CN104239731A - Direction estimation method of MIMO-UKF-MUSIC (Multiple Input Multiple Output-Unscented Kalman Filter-Multiple Signal Classification) target - Google Patents

Abstract

The invention relates to a direction estimation method of an MIMO-UKF-MUSIC (Multiple Input Multiple Output-Unscented Kalman Filter-Multiple Signal Classification) target. In the actual direction finding process, according to an MUSIC spatial spectrum estimation algorithm, fluctuating target signal-to-noise ratio and time-varying strong interference commonly occur, radar signals of a plurality of targets of an MIMO sky wave radar are coherent signals, the disadvantageous conditions need to be deeply analyzed, and the estimation performance of an MUSIC algorithm under non-ideal conditions is improved. The direction estimation method comprises the following steps of 1, acquiring a plurality of groups of DOA (Direction of Arrival) estimation values and a plurality of groups of DOD (Direction of Departure) estimation values in a plurality of sensor arrays under an MIMO system; 2, filtering the DOA estimation values and the DOD estimation values respectively to obtain a plurality of groups of DOA estimation filtering values and a plurality of groups of DOD estimation filtering values, and averaging the plurality of groups of DOA estimation filtering values and the plurality of groups of DOD estimation filtering values to obtain final DOA estimation and final DOD estimation. The direction estimation method of the MIMO-UKF-MUSIC target is applicable to the field of array signal processing.

Description

MIMO-UKF-MUSIC target direction method of estimation

Technical field

The present invention relates to a kind of MIMO-UKF-MUSIC target direction method of estimation, belong to Array Signal Processing field.

Background technology

MIMO (Multiple input muitiple output, multiple-input and multiple-output) radar system is a kind of brand-new concept of field of radar, and oneself is through demonstrating great potentiality and advantage.MIMO radar, with the feature of its multiple-input and multiple-output, has unique advantage increasing in spatial sampling, the antagonism of anti-stealthy, antielectron and interference etc.In order to ensure its own existence ability, grasp electromagnetism processed power, realize the object to target effective monitoring, radar system needs to have good bearing resolution, can distinguish location quickly and accurately to target azimuth.Therefore, good target Bearing Estimation performance is just of great significance MIMO radar system detection of a target information tool.

Current is that the spatial spectral estimation algorithm theory represented has reached its maturity with multiple signal classification (MUSIC), but in actual direction-finding system, often exist fluctuating target signal to noise ratio and time the strong jamming that becomes, and for MIMO folded Clutter in Skywave Radars, the radar signal being positioned at multiple targets of same distance-speed resolution element is coherent signal etc., and these adverse condition all can make the estimated performance of MUSIC algorithm be affected even to lose efficacy.So analysing in depth it and improving the estimated performance of MUSIC algorithm under non-ideal condition is the problem needing solution badly.

Summary of the invention

The object of the invention is in order to solve in existing target direction estimation procedure be vulnerable to fluctuating signal to noise ratio (S/N ratio) and time become the shortcoming of strongly disturbing impact, and a kind of method of estimation to target direction is proposed.

(MUSIC, refers to MUSIC algorithm to a kind of MIMO-UKF-MUSIC target direction method of estimation, is the abbreviation of English Multiple Signal Classification, multi-signal sorting algorithm; UKF is the abbreviation of English Unscented Kalman Filter, harmless Kalman filtering), described target direction method of estimation is realized by following steps:

Step one, obtains many group DOA estimated values and many group DOD estimated values under MIMO system in multisensor array,

Step 2, respectively many group DOA are obtained to DOA estimated value and the filtering of DOD estimated value afterwards and estimate that filter value and many group DOD estimate filter value, ask for many group DOA estimation filter values and organize the average that DOD estimates filter value more, obtaining final DOA and estimate and final DOD estimation;

Often organize described DOA in step one and estimate that the process of asking for of filter value is:

Step is one by one: utilizing front-rear space smooth MUSIC algorithm to carry out DOA estimation by receiving transmitting that battle array receives, the DOA estimated value obtained being carried out filtering and noise reduction by UKF wave filter, obtains DOA and estimate filter value;

Step one two: the DOA utilizing step to obtain one by one estimates filter value structure blocking matrix, carries out obstruction suppress the signal of interference radiating way;

