\right),...,a_{k,l,M}\left({\mathrm{\θ}}_{k,l}\left(t\right)\right)]}^T/\sqrt{M}$ Be that corresponding k user l footpath signal direction of arrival is θ _{K, l}(t) array vector.

Like this, base station array antenna receives total signal and is

$r\left(t\right)=\underset{k=0}{\overset{K}{\mathrm{\Σ}}}{x}_k\left(t\right)*h_k\left(t\right)+w\left(t\right)$

$=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{A}_k{b}_k\left(n\right)\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{\mathrm{\α}}_{k,l}{a}_{k,l,m}{s}_k(t-n{T}_b-{\mathrm{\τ}}_{k,l})+w\left(t\right)$

[formula 4]

Wherein, * represents convolution, μ _{K, l}=A _kα _{K, l}Be the complex magnitude of k user l footpath signal of being received, w (t) is that average is 0, and covariance matrix is σ ²I _MAdditivity from Gaussian noise vector, I _MIt is the unit matrix of M * M.

2) calculating of matched filter banks output

Use r _m(t) signal of m array element reception of expression array antenna.The signal phasor that receives from formula 4, the m array elements can be expressed as

$r_m\left(t\right)=\underset{n=0}{\overset{N-1}{\mathrm{\Σ}}}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{A}_k{b}_k\left(n\right)\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{c}_{k,l,m}{s}_k(t-n{T}_b-{\mathrm{\τ}}_{k,l})+w_m\left(t\right),m=1,...M$

[formula 5]

Wherein, c _{K, l}, m=α _{K, l}α _{K, l, m}It is the complex gain that comprises multiple decay and direction of arrival.

All users and multipath composition thereof can provide sufficient statistical information in the output of all array element matched filter banks of array antenna to detecting numerical chracter.At t ∈ [nT _b, (n+1) T _b) time interval sampling that is matched with last k the user l of array element m footpath frequency spreading wave is output as

$y_{k,l,m}^{\left(n\right)}={\∫}_{n{T}_b+{\mathrm{\τ}}_{k,l}}^{(n+1)T_b+{\mathrm{\τ}}_{k,l}}{r}_m\left(t\right)s_k(t-n{T}_b-{\mathrm{\τ}}_{k,l})\mathrm{dt}$

[formula 6]

The time delay expansion maximum of supposing Any user is no more than P mark space, and P is a positive integer, has

τ _{K, l}≤ PT _s, 1≤k≤K; 1≤l≤L [formula 7]

The received signal of this expression array antenna will be handled in length is the processing window of N=2P+1.Be the k user's of demodulation the n symbol, the user's matched filter output in corresponding each footpath of each bay is that formula 6 can be written as again

$y_{k,l,m}^{\left(n\right)}=\underset{i=-P}{\overset{P}{\mathrm{\Σ}}}\underset{k^{\′}=1}{\overset{K}{\mathrm{\Σ}}}{A}_{k^{\′}}{b}_{k^{\′}}(n+i)\underset{l^{\′}=1}{\overset{L}{\mathrm{\Σ}}}{c}_{k^{\′},l^{\′},m}{R}_{(k,l),(k^{\′},l^{\′}),m}\left(i\right)+w_{k,l,m}\left(n\right),l=1,\·\·\·,L;m=1,\·\·\·,M$

[formula 8]

Wherein, R _{(k, l), (k ' l '), m}(i) be the cross-correlation that array element m goes up user's inhibit signal waveform, it is defined as

$R_{(k,l)(k^{\′},l^{\′}),m}\left(i\right)={\∫}_{-\∞}^{+\∞}{s}_k(t-n{T}_b-{\mathrm{\τ}}_{k,l,m})s_{k^{\′}}(t-n{T}_b+i{T}_b-{\mathrm{\τ}}_{k^{\′},l^{\′},m})\mathrm{dt}$ [formula 9]

w _{K, l, m}(n) expression array element m goes up the k user l footpath in symbol n sampling noiset matched filtering output therebetween

