KR20220117468A - Method and apparatus for providing robust weighted music algorithm based on mm and skip estimation method - Google Patents

Abstract

Disclosed are a method and a device for providing a robust weighted music algorithm based on an MM and a skip estimation method. According to one embodiment of the present invention, a method for providing a robust weighted music algorithm implemented through a computer device includes the steps of: determining whether an observed value is an outlier by a statistical test based on at least one of the MM estimation method and skip estimation method; and updating the observed value determined as the outlier by replacing the observed value with the MM algorithm estimate value or skip algorithm estimate value, if it is judged to be the outlier by the above statistical test.

Description

Method and apparatus for providing robust weight MUSIC algorithm based on MM and Skip estimation method

The following embodiments relate to a robust weight MUSIC algorithm based on MM and Skip estimation, and more specifically, a method and apparatus for providing a robust weight MUSIC algorithm based on MM and Skip estimation to reduce performance degradation on the estimation performance of impact noise. is about

Direction-of-arrival (DOA) estimation is a technique for determining the direction of an incident signal using measurements taken from an array of sensors, such as a linear, circular, or rectangular array. DOA estimation is of considerable interest in various research fields such as mobile communications, telecommunications, radar and sonar. The problem of DOA estimation in the presence of Gaussian noise under the far-field and narrow-band assumptions has been intensively studied in previous techniques. However, some published problems exist, and an important task among the DOA estimation problems is to estimate the DOA of a signal in the presence of thick-tailed outliers. An outlier is a data point that differs significantly from other measurements. For example, the line-of-sight (LOS) path between the source and receiver may be blocked in indoor or urban scenarios. In this case, the accuracy of the conventional DOA estimation based on the Gaussian distribution is severely degraded because these outliers do not match the Gaussian distribution.

In general, heavy tailed and measurement noise were modeled as Caucy, Student-t, or α-stable noise. Mixture distributions can also be used, but the determination of the number of mixture components is not trivial. In addition, robust statistical-based methods such as skip filters and MM estimators provide satisfactory estimation performance in the presence of heavy-tailed impact noise.

As such, in general, observation noise is modeled with a Gaussian distribution, but the effect of impulsive noise cannot be neglected. In the existing robust sample covariance matrix, a robust sample covariance matrix was obtained by giving weights to the data to have a smaller weight for outliers. In addition, MM algorithm estimates and skip algorithm estimates have been widely used in the field of robust signal processing due to their accuracy and efficiency. Although this technique performs better than when using the sample variance matrix obtained under Gaussian noise, performance improvement is required in the presence of impact noise.

Esa Ollila, “Adaptive Lasso based on joint M-estimation of regression and scale,” IEEE Trans, Signal Processing, Sep., 2001. J. Li and P. Stoica, Robust adaptive beamforming, Wiley, 2005.

The embodiments describe a method and apparatus for providing a robust weight MUSIC algorithm based on MM and Skip estimation, and more specifically, a robust angle of arrival (DOA) using a skip filter and an MM estimator to determine an outlier tolerant sample covariance matrix (SCM). ) to provide estimation algorithm technology.

Examples are robust based on the MM and Skip estimation methods in which observation values determined as outliers by statistical testing by MM or skip algorithm are replaced with MM algorithm estimates or skip algorithm estimates, and observation values determined as normal values are used as they are. An object of the present invention is to provide a method and apparatus for providing a weighted MUSIC algorithm.

According to an embodiment, a method for providing a robust weight MUSIC algorithm implemented through a computer device includes: determining whether an observation value is an outlier by a statistical test based on at least one of an MM estimation method and a skip estimation method; and when it is determined as an outlier by the statistical test, the observation value determined as an outlier is replaced with an MM algorithm estimate or a skip algorithm estimate and updated.

The method may further include the step of using the observation value determined as the normal value as it is when it is determined as the normal value by the statistical test.

