CN106169941A - A kind of DOA method of estimation of round battle array based on noise power - Google Patents

Abstract

The invention belongs to wireless communication technology field, particularly relate to one and utilize noise to carry out the algorithm of subspace estimation true to signal in wireless multiple-input and multiple-output (Multiple Input Multiple Output, MIMO) communication system.The present invention proposes a kind of DOA estimation method based on noise power estimation signal subspace, utilize the characteristic that white Gaussian noise power is constant, subspace DOA estimation method is become by utilizing the existence of noise to isolate signal subspace and noise subspace by separating merely signal subspace with noise subspace, thus utilize the correlation theory of matrix theory, in the case of channel is poor, performance is improved 50%.

Description

A kind of DOA method of estimation of round battle array based on noise power

Technical field

The invention belongs to wireless communication technology field, particularly relate to a kind of at wireless multiple-input and multiple-output (Multiple Input Multiple Output, MIMO) communication system utilize noise to carry out the calculation of subspace estimation true to signal Method.

Background technology

One important branch Array Signal Processing in signal processing field is in biomedical engineering commercial communication, military affairs All many-sides such as electronic warfare have a wide range of applications.At present, Array Signal Processing is roughly divided into two mains direction of studying: adaptive Answer airspace filter and Estimation of Spatial Spectrum.

Signal is processed and has good application prospect in modern communications by first array, and in recent years Arrive swift and violent development.Along with people propose higher standard, common communication to the capacity of communication system and the continuous of quality Antenna can not fully meet the highest performance indications requirement.Therefore, this just requires that we are necessary to take some skills Other potential of antenna is all excavated by art means and method, thus forms the antenna system meeting modern communications demand.So In order to meet the requirement that capacity and quality are improved constantly by Modern Communication System, the design of array antenna the most just becomes a kind of stream The method of row.Difference due to communication need, then be accomplished by difform array structure to meet different wave cover Form.It is understood that the conformal array such as circular cone or ring array has the covering power of 360 orientation angles, they can form one entirely To wave beam, multiple wave beam or a narrow beam and 360 orientation angles can be directed to.And hemisphere can be realized cover Array antenna system will highlight its glamour in the application of antenna of mobile communication base station day by day.It addition, some telecommunication satellites are also Begin with conformal array as antenna.

DOA algorithm based on array the earliest is conventional beamformer (CBF) method, also referred to as Barlett beam-forming schemes. This method is the simple Extended forms in a kind of spatial domain in conventional Time-domain Fourier spectrum method of estimation, i.e. receives by each array element in spatial domain Data replace conventional Time-domain process in time domain data.As limiting with time domain Fourier, this method is extended to spatial domain After, the angular resolution of array is similarly subjected to the restriction of spatial domain " Fourier's limit ".Spatial domain " Fourier's limit " is exactly the physics of array Aperture limits, and often claims " Rayleigh (Rayleigh) limit ".In other words, the extraterrestrial target being pointed to a beam angle can not be differentiated. So, the effective ways improving spatial processing precision increase antenna aperature (being equivalent to laugh at beam angle) exactly, thus reach Put forward high-precision purpose.But for many actual application environment, it is unpractical for increasing antenna aperature, so needing more preferably Ground algorithm improves the precision that orientation is estimated.Therefore, after CBF method proposes, many Time-Domain Nonlinear Power estimation methods are all In succession being applied among the process of spatial domain signal, create what is called " High-Resolution Spectral Estimation method ", representative has: The harmonic analysis method of Pisarenko, the maximum entropy method (MEM) of Brug, the least variance method etc. of Capon.In research subsequently, different Scholar propose different methods and the theoretical Rayleigh limit breaking through Array Signal Processing.

Later 1970s, Schmidt-R-O of the U.S. et al. proposes multiple signal classification algorithm (MUSIC), this algorithm successfully breaks through the Rayleigh limit of Array Signal Processing, has stepped important to modern super-resolution direction-finding technology One step.Also the rise indicating proper subspace algorithm while of appearance of MUSIC algorithm.The common feature of subspace Direction Finding Algorithm It is through the mathematic decomposition (feature decomposition, singular value decomposition, QR decomposition etc.) to array received data matrix, data will be received It is decomposed into two mutual orthogonal subspaces.Referred to as signal subspace consistent with array manifold space in the two subspace, another The individual referred to as noise subspace orthogonal with signal subspace.Sub space decomposition i.e. utilizes the characteristic that two subspaces are mutually orthogonal Special composition is composed, and completes direction finding by spectrum peak search and works.Additionally, as a kind of representative subspace method, ESPRIT algorithm is suggested the most subsequently and has obtained studying widely.