$w_{k,l,m}\left(n\right)={\∫}_{n{T}_b+{\mathrm{\τ}}_{k,l}}^{(n+1)T_b+{\mathrm{\τ}}_{k,l}}{w}_m\left(t\right)s_k(t-n{T}_b-{\mathrm{\τ}}_{k,l})\mathrm{dt}$ [formula 10]

{ w _{K, l, m}(n) } for average be 0 multiple Gaussian random process, its covariance is

[formula 11]

Define t ∈ [nT now _b, (n+1) T _b] between symbolic vector

$b^{\left(n\right)}\stackrel{\mathrm{def}}{=}{(b_1^{\left(n\right)},b_2^{\left(n\right)},\·\·\·b_K^{\left(n\right)})}^T\∈{\mathrm{\Ξ}}^K$

Therefore, handling window scope preface received signal even at one can be expressed as

$b\stackrel{\mathrm{def}}{=}{({\left(b^{(n-p)}\right)}^T,\·\·\·,{\left(b^{\left(n\right)}\right)}^T,\·\·\·{\left(b^{(n+p)}\right)}^T)}^T\∈{\mathrm{\Ξ}}^{\mathrm{NK}}$

Similarly, the last sampling matched filtering output vector between symbol n of definition array element m is

$y_{k,m}^{\left(n\right)}={(y_{k,1,m}^{\left(n\right)},y_{k,2,m}^{\left(n\right)}\·\·\·y_{k,L,m}^{\left(n\right)})}^T\∈C^L$ [formula 12]

$y_m^{\left(n\right)}={({\left({\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">y</figure-callout>}}_{1,m}^{\left(n\right)}\right)}^T,{\left({\mathrm{<figure-callout\; id="2"\; label="y"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">y</figure-callout>}}_{2,m}^{\left(n\right)}\right)}^T,\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;\&CenterDot;,{\left(y_{K,m}^{\left(n\right)}\right)}^T)}^T\&Element;C^{\mathrm{KL}}$ [formula 13]

They are linked as a preface of handling the window scope

$y_m={({\left(y_m^{(n-p)}\right)}^T,\&CenterDot;\&CenterDot;\&CenterDot;,{\left(y_m^{\left(n\right)}\right)}^T,\&CenterDot;\&CenterDot;\&CenterDot;,{\left(y_m^{(n+p)}\right)}^T)}^T\&Element;C^{\mathrm{NKL}}$ [formula 14]

Like this, array element m goes up the symbolic representation that total matched filtering sampling output can matrix and is

y _m=R _mC _mA _mB+w _m[formula 15]

Wherein, cross-correlation matrix R _mCan be expressed as

$R_m=\left(\begin{array}{ccccc}{R}_m\left(0\right)& R_m(-1)& R_m(-2)& ...& 0_k\\ R_m\left(1\right)& R_m\left(0\right)& R_m(-1)& ...& 0_k\\ .& .& .& .& .\\ .& .& .& .& .\\ .& .& .& .& .\\ 0_k& 0_k& 0_k& ...& R_m\left(0\right)\end{array}\right)\&Element;C^{\mathrm{NKL}\&times;\mathrm{NKL}}$