분포로 모델링되는 충격 잡음의 추정 성능에 미치는 성능 저하를 감소시키도록, 업데이트된 상기 관측값은 샘플 공분산 행렬(SCM)을 결정하기 위해 사용될 수 있다.The step of updating is, Student-t distribution or

The updated observations can be used to determine a sample covariance matrix (SCM) to reduce performance degradation on the estimation performance of impact noise modeled as a distribution.

The method may further include coupling the sample covariance matrix (SCM) to a multiple signal classification (MUSIC) algorithm.

The updating may include removing the observation value determined as the outlier, calculating an average value of the remaining observation values, and replacing the observation value determined as the outlier value with the average value of the remaining observation values to be updated.

An apparatus for providing a robust weight MUSIC algorithm according to another embodiment includes: an outlier determining unit configured to determine whether an observation value is an outlier by a statistical test based on at least one of an MM estimation method and a skip estimation method; and an observation value update unit in which the observation value determined as an outlier is replaced with an MM algorithm estimation value or a skip algorithm estimation value and updated when it is determined as an outlier by the statistical test.

When it is determined as a normal value by the statistical test, the observed value determined as a normal value may be used as it is.

분포로 모델링되는 충격 잡음의 추정 성능에 미치는 성능 저하를 감소시키도록, 업데이트된 상기 관측값은 샘플 공분산 행렬(SCM)을 결정하기 위해 사용될 수 있다.The observation value update unit may include a Student-t distribution or

The updated observations can be used to determine a sample covariance matrix (SCM) to reduce performance degradation on the estimation performance of impact noise modeled as a distribution.

The sample covariance matrix (SCM) may further include a MUSIC connector connected to a multiple signal classification (MUSIC) algorithm.

The observation value update unit may update the observation value determined as the outlier by removing the observation value determined as the outlier, calculating an average value of the remaining observation values, and replacing the observation value determined as the outlier value with the average value of the remaining observation values.

According to the embodiments, observation values determined as outliers by statistical testing by the MM or skip algorithm are replaced with MM algorithm estimates or skip algorithm estimates, and observation values determined as normal values are used as they are in the MM and Skip estimation methods. A method and apparatus for providing a robust weight MUSIC algorithm based on the present invention can be provided.

According to embodiments, it is possible to provide a method and apparatus for providing a robust weight MUSIC algorithm based on MM and Skip estimation, which improves performance with respect to a thick-tailed noise distribution.

1 is a block diagram illustrating an apparatus for providing a robust weight MUSIC algorithm based on an MM and a skip estimation method according to an embodiment.
2 is a flowchart illustrating a method for providing a robust weight MUSIC algorithm based on MM and Skip estimation according to an embodiment.
3 is a graph showing a MUSIC result for a Student-t (degree of freedom = 3) distribution according to an embodiment.
4 is a graph illustrating a MUSIC result in SS noise (=1) according to an embodiment.
5 is a graph illustrating an RMSE estimation result of an angle of arrival in heavy-tailness according to an embodiment.
6 is a graph illustrating an RMSE estimation result of an angle of arrival according to the number of receivers according to an embodiment.
7 is a graph illustrating the number of samples and an RMSE estimation result according to an embodiment.

Hereinafter, embodiments will be described with reference to the accompanying drawings. However, the described embodiments may be modified in various other forms, and the scope of the present invention is not limited by the embodiments described below. In addition, various embodiments are provided in order to more completely explain the present invention to those of ordinary skill in the art. The shapes and sizes of elements in the drawings may be exaggerated for clearer description.

In general, observation noise is modeled with a Gaussian distribution, but the reality is that the effect of impulsive noise cannot be ignored. When it is determined as an outlier by statistical test, a more accurate covariance matrix can be obtained because high-quality data is used by replacing it with the MM estimate or the skip algorithm estimate.

