CN1388655A - Wave arrival direction estimating method based on route selection - Google Patents

Abstract

The present invention is one wave arrival direction estimating method based on route selection. The method uses the statistic signal processing measures and route selection result, estimates data via combining several channels and combines the channel estimation data of several time instants. The method has less noise influence, raised S/N ratio of wave arrival direction estimating data, raised correctness and stability of wave arrival direction estimation, and thus improved radio communication receiving performance.

Description

A kind of Wave arrival direction estimating method based on Path selection

The present invention relates to the reception technique in the wireless communication system receiver, relate in particular to direction of arrival (DOA) estimation technique of wireless communication system upward signal.

At present, Chang Yong DOA method of estimation is rapid fourier change (FFT) method.The advantage of this method is to realize simply can't determining signal number and the bigger problem of estimation error variance but exist, and because the FFT method is not a kind of statistical method, its performance is subjected to The noise very big, so the fluctuation of estimated value is very big.In wireless communication system, adopt said method to carry out the estimation of the direction of arrival of signal, can not utilize received signal in a period of time to reduce the error of estimated result well.If utilize all channel evaluation data that obtain based on pilot code to carry out the estimation of direction of arrival, then, cause the reduction of actual signal to noise ratio when handling the most at last because the energy of signal is very low in some channel estimating.

Problem at above-mentioned prior art exists the purpose of this invention is to provide a kind of direction of arrival estimation approach based on Path selection, uses this method can improve the signal to noise ratio of direction of arrival data estimator, thereby improves the accuracy and the stability of estimated result.

For achieving the above object, the technical solution used in the present invention is: a kind of Wave arrival direction estimating method based on Path selection, and this method may further comprise the steps:

(1) judge that each array element of antenna is equidistant linear the placement or equidistant being placed on the annulus,, change step (3) and continue operation if equidistant linearity is placed, otherwise the following operation of continuation;

(2) for loop aerial, at channel estimating vector y ₀(m), utilize the angle of pitch of direction of arrival of signal known or that estimated, use following formula y (m)=Ty ₀(m), m ∈ ∧) do linear transformation, obtain equivalent signal column vector y (m);

(3) definite number of parameters N that will estimate _p

(4) utilize equivalent signal column vector y (m) structure dimensionality reduction signal vector z _i(m), i=1,2 ..., P-N _P+ 1, according to formula $R_{\mathrm{SS}}\left(m\right)=\frac{1}{P-N_p+1}\underset{i=1}{\overset{P-N_p+1}{\mathrm{\Σ}}}{z}_i\left(m\right)z_i{\left(m\right)}^H$ Calculate instantaneous correlation matrix R _SS(m) estimation;

(5) rearrange the result that above-mentioned steps (4) calculates, according to formula R (l)=λ _fR (l-1)+R _SS(l) calculate correlation matrix R (l), obtain the parametric polynomial vector that needs $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T$ Estimation;

(6) the polynomial parameters value w that vector comprised that utilizes previous step to obtain suddenly ₁, w ₂,

, be constructed as follows described equation:

,, obtain phase angle φ by finding the solution of this equation ₁, φ ₂,

(7) for loop aerial, the direction of arrival of signal azimuth is estimated as φ ₁, φ ₂,

For linear antenna, through type ${\mathrm{\θ}}_i={\sin }^{-1}\left(\frac{\mathrm{\λ}{\mathrm{\φ}}_i}{2\mathrm{\πd}}\right)$ , i=1,2,

Obtain required estimated signals arrival direction angle.

Above the described parametric polynomial vector that needs of obtaining $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T$ Estimation obtain by following alternative manner:

If the initial condition of this method is: w (0), residual vector g (0), gradient vector p (1)=g (0), l=1, calculate successively and upgrade step-length: $\mathrm{\α}\left(l\right)=\mathrm{\η}\frac{p{\left(l\right)}^{T}g(l-1)}{p{\left(l\right)}^{T}R\left(l\right)p\left(l\right)},({\mathrm{\λ}}_f-0.5)\≤\mathrm{\η}\≤{\mathrm{\λ}}_f$ Desired parameters multinomial vector update calculation is:

W (l)=w (l-1)+α (l) p (l) residual vector more new formula is:

G (l)=λ _fG (l-1)-a (l) R (l) p (l)-R _SS(l) w (l-1) gradient adjustment step size computation is as the following formula: $\mathrm{\β}\left(l\right)=\frac{{(g\left(l\right)-g(l-1))}^{T}g\left(l\right)}{g{(l-1)}^{T}g(l-1)}$ The gradient vector update calculation:

P (l)=g (l)+β (l) p (l) obtains parameters needed multinomial vector by iterative computation repeatedly $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T.$

The technical scheme that is adopted by the invention described above as can be seen, because the present invention adopts the method for utilizing statistical signal to handle, utilize the result of Path selection simultaneously, in conjunction with many group channel evaluation data, the channel evaluation data in a plurality of moment is combined well, avoided the easy shortcoming affected by noise of FFT method commonly used, improve the signal to noise ratio of direction of arrival data estimator and improved correctness and stability that the signal direction of arrival is estimated, thereby can improve the performance that radio communication receives.

The present invention is described in further detail below in conjunction with drawings and Examples.

Fig. 1 is the schematic flow sheet of the inventive method.

In general, in the actual wireless communication system, it is more slowly that the direction of arrival of signal changes, if adopt the FFT method, then can only use transform method to each channel evaluation data constantly obtains respectively to estimate, and can't utilize these information that slowly change to improve estimated result, so estimated result is affected by noise bigger.The present invention adopts the statistical signal processing method to address the above problem.The applied statistics signal processing need be separated signal better from noise, therefore need to use earlier routing resource, for example, use " broadband directly expands-code division multiple access (DS-CDMA) mobile communication in merge machine " (S Fukumoto based on the RAKE of matched filter, M Sawahashi, and FAdachi. " Matched filter-based RAKE combiner for wideband DS-CDMAmobile radio ", IEICE Trans Commun, Vol.E81-B, No.7, July 1998, pp.1384～1391) method that provides in the literary composition, to comprise the low channel evaluation data of signal energy and at first separate and give up, guarantee that the data that actual treatment adopts have enough signal to noise ratios.

Specifically, method of the present invention is the processing of the baseband signal that receives at each array element of antenna, obtains the channel estimating on each array element of unique user by frequency pilot sign, with symbol h ⁽ⁿ⁾(m) expression, n=1,2 ... √ N represents different array elements, m=1, and 2 ... W represents different samplings constantly, comprises the information of signal direction of arrival in these channel evaluation data, need integrate and use the estimation that certain method of estimation obtains needed signal direction of arrival.

If symbol h ⁽ⁿ⁾(m) channel evaluation data of expression through remaining after the Path selection, n=1,2 ..., N remains unchanged, and m ∈ ∧ represents the sampling instant with enough high s/n ratios that Path selection keeps.

At the antenna pattern difference, the present invention's step thereafter also will adopt different processing.Each array element equidistantly distributes the antenna on annulus, promptly the received signal of loop aerial with respect to each array element linear isometry from the antenna of placing, i.e. the received signal of linear antenna, need carry out following processing in advance to signal:

The annular radii of supposing loop aerial is r, carrier wavelength lambda, the arriving signal angle of pitch can obtain by estimation, now supposes this pitching angle theta ∈ [0, pi/2], make M for being not more than min{2 π r/ λ, the maximum integer of N/2}, min{a, a is got in the b} representative, less number among the b, ξ=2 π r sin θ/λ; $T=\mathrm{diag}[j^{-M}{J}_M^{-1}\left(\mathrm{\ζ}\right),...,j^{-1}{J}_1^{-1}\left(\mathrm{\ζ}\right),J_0^{-1}\left(\mathrm{\ζ}\right),j^{-1}{J}_1^{-1}\left(\mathrm{\ζ}\right),...,j^{-M}{J}_M^{-1}\left(\mathrm{\ζ}\right)\}{V}^H,$ Wherein, j=

J _mBe that exponent number is the first kind Bei Sheer function of m (ξ), on behalf of main diagonal element, diag{a, b} be respectively a, the diagonal matrix of b, V ^HThe conjugate transpose of matrix V is got in representative.Then the matrix of (2M+1) * N is:

The preliminary treatment that the loop aerial received signal is done is exactly: the channel estimating column vector y of N * 1 that antenna is received ₀(m) do as down conversion

y(m)＝Ty ₀(m)，m∈∧ (1)

In the following formula, y ₀(m)=[h ⁽¹⁾(m) h ^(N)(m)] ^T, subscript T represents matrix transpose operation.The element number of equivalent signal column vector y (m) after the conversion is P=2M+1, defines y (m)=[x here ⁽¹⁾(m) ... √ x ^(P)(m)] ^T