$R_m\left(i\right)=\left(\begin{array}{cccc}{\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">\; <figure-callout\; id="1"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">R</figure-callout>\; </figure-callout>}}_{\mathrm{1,1},m}\left(i\right)& {\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">\; <figure-callout\; id="2"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">R</figure-callout>\; </figure-callout>}}_{\mathrm{1,2},m}\left(i\right)& ...& {\mathrm{<figure-callout\; id="1"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">R</figure-callout>}}_{1,K,m\left(i\right)}\\ {\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">\; <figure-callout\; id="1"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">R</figure-callout>\; </figure-callout>}}_{\mathrm{2,1},m}\left(i\right)& {\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">\; <figure-callout\; id="2"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">R</figure-callout>\; </figure-callout>}}_{\mathrm{2,2},m}\left(i\right)& ...& {\mathrm{<figure-callout\; id="2"\; label="R"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">R</figure-callout>}}_{2,K,m}\left(i\right)\\ .& .& .& .\\ .& .& .& .\\ .& .& .& .\\ R_{K,1,m}\left(i\right)& R_{K,2,m}\left(i\right)& ...& R_{K,K,m}\end{array}\right)\&Element;C^{\mathrm{KL}\&times;\mathrm{KL}}$

Wherein, matrix R _{K, k ', m (i)}∈ R ^{L * L}Be cross-correlation matrix, its element R _{(k, l), (k ', l '), m}(i) definition is identical with formula 9.

Because τ _{K, l}≤ PT _bWith [0, T _b] outer s _k(t) be 0, so have

R _m(i)＝0，|i|＞P

$R_m(-i)=R_m^T\left(i\right)$

Channel matrix C _mCan be expressed from the next

$C_m=\mathrm{diag}{(C_m^{(n-p)},C_m^{(n-p+1)},\&CenterDot;\&CenterDot;\&CenterDot;,C_m^{(n+p)})}^T\&Element;C^{\mathrm{NKL}\&times;\mathrm{NK}}$

$C_m^{\left(n\right)}=\mathrm{diag}({\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">c</figure-callout>}}_{1,m}^{\left(n\right)},{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">c</figure-callout>}}_{2,m}^{\left(n\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{K,m}^{\left(n\right)})\&Element;C^{\mathrm{KL}\&times;K}$

$c_{k,m}^{\left(n\right)}={(c_{k,1}^{\left(n\right)},c_{k,2}^{\left(n\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{k,L}^{\left(n\right)})}^T\&Element;C^L.$

The signal amplitude matrix can be expressed as

A _m＝diag(A _m，A _m，…，A _m)∈C ^NK×NK

A _m＝diag(A ₁，A ₂，…，A _K)∈C ^K×K

Use w _mExpression array element m goes up the noise vector of matched filter output, and then the preface of all sensing array elements of array antenna connects the matched filtering output vector and is

Y=RCAB+W [formula 16]

Wherein, y is defined as

$y={(y_1^T,y_2^T,\&CenterDot;\&CenterDot;\&CenterDot;,y_m^T)}^T\&Element;C^{\mathrm{MNKL}}$

And R=diag (R ₁, R ₂..., R _M) ∈ C ^{MNKL * MNKL}, C=diag (C ₁, C ₂..., C _M) ∈ C ^{MNKL * MNKL},

A＝didg(A ₁，A ₂，…，A _M)∈C ^MNK×MNK，B＝(b ^T，b ^T，…，b ^T)∈Ξ ^MNK，

$W=(w_1^T,w_2^T,\&CenterDot;\&CenterDot;\&CenterDot;,w_M^T)\&Element;C^{\mathrm{MNKL}}.$

The calculating of decorrelator output when B. empty

For the contrary output that is added on matched filter of lotus root user data information decorrelation detector with correlation matrix is conciliate in the interference of eliminating multipath signal.Because correlation matrix R is positive definite always in practice, its Linear Mapping has the character of Hermitian and positive definite, and therefore, they contrary always exists.Multiply by inverse matrix R at the two ends of formula 16 ^-1Decorrelation is output as when obtaining sky

Z=CAb+R ^-1W [formula 17]

Wherein,

$Z=(Z_1^T,Z_2^T,\&CenterDot;\&CenterDot;\&CenterDot;,Z_M^T)\&Element;C^{\mathrm{MNKL}}$