분포로 모델링되는 충격 잡음의 추정 성능에 미치는 성능 저하를 감소시키기 위하여 MM 및 skip 추정법에 기반한 강인한 알고리즘을 제안한다. 즉, MM 또는 skip 알고리즘에 의하여 통계적 검정에 의하여 이상치(outlier)로 판정된 관측값은 MM 알고리즘 추정값이나 skip 알고리즘 추정값으로 대체되고 정상치로 판정된 관측값은 그대로 사용된다. In the example below, the Student-t distribution and

We propose a robust algorithm based on MM and skip estimation to reduce the performance degradation on the estimation performance of impact noise modeled as a distribution. That is, the observation value determined as an outlier by statistical test by the MM or skip algorithm is replaced with the MM algorithm estimate value or the skip algorithm estimate value, and the observation value determined as a normal value is used as it is.

According to the examples, it was confirmed through computer simulation that a noise distribution with a thick tail exhibited superior performance compared to the conventional method.

1 is a block diagram illustrating an apparatus for providing a robust weight MUSIC algorithm based on an MM and a skip estimation method according to an embodiment.

Referring to FIG. 1 , an apparatus 100 for providing a robust weight MUSIC algorithm according to an embodiment may include an outlier determination unit 110 and an observation value update unit 120 . According to an embodiment, the apparatus 100 for providing a strong weight MUSIC algorithm may further include a MUSIC connection unit 130 .

The outlier determination unit 110 may determine whether an observation value is an outlier by a statistical test based on at least one of an MM estimation method and a skip estimation method.

When the observation value update unit 120 is determined to be an outlier by statistical testing, the observation value determined as an outlier may be replaced with an MM algorithm estimate or a skip algorithm estimate and updated.

In addition, in the MUSIC connector 130 , a sample covariance matrix (SCM) may be connected to a multiple signal classification (MUSIC) algorithm.

2 is a flowchart illustrating a method for providing a robust weight MUSIC algorithm based on MM and Skip estimation according to an embodiment.

Referring to FIG. 2 , in the method for providing a robust weight MUSIC algorithm implemented through a computer device according to an embodiment, it is determined whether an observation value is an outlier by a statistical test based on at least one of an MM estimation method and a skip estimation method. Determining (S110), and when it is determined as an outlier by statistical test, the observation value determined as an outlier is replaced with an MM algorithm estimate or a skip algorithm estimate and updated (S120).

The method may further include the step of using the observed value determined as the normal value as it is when it is determined as the normal value by the statistical test.

In addition, the method may further include a step S130 in which the sample covariance matrix (SCM) is connected to a multiple signal classification (MUSIC) algorithm.

A method of providing a strong weight MUSIC algorithm implemented through a computer device according to an embodiment may be described using the apparatus 100 for providing a strong weight MUSIC algorithm described with reference to FIG. 1 as an example.

In step S110 , the outlier determination unit 110 may determine whether an observation value is an outlier by a statistical test based on at least one of an MM estimation method and a skip estimation method.

In step S120 , when the observation value update unit 120 is determined to be an outlier by statistical test, the observation value determined as an outlier may be replaced with an MM algorithm estimate or a skip algorithm estimate and updated. On the other hand, when it is determined as a normal value by a statistical test, the observed value determined as a normal value may be used as it is.

분포로 모델링되는 충격 잡음의 추정 성능에 미치는 성능 저하를 감소시키도록, 업데이트된 관측값은 샘플 공분산 행렬(SCM)을 결정하기 위해 사용될 수 있다.The observation value update unit 120 is a Student-t distribution or

The updated observations can be used to determine a sample covariance matrix (SCM) to reduce performance degradation on the estimation performance of impact noise modeled as a distribution.

Also, the observation value update unit 120 may update the observation values determined as outliers by removing the observation values determined as outliers, obtaining an average value of the remaining observation values, and replacing the observation values determined as outliers with the average values of the remaining observation values.

In step S130 , the MUSIC connector 130 may connect a sample covariance matrix (SCM) to a multiple signal classification (MUSIC) algorithm.