Step one three: again carry out DOA estimation by front-rear space smooth MUSIC algorithm by blocking the signal suppressed, afterwards by UKF filter filtering, obtains DOA and estimates filter value;

Often organize described DOD and estimate that the process of asking for of filter value is:

Step one four: utilize front-rear space smooth MUSIC algorithm to carry out DOD estimation the Received signal strength that transmitting battle array receives, the DOD estimated value obtained is carried out filtering and noise reduction by UKF wave filter, obtains DOD and estimates filter value;

The step First Five-Year Plan: the DOD utilizing step one four to obtain estimates filter value structure blocking matrix, carries out obstruction suppress the signal of interference radiating way;

Step one six: again carry out DOD estimation by front-rear space smooth MUSIC algorithm by blocking the signal suppressed, afterwards by UKF filter filtering, obtains DOD and estimates filter value.

Beneficial effect of the present invention is: 1, this algorithm adopts front-rear space smooth algorithm to process the reception data covariance matrix in MUSIC algorithm, the problem of traditional MUSIC algorithm estimated performance degradation when signal coherence under solving MIMO folded Clutter in Skywave Radars system; 2, by UKF wave filter, filtering is carried out to the measurement of MUSIC algorithm, effectively improve the estimated accuracy of MIMO folded Clutter in Skywave Radars, achieve good denoising effect; 3, utilize the filter value measured to build blocking matrix, interference is effectively suppressed, and relative to traditional MUSIC algorithm, block and interference is obviously reduced the impact that target direction is estimated, improve the precision that target direction is estimated.

Accompanying drawing explanation

Fig. 1 is the schematic block diagram of the UKF-MUSIC algorithm that the present invention relates to;

Fig. 2 is that MIMO folded Clutter in Skywave Radars of the present invention is launched battle array and received the schematic diagram of corresponding relation between a burst of unit;

Fig. 3 is the method for estimation schematic block diagram of MIMO folded Clutter in Skywave Radars emission angle DOA of the present invention;

Fig. 4 is the method for estimation schematic block diagram of MIMO folded Clutter in Skywave Radars emission angle DOD of the present invention;

Fig. 5 is MUSIC algorithm single DOD estimated result, and label 1 place represents that target 2 is hidden by strong jamming under traditional MUSIC algorithm, and the signal of shift phenomenon appears in spectrum peak; In figure, horizontal ordinate represents DOD value, ordinate representation space spectrum normalization amplitude;

Fig. 6 is algorithm single DOD estimated result of the present invention, and label 2 place represents in the inventive method, disturbs suppressed, and the direction of target 2 is effectively estimated; In figure, horizontal ordinate represents DOD value, ordinate representation space spectrum normalization amplitude;

Fig. 7 is MUSIC algorithm single DOA estimated result, and label 3 represents strong jamming direction, and label 4 represents target 2 direction (under traditional MUSIC algorithm, interference does not get clogged); In figure, horizontal ordinate represents DOA value, ordinate representation space spectrum normalization amplitude;

Fig. 8 is algorithm single DOA estimated result of the present invention, and label 5 represents target 2 direction (under the inventive method, interference gets clogged); In figure, horizontal ordinate represents DOA value, ordinate representation space spectrum normalization amplitude.

Embodiment

Embodiment one:

The MIMO-UKF-MUSIC target direction method of estimation of present embodiment, as shown in Figure 3 and Figure 4, described target direction method of estimation is realized by following steps:

Step one: obtain many group DOA estimated values and many group DOD estimated values under MIMO system in multisensor array,

Step 2: respectively many group DOA are obtained to DOA estimated value and the filtering of DOD estimated value afterwards and estimate that filter value and many group DOD estimate filter value, ask for many group DOA estimation filter values and organize the average that DOD estimates filter value more, obtaining final DOA and estimate and final DOD estimation;

As Fig. 1, often organize described DOA in step one and estimate that the process of asking for of filter value is:

Step is one by one: utilizing front-rear space smooth MUSIC algorithm to carry out DOA estimation by receiving transmitting that battle array receives, the DOA estimated value obtained being carried out filtering and noise reduction by UKF wave filter, obtains DOA and estimate filter value;

Step one two: the DOA utilizing step to obtain one by one estimates filter value structure blocking matrix, carries out obstruction suppress the signal of interference radiating way;