$Z_m={({\left(Z_m^{(n-p)}\right)}^T,\&CenterDot;\&CenterDot;\&CenterDot;,{\left(Z_m^{\left(n\right)}\right)}^T,\&CenterDot;\&CenterDot;\&CenterDot;{\left(Z_m^{(n+p)}\right)}^T)}^T\&Element;C^{\mathrm{NKL}}$ [formula 18]

$Z_m^{\left(n\right)}={({\left({\mathrm{<figure-callout\; id="1"\; label="Z"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">Z</figure-callout>}}_{1,m}^{\left(n\right)}\right)}^T,{\left({\mathrm{<figure-callout\; id="2"\; label="Z"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">Z</figure-callout>}}_{2,m}^{\left(n\right)}\right)}^T,\&CenterDot;\&CenterDot;\&CenterDot;{\left(Z_{K,m}^{\left(n\right)}\right)}^T)}^T\&Element;C^{\mathrm{KL}}$

[formula 19]

$Z_{k,m}^{\left(n\right)}={(Z_{k,1,m}^{\left(n\right)},Z_{k,2,m}^{\left(n\right)},\&CenterDot;\&CenterDot;\&CenterDot;,Z_{k,L,m}^{\left(n\right)})}^T\&Element;C^L$

[formula 20]

Obviously, Z _{K, m} ⁽ⁿ⁾Each element contain that array element m goes up single decoupled multipaths signal of user k and array element m goes up other the multipath interference signal that can separate lotus root and has been eliminated.Therefore, each multipath signal that can separate lotus root can be detected independently.

2, the user multipaths signal direction of arrival is estimated

A. the calculating of covariance matrix

For estimating the direction of arrival θ of k user l footpath signal _{K, l}Need at first estimate its pairing covariance matrix of separating the output of lotus root array.In practice, the exact value of covariance matrix is unavailable, must estimate by the data that receive.Typical in the case processing mode is exactly the employing estimation that forms true covariance matrix by Q continuous observation or sampling.For obtaining continuous Q observation or sampling, need be that Q the continuous processing window of N=2P+1 is handled to length, the adjacent processing window symbol that slides over each other.Can only open a long processing window when concrete the processing, be called the long process window, be Q observation of equidistant intercepting or sampling about reference point to handle window center then.Adopt this method can obtain the better estimation of covariance matrix.

Can estimate on m the array element that from formula 20 k user l footpath signal separate lotus root signal Z when q mark space empty _{K, l, m} ^(q)K user l footpath signal is separated the signal of lotus root signal on all array elements of array antenna when q mark space empty pile up, can obtain M * 1 vector

$Z_{k,l}^{\left(q\right)}={(Z_{k,l,1}^{\left(q\right)},Z_{k,l,2}^{\left(q\right)}\&CenterDot;\&CenterDot;\&CenterDot;,Z_{k,l,M}^{\left(q\right)})}^T$ [formula 21]

Separate being estimated as of lotus root signal covariance matrix when therefore, adopting directly empty of corresponding k the user l of long process window processing method

${\hat{R}}_{k,l}=\frac{1}{Q}\{Z_{k,l}^{(r-\frac{Q-1}{2})}{\left(Z_{k,l}^{(r-\frac{Q-1}{2})}\right)}^H+\&CenterDot;\&CenterDot;\&CenterDot;+Z_{k,l}^{\left(r\right)}{\left(Z_{k,l}^{\left(r\right)}\right)}^H+\&CenterDot;\&CenterDot;\&CenterDot;+Z_{k,l}^{(r+\frac{Q-1}{2})}{\left(Z_{k,l}^{(r+\frac{Q-1}{2})}\right)}^H\}$

[formula 22]

Wherein, Q is assumed to be odd number, and r is a reference point.