Therefore, when it is determined as an outlier by a statistical test, a more accurate covariance matrix can be obtained because good quality data is used by substituting the MM estimate or the skip algorithm estimate.

Hereinafter, a method and apparatus for providing a robust weight MUSIC algorithm based on MM and Skip estimation according to an embodiment will be described in more detail.

Embodiments may use a skip filter and an MM estimator to remove the adverse effect of outliers and determine a robust sample covariance matrix (SCM). An outlier tolerant sample covariance matrix (SCM) can be determined based on the MM estimate, and the filter can be skipped by replacing observations predicted as outliers with robust statistics (eg MM estimates). This sample covariance matrix (SCM) is then connected to a multiple signal classification (MUSIC) algorithm. It should be noted that this sample covariance matrix (SCM) can be connected to any subspace estimator, such as estimating signal parameters via minimum standard or ESPRIT (rotational invariance technique).

The embodiments can be summarized as follows.

Embodiments apply MM estimator and skip filter in DOA estimation and beamforming context. MM estimator and skip filter have not yet been utilized to solve DOA estimation and beamforming problems.

Examples are based on statistical tests that samples predicted as outliers are replaced with MM estimates and these updated samples are used to determine a robust sample covariance matrix (SCM). This sample covariance matrix (SCM) is used by the MUSIC method. Below, this algorithm is abbreviated as the MUSIC-MM-sub algorithm.

In the second algorithm below, samples predicted as outliers are removed based on a statistical test and the remaining samples are averaged. Then, instead of the outliers, these averages are substituted and these updated snapshots are used to determine the sample covariance matrix (SCM). Then, this sample covariance matrix (SCM) is adopted for the MUSIC method. Hereinafter, this algorithm is abbreviated as the MUSIC-MM-skip method.

The purpose of the DOA estimation method in narrowband and far-domain assumptions is to accurately predict the angle of incidence of a point object so that the mean square error (MSE) is minimized or the signal and noise subspaces are orthogonal. In the context of the same linear arrangement (ULA) as the LOS/NLOS mixed DOA estimation, the measurement equation is determined as

는 MХK 행렬이고

로 정의되며

는 MХ1 배열 조향 벡터를 나타낸다. here,

is the MХK matrix and

is defined as

denotes the MХ1 array steering vector.

이고,

는 신호의 파장이며, d는 요소 공간간 거리, j는 가상 유닛, sj는 j번째 신호 스냅샷으로

로 정의되며 nj는 j번째 복소값 잡음 벡터이다. 잡음 벡터는 LOS 환경에서 원형 대칭 가우시안으로 모델링된 반면, LOS/NLOS 환경에서는 Cauchy, Student-t 또는 α-stable 잡음으로 표현되었다. 이미터[emitter] 수(K)는 이미 알려진 것으로 가정한다. 여기서 목표는 heavy-tailed 잡음에서 이미터의 DOA를 정확하게 추정하는 것이다. 본 실시예에서, 벡터는 소문자 굵은 글꼴 문자, 대문자 굵은 글꼴 문자가 있는 행렬, 연산자

는 벡터/매트릭스 전치를,

는 벡터/매트릭스 복소 결합 전치를 나타낸다.

ego,

is the wavelength of the signal, d is the distance between element spaces, j is the virtual unit, and sj is the j-th signal snapshot.

, where nj is the j-th complex-valued noise vector. The noise vector was modeled as a circularly symmetric Gaussian in the LOS environment, whereas it was expressed as Cauchy, Student-t, or α-stable noise in the LOS/NLOS environment. It is assumed that the number of emitters (K) is already known. The goal here is to accurately estimate the emitter's DOA in heavy-tailed noise. In this embodiment, the vector is a matrix with lowercase bold font characters, uppercase bold font characters, operators

is the vector/matrix transpose,

represents the vector/matrix complex bond transpose.

Existing robust covariance matrices and MM estimators are described.

Kendall's Tau Covariance Matrix (TCM)

The tau covariance matrix (TCM) is defined as follows.