Step one three: again carry out DOA estimation by front-rear space smooth MUSIC algorithm by blocking the signal suppressed, afterwards by UKF filter filtering, obtains DOA and estimates filter value;

As Fig. 2, often organize described DOD and estimate that the process of asking for of filter value is:

Step one four: utilize front-rear space smooth MUSIC algorithm to carry out DOD estimation the Received signal strength that transmitting battle array receives, the DOD estimated value obtained is carried out filtering and noise reduction by UKF wave filter, obtains DOD and estimates filter value;

The step First Five-Year Plan: the DOD utilizing step one four to obtain estimates filter value structure blocking matrix, carries out obstruction suppress the signal of interference radiating way;

Step one six: again carry out DOD estimation by front-rear space smooth MUSIC algorithm by blocking the signal suppressed, afterwards by UKF filter filtering, obtains DOD and estimates filter value.

Embodiment two:

MIMO-UKF-MUSIC target direction method of estimation described in embodiment one, step one by one front-rear space smooth MUSIC algorithm described in middle step one four is specially:

The first, carry out the calculating of covariance matrix: wherein,

Forward direction smoothed covariance is: p represents the submatrix number of division and p=M-m+1, M are array number, and m is the array number of each submatrix, wherein, and S _irepresent submatrix data covariance, A ^hrepresent the conjugate transpose of array steering vector matrix A, I representation unit matrix, P ^frepresent forward direction smoothed covariance matrix and $P^f=\frac{1}{p}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{D}^{i-1}P{\left(D^{i-1}\right)}^H,$ P representation signal covariance matrix, $D=\left[\begin{array}{cccc}{e}^{j{\mathrm{\&beta;}}_1}& 0& ...& 0\\ 0& e^{j{\mathrm{\&beta;}}_2}& ...& 0\\ ...& ...& ...& ...\\ 0& 0& ...& e^{j{\mathrm{\&beta;}}_D}\end{array}\right],$ Wherein d is array element distance, θ _i(i=1,2 ..., D, D<m) and be the angle of incident direction, λ is incoming signal wavelength;

Backward smoothed covariance is: wherein, p represents the submatrix number of division, and p=M-m+1, M are array number, and m is the array number of each submatrix, and p represents the submatrix number of division, wherein, represent p-i+1 submatrix data covariance, A ^hrepresent the conjugate transpose of array steering vector matrix A, I representation unit matrix, Pb represent backward smooth signal covariance matrix and p representation signal covariance matrix,

In described forward direction smoothed covariance and backward smoothed covariance, each submatrix covariance matrix is:

$S_k\stackrel{\mathrm{\&Delta;}}{=}\stackrel{\&OverBar;}{X_k{X}_k^{*}}={\mathrm{AD}}^{(k-1)}P\left[{\left(D^{-1}\right)}^H\right]A^H+{\mathrm{\&sigma;}}^{2}I,$

Signal covariance matrix $P\stackrel{\mathrm{\&Delta;}}{=}\stackrel{\&OverBar;}{{\mathrm{FF}}^{*}},$ Wherein, incoming signal $F={[f_1\left(\hat{i}\right),f_2\left(\hat{i}\right),...,f_D\left(\hat{i}\right)]}^T,$ represent incident signal component, defining variable to describe two-dimensional distance velocity spectrum signal spectrum peak position, wherein r _binrepresent range gate, v _binexpression speed door,

Submatrix receives data: $X_k\left(\hat{i}\right)=[x_k,x_{k+1},...,x_{k+m-1}]={\mathrm{AD}}^{(k-1)}F\left(\hat{i}\right)+{\mathrm{\&eta;}}_k\left(\hat{i}\right),$ Wherein, represent additive white Gaussian noise, A be array guiding matrix for and A=[a (θ ₁), a (θ ₂) ..., a (θ _d)], a (θ _i) for steering vector and a (θ _i)=[exp (-j ω τ _1i), exp (-j ω τ _2i) ..., exp (-j ω τ _ki) ..., exp (-j ω τ _mi)] ^t; τ _kifor delay time and m is the array number of each submatrix, and d is array element distance, θ _i(i=1,2 ..., D, D<m) and be the angle of incident direction, ω is angular frequency and ω=2 π f, f are carrier frequency, and D is signal source number, and j is imaginary unit, and c is the light velocity, and λ is incoming signal wavelength;