B. the feature decomposition of covariance matrix

To resulting sampling Estimation of covariance matrix Carry out feature decomposition, obtain covariance matrix R _{K, l}Principal eigenvector e _{1, k, l}Estimation ê _{1, k, l}

C. the user multipaths signal direction of arrival is estimated

At last, realize the estimation of user multipaths signal direction of arrival.At the estimation ê that obtains k user l footpath covariance matrix feature decomposition principal eigenvector of de-correlated signals when empty _{1, k, l}After, can realize k user l footpath signal direction of arrival θ _{K, l}Estimation

Concrete steps are:

At first, structure spatial spectrum function,

$P_{\mathrm{MU}}\left({\mathrm{\&theta;}}_{k,l}\right)=[1-{\left|\right|a_{k,l}^H\left({\mathrm{\&theta;}}_{i,j}\right){\hat{e}}_{1,k.l}\left|\right|}^2{]}^{-1}$ [formula 23]

Wherein, reciprocity pitch arrays antenna has

$a_{k,l}^H\left({\mathrm{\&theta;}}_{k,l}\right)=[1,e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">j</figure-callout>}\frac{2\mathrm{\&pi;d}}{\mathrm{\&lambda;}}\sin \left({\mathrm{\&theta;}}_{k,l}\right)},...,e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="C0314968800132.png,C0314968800142.png"\; state="\{\{state\}\}">j</figure-callout>}\frac{2\mathrm{\&pi;d}}{\mathrm{\&lambda;}}(M-1)\sin \left({\mathrm{\&theta;}}_{k,l}\right)}]/\sqrt{M}$

Then, search volume spectral function P _MUThe spectrum peak direction of arrival that just can obtain corresponding k user l footpath signal estimate.The direction of arrival θ of corresponding k user l footpath signal _{K, l}Be estimated as

${\widehat{\mathrm{\&theta;}}}_{k,l}=\arg \underset{{\mathrm{\&theta;}}_{k,l}}{\max }\left\{\right[1-{\left|\right|a_{k,l}^H\left({\mathrm{\&theta;}}_{k,l}\right){\hat{e}}_{1,k,l}\left|\right|}^2{]}^{-1}\}$ [formula 24]

Beneficial effect of the present invention: at first since the estimation of user multipaths signal direction of arrival be based on user multipaths signal separate that lotus root output realizes, eliminated the interference of other multipath signals, therefore, the advantage of this method is exactly that estimation performance is good.Secondly, this method does not require the number of signals of the array number of array antenna more than user and multipath thereof; Simultaneously, this method does not need to detect the signal source number of user and multipath thereof, and these are that traditional multiple signal classification method is necessary.In addition, owing to only need the single spatial spectrum peak of search, therefore, corresponding estimated result is very reliable.These all make the present invention be very suitable for the demanding situation of user's multipath direction of arrival estimated accuracy, particularly when the array antenna CDMA system adopts the multiuser detection that decorrelation is handled when empty.

Description of drawings

Fig. 1 is the device general diagram.

Fig. 2 is a matched filter banks schematic diagram after each array element.

Fig. 3-the 6th, the performance legend of Wave arrival direction estimating method that the present invention carries.

Embodiment

The present invention is described in further detail below in conjunction with accompanying drawing.

The method according to this invention is applicable to the CDMA mobile communication system of any employing array antenna.

Fig. 1 has provided this method and has been applied to the device detailed structure schematic diagram that the array antenna direct sequence spread spectrum codes is divided multiple access DS-CDMA mobile communication system.

This device comprises: decorrelation processor A and user multipaths signal direction of arrival estimator B two parts are formed during user multipaths signal empty, the input of the input termination baseband sampling signal of decorrelation processor A when empty, the input of the output termination user multipaths signal direction of arrival estimator B of decorrelation processor A when empty.

The decorrelation processor A comprises that user multipaths signal based on matched filter banks separates and decorrelation processing module A102 when computing module A101 and sky when empty, and user multipaths signal direction of arrival estimator B comprises user's decoupled multipaths signal correlation matrix computing module B101, covariance matrix feature decomposition module B102, user multipaths signal direction of arrival estimation module B103.Wherein separate and computing module A101, decorrelation processing module A102, user's decoupled multipaths signal correlation matrix computing module B101, covariance matrix feature decomposition module B102, user multipaths signal direction of arrival estimation module B103 are connected in series successively in proper order when empty based on the user multipaths signal of matched filter banks.