이다.here,

to be.

The tau covariance matrix (TCM) is known to produce convergence estimates for the eigenvectors and consequently the subspace. This covariance matrix estimate can be connected to any sub-spatial estimator (eg MUSIC). In addition, it is possible to process correlated signals when performing spatial smoothing.

t1M covariance matrix

The t1M covariance matrix is determined as follows.

이다. 가중치 함수가 u(t)인 M 추정기이다. 이 가중치 함수의 선택은 복잡한 다변량 Cauchy 잡음이 있는 경우 최대우도 추정기(MLE)를 산출한다. 또한, 이 t1M 공분산 행렬 추정값을 사용하면 다변량 가우시안 샘플의 경우에 효율적이다. t1M 공분산 행렬은 닫힌 형태가 없기 때문에 고정점 알고리즘을 사용하여 재귀적으로 구해진다. 고정점 반복은 함수의 고정점이 계산되는 기법이다. 보다 구체적으로, 실제 값을 갖는 실수에 의존하며, f 도메인에서 초기 점 xo가 주어지면, 고정점 반복은 xj+1 = f(xj), j =1,...P이며, 이것은 점 x로 수렴될 것으로 예상된다.here,

to be. An M estimator whose weight function is u(t). The choice of this weighting function yields a maximum likelihood estimator (MLE) in the presence of complex multivariate Cauchy noise. Also, using this t1M covariance matrix estimate is efficient in the case of multivariate Gaussian samples. Since the t1M covariance matrix has no closed form, it is obtained recursively using the fixed-point algorithm. Fixed-point iteration is a technique in which the fixed point of a function is computed. More specifically, it depends on real-valued real numbers, given an initial point xo in the f domain, the fixed-point iteration is xj+1 = f(xj), j = 1,...P, which is is expected to converge.

Tyler's covariance matrix

Tyler's covariance matrix is defined as

It is an M estimator whose weight function is u(t)=M/t. An important property of Tyler's estimate of covariance is that the asymptotic distribution is not distributed with respect to the ranks of the elliptical population. In addition, it is kept to a fine sample size. Since Tyler's covariance matrix has no closed form, it is determined recursively using a fixed-point algorithm.

MM Estimator

The MM estimate for the position parameter can be considered as a weighted average of the observations. Embodiments use a constrained cost function ρ. The widely used cost functions belong to the bisquare family of functions.

일 때 미분

를 가지고, I(·)는 지시 함수를 나타내며,

은 대부분의 경우에 존재한다. 우리의 목표는 다음과 같이 위치 모수(μ)와 관련하여 비용 함수(ρ)를 최소화하는 것이다.

differential when

, I(·) denotes an indication function,

is present in most cases. Our goal is to minimize the cost function (ρ) with respect to the position parameter (μ) as follows:

Here, σ can be estimated as the standard deviation of xi. If ρ is differentiable, differentiating [Equation 6] with respect to the output μ can be expressed as follows.

라 가정할 수 있다.At this time,

can be assumed.

Then, [Equation 7] can be written as follows.

)는 [수학식 9]의 양쪽에 존재하는

처럼 반복적으로 구한다.Or equally, this value represents the estimate as a weighted average. Since W(x) is a non-increasing function of IxI, the external observations are given less weight. position estimate (

) is present on both sides of [Equation 9]

Iteratively, as

Here, the scale parameter σ is estimated based on the S estimator. The MM estimate is summarized in Algorithm 1 below.

[Table 1]

Hereinafter, the proposed DOA estimation and beamforming method will be described.

In general, observation noise for stationary values is assumed to be circularly symmetric Gaussian noise in the context of DOA estimation. However, due to the impact characteristics of artificial electromagnetic interference and natural noise, it may be a non-Gaussian noise with a heavy tailed distribution, such as Cauchy, Student-t, or α-stable noise. Therefore, we can provide a robust DOA estimation algorithm to solve the non-Gaussian heavy tailed impact noise. Below, the proposed MM method-based DOA estimation technique is described in detail.