The second, to S ^fbcarry out feature decomposition, obtain S ^fb=U Σ U ^h, U is feature matrix, Σ be the diagonal matrix that is made up of eigenwert and meet λ ₁>=λ ₂>=...>=λ _d>=λ _d+1=...=λ _m=σ ², by decompose after individual minimal eigenvalue λ _mincharacteristic of correspondence vector composition noise subspace E _n, obtaining space spectral function is $P_{\mathrm{MUSIC}}\left(\mathrm{\&theta;}\right)=\frac{1}{a^{*}\left(\mathrm{\&theta;}\right)E_N{E}_N^{*}a\left(\mathrm{\&theta;}\right)},$ Then DOA estimated value is: ${\mathrm{\&theta;}}_{\mathrm{MUSIC}}=\arg \min a^{*}\left(\mathrm{\&theta;}\right)E_N{E}_N^{*}a\left(\mathrm{\&theta;}\right);$ Wherein, represent the estimate amount of incoming signal, E _nfor the spatial noise that the noise feature vector after covariance matrix feature decomposition is formed, for E _ncomplex conjugate, a (θ) is steering vector, a ^*(θ) be the complex conjugate of a (θ);

3rd, by DOA estimated values theta obtained in the previous step _mUSICfiltering is carried out by UKF wave filter.

Embodiment three:

MIMO-UKF-MUSIC target direction method of estimation according to embodiment two, described submatrix number p is: p=D+1; Wherein, D is signal source number.

Embodiment four:

MIMO-UKF-MUSIC target direction method of estimation according to embodiment one, two or three, described in step one three and step one six, UKF wave filter carries out filtering and noise reduction, and k+1 moment target discrete state equation is: X (k+1)=Φ (k) X (k)+G (k) V (k); Wherein, Φ (k) represents the state-transition matrix on k moment n × n rank, and X (k) maintains system state vector for n, and G (k) is that n × n ties up process noise distribution matrix, V (k) ties up process noise for n, is zero-mean, covariance gauss white noise,

When surveyed angle on target fix or linear change time, its system state vector is state-transition matrix F and G is respectively: $F=\left[\begin{array}{cc}1& T\\ 0& 1\end{array}\right],$ $G\left(k\right)=\left[\begin{array}{cc}{0.5T}^2& 0\\ T& 0\end{array}\right];$

Observation equation is: Z (k)=X (k)+W (k), and wherein, Z (k) is m dimensional vector, represents the measurement vector in k moment, and W (k) is that m ties up measurement noise;

Measurement noises variance is: for the covariance of angle estimation error.

Embodiment five:

MIMO-UKF-MUSIC target direction method of estimation according to embodiment four, if undesired signal direction is θ described in step one two and step First Five-Year Plan _j, order structure blockage factor b=exp (j ω τ ₁), structure blocking matrix when estimating DOA, M=M _r, the signal after obstruction is y _dOA=Tx; When estimating DOD, M=M _t, the signal after obstruction is y _dOD=Tx; Wherein, x represents that radar receives data.

Embodiment six:

MIMO-UKF-MUSIC target direction method of estimation according to embodiment one, two, three or five, by M _tunit's even linear array launches battle array, M _runit's even linear array receives in the MIMO system lower sensor array system of battle array composition, for standard list base MIMO folded Clutter in Skywave Radars, because the distance of launching battle array and reception battle array is very little to the distance of radar relative to the detection of a target, so the relative equidistant transmitting battle array of MIMO folded Clutter in Skywave Radars and reception battle array, the transmit direction DOD of its correspondence and receive direction DOA approximately equal, launch battle array and receive corresponding relation between a burst of unit as shown in Figure 2, described transmitting battle array array number is M _t, described reception battle array array number is M _r, then receiving a burst of unit described in each has M _tthe a burst of unit of individual transmitting is corresponding with it, any one reception of described reception battle array array element k (k=1,2 ... M _r) there is a M _rthe DOD that a burst of unit of individual reception estimates _kcorresponding with it, obtain: by M _tunit's even linear array launches battle array, M _runit's even linear array receives in the MIMO system lower sensor array system of battle array composition and obtains:

M _tgroup M _runit's transmitting battle array estimates that the DOA obtained estimates that angle expression formula is DOA _k(k=1,2 ... M _t), then final DOA is estimated as $\mathrm{DOA}=\frac{1}{M_t}\underset{k=1}{\overset{M_t}{\mathrm{\&Sigma;}}}{\mathrm{DOA}}_k;$ Calculation process as shown in Figure 3;

M _rgroup M _tunit's transmitting battle array estimates that the DOD obtained estimates that angle expression formula is DOD _k(k=1,2 ... M _r), then final DOD is estimated as $\mathrm{DOD}=\frac{1}{M_r}\underset{k=1}{\overset{M_r}{\mathrm{\&Sigma;}}}{\mathrm{DOD}}_k,$ Calculation process as shown in Figure 4.

Experiment 1: the precision that checking DOA estimates and the relation between signal number D and submatrix number p.

The coherent signal that one group of incident direction is different is set, signal to noise ratio (S/N ratio) is 5dB, even linear array array number is 20, front-rear space smooth MUSIC algorithm is utilized to carry out DOA estimation to signal arrival bearing, when changing signal number D and submatrix number p respectively, carry out 400 independent emulation experiments, add up the RMSE result obtained, as shown in table 1:

The relation of table 1DOA estimated accuracy and signal number D and submatrix number p

(D is signal source number, and p is submatrix number, and in table, data are RMSE)

Interpretation of result: from table data draw, when the submatrix number chosen be p=D+1 (D is signal source number) and neighbouring time, the estimation RMSE of algorithm is minimum, and precision is higher.So before application backward space smoothing MUSIC algorithm time, the basis for selecting of submatrix number is p=D+1.

Experiment 2: set the transmitting array number of MIMO folded Clutter in Skywave Radars as M _t=8, reception array number is M _r=32, array element is spaced apart d _t=d _r=λ/2, the direction of 3 signal sources is respectively-60 °, 0 °, and 70 ° (being followed successively by target 1, target 2, target 3), undesired signal direction rises and falls between 20-25dB for-2.5 ° of signal to noise ratio (S/N ratio)s, and fast umber of beats is 100.The result is as shown in Fig. 5-Fig. 8.

(1) MIMO folded Clutter in Skywave Radars launches battle array DOD estimation, as shown in Figure 5, Figure 6:

Table 2MIMO folded Clutter in Skywave Radars DOD estimates that the RMSE of MUSIC algorithm and algorithm of the present invention compares

(2) MIMO folded Clutter in Skywave Radars receives battle array DOA estimation, as shown in Figure 7, Figure 8:

Table 3MIMO folded Clutter in Skywave Radars DOA estimates that the RMSE of MUSIC algorithm and algorithm of the present invention compares

Note: RMSE (Root Mean Square Error, root-mean-square error).

Can see, when DOD estimates (as shown in Figure 5, Figure 6), for traditional MUSIC algorithm, the disturbed covering of target 2, there is larger skew in its spatial spectrum peak, cannot effectively estimate its direction; And algorithm of the present invention is by after filtering, obstruction process, target 2 effectively can be detected, (see table 2) is also improved to the estimated accuracy of other target directions simultaneously; When DOA estimates (as shown in Figure 7, Figure 8), much larger relative to transmitting battle array owing to receiving battle array element number of array, so for traditional MUSIC algorithm, also target 2 can be distinguished, but after application this method, improve its direction estimation precision (see table 3).

Claims (6)

1. a MIMO-UKF-MUSIC target direction method of estimation, is characterized in that: described target direction method of estimation is realized by following steps:

Step one: obtain many group DOA estimated values and many group DOD estimated values under MIMO system in multisensor array,

Step 2: respectively many group DOA are obtained to DOA estimated value and the filtering of DOD estimated value afterwards and estimate that filter value and many group DOD estimate filter value, ask for many group DOA estimation filter values and organize the average that DOD estimates filter value more, obtaining final DOA and estimate and final DOD estimation;

Often organize described DOA in step one and estimate that the process of asking for of filter value is:

Step is one by one: utilizing front-rear space smooth MUSIC algorithm to carry out DOA estimation by receiving transmitting that battle array receives, the DOA estimated value obtained being carried out filtering and noise reduction by UKF wave filter, obtains DOA and estimate filter value;