Concrete signal processing is as follows: at first, the base-band analog signal that each array element of array antenna is received into becomes digital signal after modulus A/D conversion, the user multipaths signal that enters then based on matched filter banks separates and calculator modules A101, this module realizes the separation of user multipaths signal by matched filter banks according to the different delay of user multipaths signal, while connects matched filtering output vector y, decorrelation processing module A102 when its result offers sky according to the preface that formula 16 calculates all sensing array elements of array antenna.When sky, in the decorrelation processing module preface is connected matched filtering output vector y and carry out space-time joint decorrelation processing, eliminate the interference of other multipath signals, and the lotus root of separating of the desired user multipath signal of other multipaths interference that are eliminated is exported Z _{K, l} ^(q)In user multipaths signal correlator block B101, export Z according to the lotus root of separating of 22 pairs of desired user multipath signals of formula _{K, l} ^(q)Carry out correlation computations, obtain the desired user multipath signal and separate lotus root and export true covariance matrix R _{K, l}Employing estimate Covariance matrix feature decomposition module B102 receives the output of user multipaths signal correlator block B101 and the employing of desired user decoupled multipaths signal covariance matrix is estimated

Matrix carries out feature decomposition, obtains principal eigenvector e _{1, k, l}Estimation ê _{1, k, l}, and with ê _{1, k, l}Be input to user multipaths signal direction of arrival estimator module B103.User multipaths signal direction of arrival estimator module B103 receives ê _{1, k, l}After, obtain desired user multipath signal direction of arrival θ _{K, l}Estimation

Fig. 3-6 has provided based on array antenna DS-CDMA system user multipath signal direction of arrival estimation under the asynchronous multipath channel of space-time joint decorrelation output the result of signal to noise ratio snr.Number of users, spread processing gain and array structure have been provided respectively to estimated Effect on Performance.For estimate estimation performance to 50 times independently simulation result carried out on average, and adopted standard error SD as evaluation index, it is defined as

$\mathrm{SD}=\sqrt{\frac{1}{U}\underset{u=1}{\overset{U}{\mathrm{\&Sigma;}}}{({\widehat{\mathrm{\&theta;}}}_{k,l}-{\mathrm{\&theta;}}_{k,l})}^2}$ [formula 25]

Wherein, U is the number of times of independent experiment.

Concrete simulated conditions is as follows: the employing array element distance is that the evenly equally spaced 5 array element linear array antennas in half wavelength lambda/2 receive the BPSK multipath signal.Be without loss of generality, suppose that channel is two footpath rayleigh fading channels, each directly has identical energy, and the relative delay of multipath signal is in a mark space.Spreading code is that the gain of picked at random is 16 frequency expansion sequence, obtains Estimation of covariance matrix with 20 times observations or sampling length Might as well suppose to have respectively 3 and 6 users to be randomly dispersed in the sub-district, user's footpath direction of arrival that expectation is estimated is 27 °.

As seen from Figure 3, method proposed by the invention is estimated very effective, accurate to the direction of arrival of user multipaths signal.And along with the increase estimation performance of number of users almost is identical.This is can eliminate multiple access interference MAI because decorrelation is handled on principle.

Decorrelation output and direct DOA estimated performance when Fig. 4 has compared based on sky based on matched filter banks output.Decorrelation output estimation performance is much more accurate than directly exporting estimation performance based on matched filter banks when Fig. 4 can be clear that based on sky, this also is can eliminate multiple access interference MAI because decorrelation is handled on principle, is inevitable and the algorithm multiple access that directly output is estimated based on matched filter banks is disturbed MAI.When signal to noise ratio increased, decorrelation output estimation performance can also be further improved during based on sky, directly but can not get improving owing to multiple access disturbs the corresponding increase of MAI based on matched filter banks output estimation performance on the contrary.