From the later stage eighties 20th century, this field occurs in that again a class subspace fitting class algorithm, wherein compares Representational algorithm has maximum likelihood (ML) algorithm, Weighted Sub-Space Fitting Direction (WSF) algorithm and multidimensional MUSIC algorithm etc..? Maximum-likelihood (ML) parameter estimation class method is a kind of typical case and practical method of estimation in parameter estimation theories, and it includes definitiveness Maximum likelihood algorithm (DML) and randomness maximum likelihood algorithm (SML).1988, Ziskind L and Max M discussed Maximum-likelihood method for parameter estimation is applied to Mutual coupling, owing to direction estimation likelihood function is nonlinear, solves it Excellent solution need to carry out multi-dimensional search, and operand is huge.Therefore follow-up study concentrates on the estimation performance of algorithm and realizes.

In order to preferably carry out airspace filter, we first can carry out Estimation of Spatial Spectrum to signal, it and adaptive array Technology is different, and the research direction that stresses of Estimation of Spatial Spectrum is that the multisensor array processing system that constituted in space is to interested The ability that the many kinds of parameters of spacing wave is accurately estimated, thus this system main purpose be estimate signal spatial domain parameter or Information source position, this is also one of vital tasks in many fields such as radar, communication, sonar.Estimation of Spatial Spectrum technology can be significantly Improve the angle estimation precision of spacing wave, angular resolution and other relevant parameter precision in system processes bandwidth, thus Its application prospect is quite varied.

Summary of the invention

The defect that during in order to overcome common DOA to estimate, expense is excessive, the present invention proposes a kind of based on noise power estimation letter The DOA algorithm for estimating in work song space.Utilize the characteristic that white Gaussian noise power is constant, by subspace DOA estimation method by merely Separate signal subspace and become empty with noise by utilizing the existence of noise to isolate signal subspace with noise subspace Between, thus utilize the correlation theory of matrix theory, in the case of channel is poor, performance is improved 50%.

In order to describe present disclosure easily, first the concept used in the present invention and term are defined.

Direction of arrival (DOA) is estimated: determine the space bit of multiple signals interested in being concurrently residing in a certain region, space Put (deflection that the most multiple signals arrive array reference array element).

Signal subspace: by the spatial organization of signal guide Linear Combination of Vectors.

Noise subspace: by the spatial organization of non-signal steering vector linear combination.

Signal to noise ratio: signal power and the ratio of noise power.

First a kind of conventional subspace DOA algorithm for estimating is as follows:

Assume that target signal source is mutual statistical independence, and additive white Gaussian noise is the noise that it receives.That Can show that signal correlation matrix P can be represented as P=E [s (n) s^H(n)]=diag{P_i}.Wherein, P_iBelieve for i-th Number mean power.Being similar to, the Correlation Moment R that can define reception signal x (n) is: R=E [x (n) x^H(n)]=APA^H+σ²I, Wherein, σ²Represent the variance of additive white Gaussian noise.In order to ensure each linear independence of direction matrix A, so N ＞ L, the most just It is to say that array number have to be larger than signal number.

In the form of above formula and MUSIC algorithm, the autocorrelation matrix of signal has an identical form, and signal The form of frequency matrix is also identical with direction matrix A, it is possible to MUSIC algorithm is applied to the place of spatial correlation matrix In should, carry out the phase information contained in estimated matrix A, further by it is estimated that the deflection of signal source, thus complete The Mutual coupling of pair signals.

For trying to achieve the phase information of echo signal, Correlation Moment R will be carried out Eigenvalues Decomposition, and by eigenvalue by dull non- Descending order, then the normalization characteristic vector that front L eigenvalue is corresponding just will open into signal subspace, rear N-L individual to Amount opens into noise subspace.

If definition noise subspace is G=u_K+1, u_K+2..., u_K+N, then it meets following relation A with array manifold A^HG= 0, i.e. each array guiding vector of signal is orthogonal with noise subspace under conditions of signal is incoherent signal.? Under normal circumstances, the time average accepting data x (n) obtained by M snap is estimated to substitute former Correlation Moment, if it is. Common SVD decompose the noise subspace produced not with array manifold strict orthogonal, therefore utilize this critical nature, structure Spectrum peak search function is as follows:Wherein, θ ∈ (-90 °, 90 °).L the peak value of MUSIC spectrum P That position is corresponding is exactly the direction of arrival θ of signal_kEstimation.

The DOA estimation method of a kind of round battle array based on noise power, specifically comprises the following steps that

S1, the n times snap data separate of array received is utilized to be calculated the space correlation square R of signal；

S2, to carrying out Eigenvalues Decomposition, obtain noise subspace matrix G；

S3, carry out spectrum peak search according to noise subspace matrix G described in space correlation square R and S2 described in S1, obtain its spectrum Peak position, thus obtain Mutual coupling value.

The invention has the beneficial effects as follows:

Performance is improved 50% in the case of channel is poor by the present invention.

Accompanying drawing explanation

Fig. 1 DOA algorithm for estimating system diagram based on circle battle array.