Step one two: the DOA utilizing step to obtain one by one estimates filter value structure blocking matrix, carries out obstruction suppress the signal of interference radiating way;

Step one three: again carry out DOA estimation by front-rear space smooth MUSIC algorithm by blocking the signal suppressed, afterwards by UKF filter filtering, obtains DOA and estimates filter value;

Often organize described DOD and estimate that the process of asking for of filter value is:

Step one four: utilize front-rear space smooth MUSIC algorithm to carry out DOD estimation the Received signal strength that transmitting battle array receives, the DOD estimated value obtained is carried out filtering and noise reduction by UKF wave filter, obtains DOD and estimates filter value;

The step First Five-Year Plan: the DOD utilizing step one four to obtain estimates filter value structure blocking matrix, carries out obstruction suppress the signal of interference radiating way;

Step one six: again carry out DOD estimation by front-rear space smooth MUSIC algorithm by blocking the signal suppressed, afterwards by UKF filter filtering, obtains DOD and estimates filter value.

2. MIMO-UKF-MUSIC target direction method of estimation according to claim 1, is characterized in that: step one by one in front-rear space smooth MUSIC algorithm described in step one four be specially:

The first, carry out the calculating of covariance matrix: wherein,

Forward direction smoothed covariance is: p represents the submatrix number of division and p=M-m+1, M are array number, and m is the array number of each submatrix, wherein, and S _irepresent submatrix data covariance, A ^hrepresent the conjugate transpose of array steering vector matrix A, I representation unit matrix, P ^frepresent forward direction smoothed covariance matrix and $P^f=\frac{1}{p}\underset{i=1}{\overset{p}{\mathrm{\&Sigma;}}}{D}^{i-1}P{\left(D^{i-1}\right)}^H,$ P representation signal covariance matrix, $D=\left[\begin{array}{cccc}{e}^{j{\mathrm{\&beta;}}_1}& 0& ...& 0\\ 0& e^{j{\mathrm{\&beta;}}_2}& ...& 0\\ ...& ...& ...& ...\\ 0& 0& ...& e^{j{\mathrm{\&beta;}}_D}\end{array}\right],$ Wherein d is array element distance, θ _i(i=1,2 ..., D, D<m) and be the angle of incident direction, λ is incoming signal wavelength;

Backward smoothed covariance is: wherein, p represents the submatrix number of division, and p=M-m+1, M are array number, and m is the array number of each submatrix, and p represents the submatrix number of division, wherein, represent p-i+1 submatrix data covariance, A ^hrepresent the conjugate transpose of array steering vector matrix A, I representation unit matrix, P ^brepresent backward smooth signal covariance matrix and p representation signal covariance matrix;

In described forward direction smoothed covariance and backward smoothed covariance, each submatrix covariance matrix is:

$S_k\stackrel{\mathrm{\&Delta;}}{=}\stackrel{\&OverBar;}{X_k{X}_k^{*}}={\mathrm{AD}}^{(k-1)}P\left[{\left(D^{-1}\right)}^H\right]A^H+{\mathrm{\&sigma;}}^{2}I,$

Signal covariance matrix $P\stackrel{\mathrm{\&Delta;}}{=}\stackrel{\&OverBar;}{{\mathrm{FF}}^{*}},$ Wherein, incoming signal $F={[f_1\left(\hat{i}\right),f_2\left(\hat{i}\right),...,f_D\left(\hat{i}\right)]}^T,$ represent incident signal component, defining variable to describe two-dimensional distance velocity spectrum signal spectrum peak position, wherein r _binrepresent range gate, v _binexpression speed door;

Submatrix receives data: $X_k\left(\hat{i}\right)=[x_k,x_{k+1},...,x_{k+m-1}]={\mathrm{AD}}^{(k-1)}F\left(\hat{i}\right)+{\mathrm{\&eta;}}_k\left(\hat{i}\right),$ Wherein, represent additive white Gaussian noise, A be array guiding matrix for and A=[a (θ ₁), a (θ ₂) ..., a (θ _d)], a (θ _i) for steering vector and a (θ _i)=[exp (-j ω τ _1i), exp (-j ω τ _2i) ..., exp (-j ω τ _ki) ..., exp (-j ω τ _mi)] ^t; τ _kifor delay time and m is the array number of each submatrix, and d is array element distance, θ _i(i=1,2 ..., D, D<m) and be the angle of incident direction, ω is angular frequency and ω=2 π f, f are carrier frequency, and D is signal source number, and j is imaginary unit, and c is the light velocity, and λ is incoming signal wavelength;