Consider above-mentioned conversion, suppose that pending data are x ⁽ⁿ⁾(m), n=1,2 ..., √ P, m ∈ ∧.For loop aerial, P's is described as defined above; For linear antenna, P=N, x ⁽ⁿ⁾(m)=h ⁽ⁿ⁾(m).In statistical signal is handled, need determine the number of parameter to be estimated, because known method how to determine number of parameters above-mentioned to be estimated has been arranged at present, therefore suppose that now this number is several N that choose in advance _p, this parameter can be determined according to system's actual conditions in the reality, also can use document " modern signal processing " (open the prominent personage, publishing house of Tsing-Hua University, May nineteen ninety-five front page) in the parametric techniques such as amount of information criterion method that propose determine.By y (m)=[x ⁽¹⁾(m) ... √ x ^(P)(m)] ^TConstruct some new dimensionality reduction signal vectors, order: $z_1\left(m\right)=[x^{\left(1\right)}\left(m\right)...x^{\left(N_p\right)}\left(m\right){]}^T$ $z_2\left(m\right)=[x^{\left(2\right)}\left(m\right)...{\∋x}^{(N_p+1)}\left(m\right){]}^T$  $Z_{P-N_p+1}\left(m\right)=[x^{(P-N_p+1)}\left(m\right)...x^{\left(P\right)}\left(m\right){]}^T$ Calculate instantaneous correlation matrix according to following formula: $R_{\mathrm{SS}}\left(m\right)=\frac{1}{P-N_p+1}\underset{i=1}{\overset{P-N_p+1}{\mathrm{\Σ}}}{z}_i\left(m\right)z_i{\left(m\right)}^H---\left(2\right)$ Rearrange and make R _SS(l)=R _SS(ml), l=1,2 ..., and satisfy l ₁＜l ₂, then Calculate

R(l)＝λ _fR(l-1)+R _SS(l) (3)

Wherein, forgetting factor λ in the following formula _fSatisfy 0＜λ _f≤ 1.Obtaining correlation matrix R (l) and instantaneous correlation matrix R _SS(l) after the result of calculation, can calculate the N that needs by the following method _p* 1 parametric polynomial vector w (l):

If the initial condition of this method is: w (0), residual vector g (0), gradient vector p (1)=g (0), l=1, calculate successively and upgrade step-length: $\mathrm{\α}\left(l\right)=\mathrm{\η}\frac{p{\left(l\right)}^{T}g(l-1)}{p{\left(l\right)}^{T}R\left(l\right)p\left(l\right)},({\mathrm{\λ}}_f-0.5)\≤\mathrm{\η}\≤{\mathrm{\λ}}_f---\left(4\right)$ Desired parameters multinomial vector update calculation is:

W (l)=w (l-1)+α (l) p (l) (5) residual vector more new formula is:

G (l)=λ _fG (l-1)-a (l) R (l) p (l)-R _SS(l) w (l-1) (6) gradient adjustment step size computation is as the following formula: $\mathrm{\β}\left(l\right)=\frac{{(g\left(l\right)-g(l-1))}^{T}g\left(l\right)}{g{(l-1)}^{T}g(l-1)}---\left(7\right)$ The gradient vector update calculation:

p(l)＝g(l)+β(l)p(l) (8)

By iterative computation repeatedly, can obtain parameters needed multinomial vector $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T.$

Pass through the estimation of the vectorial picked up signal direction of arrival of being tried to achieve at last.Utilize the N among the parametric polynomial vector w (l) _pIndividual polynomial parameters value w ₁, w ₂...,

Formation is about the equation of variable z: $w_1{z}^{N_p-1}+w_2{z}^{N_p-2}+...+w_{N_p}=0---\left(9\right)$

Find the solution this equation and obtain N _p-1 root, the phase angle of these roots is respectively φ ₁, φ ₂...,

For loop aerial, Here it is needs the azimuth of estimated signals arrival direction, and unit is a radian; For linear antenna, required estimated signals arrival direction angle is estimated and can be calculated by following formula: ${\mathrm{\θ}}_i={\sin }^{-1}\left(\frac{{\mathrm{\λ\φ}}_i}{2\mathrm{\πd}}\right),i=\mathrm{1,2},...,N_p-1---\left(10\right)$ λ is carrier wavelength equally, and d is the distance between the adjacent array element, sin ^-1(.) represents arcsin function.