Fig. 2 is the one-shot measurement figure during antenna training.

Fig. 3 is the flow chart of simulated program of the present invention.

Fig. 4 be angle of incidence be 20 measured values time (36 °, 20 °).

Fig. 5 be angle of incidence be 20 measured values time (45 °, 30 °).

Fig. 6 be angle of incidence be 20 measured values time (30 °, 45 °).

Detailed description of the invention

Below in conjunction with the accompanying drawings, the present invention is described in further detail.

First, carry out auto-correlation by the docking collection of letters number and obtain Rxx, it is assumed that target signal source is mutual statistical independence, And additive white Gaussian noise is the noise that it receives.So can show that signal correlation matrix P can be represented as P=E [s(n)s^H(n)]=diag{P_i}.Wherein, P_iMean power for i-th signal.It is similar to, reception signal x (n) can be defined Correlation Moment R be: R=E [x (n) x^H(n)]=APA^H+σ²I。

Then, first is obtained by the vector of actual signal subspace linear combination ForAssume it is incoherent between information source, and ifBe desired orientation steering vector time Waiting, following formula can be obtained:

, Wherein,It is by all steering vector linear combinations after pretreated, i.e. only by by pre- The K rank vector of the signal subspace linear combination after reason.

WhenThe when of not being the steering vector of desired orientation, then:

Wherein, WhenThe when of not being the steering vector of desired orientation,It is all with the presence of noise when With the steering vector after process and steering vectorLinear combination, i.e. comprise actual signal subspace empty with another Between K+1 rank vector,.When noise is non-existent time, no matter whether desired orientation is the steering vector of desired orientation, obtain Result is all the K rank vector combined by whole signal subspace SYSTEM OF LINEAR VECTOR, so can be in noise at this algorithm Set up the when of smaller, and need to meet L^HL=I, because only that so, through with process after making an uproar in space Sound still keeps white Gaussian noise form.

Obtain other subspace vectors, owing to whether using the steering vector of desired orientationAll comprise whole Individual actual signal subspace.So according to following formula second vector comprising actual signal subspace steering vector can obtain:

By to matrix theory learning, can obtain above-mentioned equation optimal solution is:

Want iteration to draw residue (K-1) individual orthogonal vectors, define two intermediate vector:

Wherein,It is that front n-1 auxiliary is vowed Amount baseline combines, then

Wherein,

According to formulaAnd use mathematical induction to ask :

Wherein, p Scope is [1, K-1], if i.e.The when of being the steering vector of desired orientation, vector base is K rank, if there being K + 1 vector mutually orthogonal vector individual with this K is orthogonal, then this vector must be null vector.WhenIt it not the phase The when of hoping the steering vector in direction, then because this vector base is K+1 rank, so this vector is not null vector.

The K+1 orthogonal vectors is expressed:

Give expression to the K+1 auxiliary vector long-pending after, ifBe desired orientation steering vector time Wait,So can directly useCarry out spectrum peak search.Can analyze from formula and draw Even in the presence of having noiseBut he also can be in close proximity to zero, then its inverse is also A crest can be formed at this point.

When angle of incidence is 3 respectively: (60 °, 45 °) (60 °, 20 °) (45 °, 30 °), matlab uses above-mentioned algorithm Emulation is as in figure 2 it is shown, have employed 25 array elements in simulations, and phase modulus value is equal to 10, and array element distance is 0.75 wavelength, Signal to noise ratio is 0dB, and fast umber of beats is 128.

First their incidence will will be assumed respectively with regard to this algorithm compared with classical high resolution algorithm MUSIC algorithm Angle is: (36 °, 20 °) (30 °, 45 °) (45 °, 30 °), through 20 emulation, have employed 25 array elements, phase modulus value in simulations Equal to 10, and array element distance is 0.75 wavelength, and signal to noise ratio is 0dB, and noise uses white Gaussian noise, and fast umber of beats is The simulation of 200 emulates.Then the data drawn do broken line graph such as shown in Fig. 4, Fig. 5 and Fig. 6:

From Fig. 4, Fig. 5, Fig. 6 it can be seen that in the case of precision is 1 degree, the fastest umber of beats is 200, and signal to noise ratio is During 0dB, the angle measured with auxiliary vector base DOA algorithm for estimating based on uniform circular array and actual value gap are usually 1 Degree.

Claims (1)

1. the DOA estimation method of a round battle array based on noise power, it is characterised in that specifically comprise the following steps that

S1, the n times snap data separate of array received is utilized to be calculated the space correlation square R of signal；

S2, to carrying out Eigenvalues Decomposition, obtain noise subspace matrix G；

S3, carry out spectrum peak search according to noise subspace matrix G described in space correlation square R and S2 described in S1, obtain its spectral peak position Put, thus obtain Mutual coupling value.