Fig. 5 has shown the influence of spread processing gain to estimated performance.Even the result of Fig. 5 shows that estimation performance is very accurate under the gain of low spread processing, therefore, along with the improvement of the increase estimated performance of spread processing gain will can be not clearly.

Fig. 6 has shown the influence of array structure to estimated performance.Can observe from Fig. 6, when strengthen between array element apart from the time, estimation performance can obtain improving more significantly.This be since in this case the resolving power of array antenna be enhanced.

It is worthy of note, even when two of different user or many strips directly arrive the direction of arrival of aerial array when identical, because it is that each CDMA user has been assigned with unique spreading code that method has been utilized the intrinsic characteristic of CDMA signal, therefore the direction of arrival in each son footpath still can effectively be estimated in this case.

Claims (2)

1. the CDMA multipath of decorrelation when empty method for estimating signal wave direction is characterized in that:

The output that at first utilizes the matched filter banks that connects after the array antenna array element to carry out the separation of user multipaths signal and calculate matched filter banks under the asynchronous multipath channel, the calculating of matched filter banks output are to utilize that the characteristic that is associated of transmission signals signature waveform realizes under the asynchronous multipath channel;

Secondly the space-time joint decorrelation is carried out in the output of matched filter banks and handled, the space-time joint decorrelation handles that being based on inverts to the space-time joint correlation matrix realizes;

Utilize the covariance matrix of user's decoupled multipaths signal estimating user multipath signal then, user's decoupled multipaths signal Estimation of covariance matrix utilizes long process window method to realize;

Again covariance matrix is carried out the calculating of feature decomposition afterwards, obtain the estimation of principal eigenvector;

Utilize principal eigenvector estimating user multipath signal direction of arrival at last based on the covariance matrix feature decomposition.

2. the CDMA multipath of decorrelation when empty signal direction of arrival estimation unit is characterized in that:

This device during by user multipaths signal empty decorrelation processor (A) and user multipaths signal direction of arrival estimator (B) two parts form:

Decorrelation processor (A) comprises that user multipaths signal based on matched filter banks separates and decorrelation processing module (A102) when computing module (A101) and sky when empty;

User multipaths signal direction of arrival estimator (B) comprises user multipaths signal correlation matrix computing module (B101), covariance matrix feature decomposition module (B102), user multipaths signal direction of arrival estimation module (B103);

User multipaths signal based on matched filter banks separates the output of finishing separating of user multipaths signal and calculating matched filter banks under the asynchronous multipath channel with computing module (A101), and the calculating of matched filter banks output is the characteristic realization that is associated that utilizes transmission signals signature waveform under the asynchronous multipath channel;

When empty decorrelation processing module (A102) is carried out the space-time joint decorrelation to the output of matched filter banks and is handled, and the space-time joint decorrelation handles that being based on inverts to the space-time joint correlation matrix realizes;

User multipaths signal correlation matrix computing module (B101) utilizes the covariance matrix of user's decoupled multipaths signal estimating user multipath signal, and user's decoupled multipaths signal Estimation of covariance matrix utilizes long process window method to realize;

Covariance matrix feature decomposition module (B102) is carried out the calculating of feature decomposition to covariance matrix, obtains the estimation of principal eigenvector;

User multipaths signal direction of arrival estimator module (B103) is utilized based on principal eigenvector estimating user multipath signal direction of arrival;

Wherein separate and computing module (A101), decorrelation processing module (A102), user multipaths signal correlation matrix computing module (B101), covariance matrix feature decomposition module (B102), user multipaths signal direction of arrival estimation module (B103) are connected in series successively in proper order when empty based on the user multipaths signal of matched filter banks.

Images