Being applied to TD-SCDMA (TD SDMA) system below by the present invention is described further the present invention.

In the TD-SCDMA system, at first utilize the estimation that is used for direction of arrival through the channel estimating signal that obtains after the joint-detection.What suppose that system in this example adopts is loop aerial, therefore need do as shown in the formula y (m)=Ty at above-mentioned estimated signal ₀(m), the described linear transformation of m ∈ ∧.In this example, specific implementation process of the present invention is with reference to figure 1.In step 1, judge that each array element of antenna is equidistant linear the placement or equidistant being placed on the annulus, because an antenna array element is not equidistant linear the placement, therefore continues following operation; In step 2, to channel estimating signal, utilize the angle of pitch of direction of arrival of signal known or that estimated through obtaining after the joint-detection, use following formula y (m)=Ty ₀(m), m ∈ ∧) do linear transformation, supposing to obtain the equivalent signal column vector is x ⁽ⁿ⁾(m), n=1 wherein, 2 ..., 7, m ∈ 2,3,5,7} represents the sampling instant with enough high s/n ratios that Path selection keeps, i.e. and P=7, ∧=2,3,5,7}.In step 3, determine the number of parameters N that will estimate _p, need two main directly direction of arrival of signal of estimating user in this hypothesis, then can determine N in the calculating of back _p=3.In step 4, utilize the equivalent signal column vector to be x ⁽ⁿ⁾(m) the corresponding dimensionality reduction signal vector z of structure _i(m), i=1,2 ..., 5, according to formula $R_{\mathrm{SS}}\left(m\right)=\frac{1}{P-N_p+1}\underset{i=1}{\overset{P-N_p+1}{\mathrm{\Σ}}}{z}_i\left(m\right)z_i{\left(m\right)}^H$ Estimate instantaneous correlation matrix R _SS(m).In step 5, press corresponding relation n=2 → l=1, n=3 → l=2, n=5 → l=3, n=7 → l=4 rearranges R _SS(m), obtain R _SS(l), l=1,2,3,4, according to formula R (l)=λ _fR (l-1)+R _SS(l) calculate correlation matrix R (l), R (0) can be set to null matrix, perhaps is set to give the correlation matrix estimation value that obtains in this user's the calculating of previous time slot in actual allocated, obtains the parametric polynomial vector that needs $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T$ Estimation; At parametric polynomial vector w (l)=[w ₁w ₂w ₃] ^TThe iteration estimating step in, w (0)=[100] can be set ^T, g (0)=[100] ^T,, be set to the estimated value that in the calculating of actual allocated, obtains to this user's previous time slot perhaps as correlation matrix.Because instantaneous correlation matrix R _SS(l) order is not 1, in the therefore actual iterative computation at each l, upgrade step-length, the parametric polynomial vector upgrades, residual vector upgrades, gradient is adjusted step-length, gradient vector is upgraded calculating also can iteration repeatedly.Concrete method of estimation is as follows:

If the initial condition of this method is: w (0)=[100] ^T, residual vector g (0)=[100] ^T, gradient vector p (1)=g (0), l=1, calculate successively and upgrade step-length: $\mathrm{\α}\left(l\right)=\mathrm{\η}\frac{p{\left(l\right)}^{T}g(l-1)}{p{\left(l\right)}^{T}R\left(l\right)p\left(l\right)},({\mathrm{\λ}}_f-0.5)\≤\mathrm{\η}\≤{\mathrm{\λ}}_f$ Desired parameters multinomial vector update calculation is:

W (l)=w (l-1)+α (l) p (l) residual vector more new formula is:

G (l)=λ _fG (l-1)-α (l) R (l) p (l)-R _SS(l) w (l-1) gradient adjustment step size computation is as the following formula: $\mathrm{\β}\left(l\right)=\frac{{(g\left(l\right)-g(l-1))}^{T}g\left(l\right)}{g{(l-1)}^{T}g(l-1)}$ The gradient vector update calculation:

p(l)＝g(l)+β(l)p(l)

After the estimation that obtains w (l), when the norm of residual vector g (l) enough little, such as less than predefined threshold value ε=0.01 o'clock, just can be in step 6, the polynomial parameters value w that vector comprised that utilizes previous step to obtain suddenly ₁, w ₂,

, be constructed as follows described equation: $w_1{z}^{N_p-1}+w_2{z}^{N_p-2}+...\▿+w_{N_p}=0,$

By finding the solution of this equation, obtain phase angle φ ₁, φ ₂,

Because what adopt in this example is loop antenna, so directly confirm φ in step 6 ₁, φ ₂, Be signal arrival bearing angle.If system adopts linear antenna to carry out the signal reception in this example, then to pass through formula in this step ${\mathrm{\θ}}_i={\sin }^{-1}\left(\frac{{\mathrm{\λ\φ}}_i}{2\mathrm{\πd}}\right)$ , i=1,2 ..., N _p-1 obtains required estimated signals direction of arrival.