The second, to S ^fbcarry out feature decomposition, obtain S ^fb=U Σ U ^h, U is feature matrix, Σ be the diagonal matrix that is made up of eigenwert and meet λ ₁>=λ ₂>=...>=λ _d>=λ _d+1=...=λ _m=σ ², by decompose after individual minimal eigenvalue λ _mincharacteristic of correspondence vector composition noise subspace E _n, obtaining space spectral function is $P_{\mathrm{MUSIC}}\left(\mathrm{\&theta;}\right)=\frac{1}{a^{*}\left(\mathrm{\&theta;}\right)E_N{E}_N^{*}a\left(\mathrm{\&theta;}\right)},$ Then DOA estimated value is: ${\mathrm{\&theta;}}_{\mathrm{MUSIC}}=\arg \min a^{*}\left(\mathrm{\&theta;}\right)E_N{E}_N^{*}a\left(\mathrm{\&theta;}\right);$ Wherein, represent the estimate amount of incoming signal, E _nfor the spatial noise that the noise feature vector after covariance matrix feature decomposition is formed, for E _ncomplex conjugate, a (θ) is steering vector, a ^*(θ) be the complex conjugate of a (θ);

3rd, by DOA estimated values theta obtained in the previous step _mUSICfiltering is carried out by UKF wave filter.

3. MIMO-UKF-MUSIC target direction method of estimation according to claim 2, is characterized in that: described submatrix number p is: p=D+1, D are signal source number.

4. MIMO-UKF-MUSIC target direction method of estimation according to claim 1,2 or 3, it is characterized in that: described in step one three and step one six, UKF wave filter carries out filtering and noise reduction, and k+1 moment target discrete state equation is: X (k+1)=Φ (k) X (k)+G (k) V (k); Wherein, Φ (k) represents the state-transition matrix on k moment n × n rank, and X (k) maintains system state vector for n, and G (k) is that n × n ties up process noise distribution matrix, V (k) ties up process noise for n, is zero-mean, covariance white Gaussian noise,

When surveyed angle on target fix or linear change time, its system state vector is state-transition matrix F and G is respectively: $F=\left[\begin{array}{cc}1& T\\ 0& 1\end{array}\right],$ $G\left(k\right)=\left[\begin{array}{cc}{0.5T}^2& 0\\ T& 0\end{array}\right];$

Observation equation is: Z (k)=X (k)+W (k), and wherein, Z (k) is m dimensional vector, represents the measurement vector in k moment, and W (k) is that m ties up measurement noise;

Measurement noises variance is: R=diag [σ _θ ²], σ _θ ²for the covariance of angle estimation error.

5. MIMO-UKF-MUSIC target direction method of estimation according to claim 4, is characterized in that: set undesired signal direction described in step one two and step First Five-Year Plan as θ _j, order structure blockage factor b=exp (j ω τ ₁), structure blocking matrix when estimating DOA, M=M _r, the signal after obstruction is y _dOA=Tx; When estimating DOD, M=M _t, the signal after obstruction is y _dOD=Tx; Wherein, x represents that radar receives data.

6. MIMO-UKF-MUSIC target direction method of estimation according to claim 1,2,3 or 5, is characterized in that: by M _tunit's even linear array launches battle array, M _runit's even linear array receives in the MIMO system lower sensor array system of battle array composition and obtains:

M _tgroup M _runit's transmitting battle array estimates that the DOA obtained estimates that angle expression formula is DOA _k(k=1,2 ... M _t), then final DOA is estimated as $\mathrm{DOA}=\frac{1}{M_t}\underset{k=1}{\overset{M_t}{\mathrm{\&Sigma;}}}{\mathrm{DOA}}_k;$

M _rgroup M _tunit's transmitting battle array estimates that the DOD obtained estimates that angle expression formula is DOD _k(k=1,2 ... M _r), then final DOD is estimated as $\mathrm{DOD}=\frac{1}{M_r}\underset{k=1}{\overset{M_r}{\mathrm{\&Sigma;}}}{\mathrm{DOD}}_k.$