In a word, the present invention is applicable to and need carries out the mobile communication system that DOA estimates, can provide the less DOA of variance to estimate.

Claims (4)

1, a kind of Wave arrival direction estimating method based on Path selection is characterized in that, this method may further comprise the steps:

(1) judge that each array element of antenna is equidistant linear the placement or equidistant being placed on the annulus,, change step (3) and continue operation if equidistant linearity is placed, otherwise the following operation of continuation;

(2) for loop aerial, at channel estimating vector y ₀(m), utilize the angle of pitch of direction of arrival of signal known or that estimated, use following formula y (m)=Ty ₀(m), m ∈ ∧) do linear transformation, obtain equivalent signal column vector y (m);

(3) definite number of parameters N that will estimate _p

(4) utilize equivalent signal column vector y (m) structure dimensionality reduction signal vector z _i(m), i=1,2 ..., P-N _P+ 1, according to formula $R_{\mathrm{SS}}\left(m\right)=\frac{1}{P-N_p+1}\underset{i=1}{\overset{P-N_p+1}{\mathrm{\Σ}}}{z}_i\left(m\right)z_i{\left(m\right)}^H$ Calculate instantaneous correlation matrix R _SS(m) estimation;

(5) rearrange the result that above-mentioned steps (4) calculates, according to formula R (l)=λ _fR (l-1)+R _SS(l) calculate correlation matrix R (l), obtain the parametric polynomial vector that needs $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T$ Estimation;

(6) the polynomial parameters value w that vector comprised that utilizes previous step to obtain suddenly ₁, w ₂..., , be constructed as follows described equation: $w_1{z}^{N_p-1}+w_2{z}^{N_p-2}+...+w_{N_p}=0$ ,, obtain phase angle φ by finding the solution of this equation ₁, φ ₂...,

(7) for loop aerial, the direction of arrival of signal azimuth is estimated as φ ₁, φ ₂...,

For linear antenna, through type ${\mathrm{\θ}}_i={\sin }^{-1}\left(\frac{\mathrm{\λ}{\mathrm{\φ}}_i}{2\mathrm{\πd}}\right)$ , i=1,2 ..., N _p-1 obtains required estimated signals arrival direction angle.

2, the Wave arrival direction estimating method based on Path selection according to claim 1 is characterized in that, the described parametric polynomial vector that obtains needs $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T$ Estimation obtain by following method:

If the initial condition of this method is: w (0), residual vector g (0), gradient vector p (1)=g (0), l=1, calculate successively and upgrade step-length: $\mathrm{\α}\left(l\right)=\mathrm{\η}\frac{p{\left(l\right)}^{T}g(l-1)}{p{\left(l\right)}^{T}R\left(l\right)p\left(l\right)},({\mathrm{\λ}}_f-0.5)\≤\mathrm{\η}\≤{\mathrm{\λ}}_f$ Desired parameters multinomial vector update calculation is:

W (l)=w (l-1)+α (l) p (l) residual vector more new formula is:

G (l)=λ _fG (l-1)-α (l) R (l) p (l)-R _SS(l) w (l-1) gradient adjustment step size computation is as the following formula: $\mathrm{\β}\left(l\right)=\frac{{(g\left(l\right)-g(l-1))}^{T}g\left(l\right)}{g{(l-1)}^{T}g(l-1)}$ The gradient vector update calculation:

P (l)=g (l)+β (l) p (l) obtains parameters needed multinomial vector by iterative computation repeatedly $w\left(l\right)=[w_1{w}_2...w_{N_P}{]}^T.$

3, the Wave arrival direction estimating method based on Path selection according to claim 1 is characterized in that, described definite number of parameters N that will estimate _p, can determine according to system's actual conditions.

4, the Wave arrival direction estimating method based on Path selection according to claim 1 is characterized in that, described definite number of parameters N that will estimate _p, can application message amount criterion method determine.