MUSIC-MM-sub DOA estimation algorithm

에 기초한 MM 추정값의 실수 부분은 다음과 같이 결정된다. 즉, 실수 부분의 MM 알고리즘 추정 값을 다음과 같이 나타낼 수 있다.Here, the MUSIC-MM-sub algorithm is described in detail.

The real part of the MM estimate based on That is, the MM algorithm estimation value of the real part can be expressed as follows.

이고,

는 i번째 센서에 대한 S-추정기에 기초한 표준 편차 추정값이다. 즉, S 추정값에 기반한 표준편차이다. 전술한 MM 추정값은 통계적 검정에 의하여 이상치로 추정된 값(이상치로 예측된 관측값)을 대신하여 다음과 같이 대체된다. here,

ego,

is the standard deviation estimate based on the S-estimator for the i-th sensor. That is, the standard deviation based on the S estimate. The above-mentioned MM estimate is substituted as follows in place of the value estimated as an outlier (observed value predicted as an outlier) by a statistical test.

Repeat the above-mentioned procedure to calculate the imaginary part. After that, the real part and the imaginary part are added as follows.

Then, a sample covariance matrix (SCM) is obtained as follows.

이다. 그 후에 전술한 관찰을 위한 공간 스펙트럼은 다음과 같이 정의된다. 즉, 방위각은 다음과 같은 비용 함수를 최대화하는 방위로 결정된다.here,

to be. Then the spatial spectrum for the above-mentioned observation is defined as follows. That is, the azimuth is determined as the azimuth that maximizes the following cost function.

은

의 내림차순 고유값과 관련된 잡음 부분 공간(잡음 부공간)을 나타내며,

는 조향 벡터이고, K는 알려진 신호 수이다. 마지막으로, DOA 추정값은 공간 스펙트럼의 정점으로 결정된다.here,

silver

Represents the noise subspace (noise subspace) associated with the descending eigenvalues of

is the steering vector, and K is the known number of signals. Finally, the DOA estimate is determined as the peak of the spatial spectrum.

MUSIC-MM-skip DOA estimation algorithm

In the MUSIC-MM-skip DOA estimation algorithm, in contrast to the case of the MUSIC-MM-sub method, samples predicted as outliers in the skip method are removed from the observation set.

의 평균값은 다음과 같다.observation

The average value of is as follows.

은

의 크기를 나타내며, 이는

에 속하고

는 정상치로 예측되는 관측 집합(정상치로 판정된 데이터 집합)이다. 그런 다음, 다음과 같이 이상치로 예측된 샘플(이상치로 판정된 관측값) 대신에

가 삽입된다. 즉, 다음과 같이 데이터 집합이 업데이트된다.here,

silver

represents the size of

belong to

is a set of observations predicted to be stationary (a data set determined to be stationary). Then, instead of samples predicted as outliers (observations determined to be outliers),

is inserted That is, the data set is updated as follows.

The aforementioned procedure is repeated for the imaginary part. After that, the real part and the imaginary part are added as follows. That is, the imaginary part is also obtained by the same procedure as above, and the direction that maximizes the cost function is determined as the final azimuth.

The spatial spectrum is defined as

은

의 내림차순 고유값과 관련하여 배열된 잡음 부분 공간을 나타낸다. 마지막으로, DOA 추정값은 공간 스펙트럼의 피크로 결정된다.here,

silver

Represents the noise subspace arranged with respect to the descending eigenvalues of . Finally, DOA estimates are determined as peaks in the spatial spectrum.

Proposed Beamforming Algorithm

A minimum variance distortion-free response (MVDR) beamformer (Non-Patent Document 2) is used and utilizes the proposed strong covariance. Consequently, the spatial spectrum is expressed as

Convergence properties of MM estimators

The convergence of the MM algorithm depends on two steps. The first step is the S-estimation (M-estimation of the scale parameter), and the second step is the M-estimation of the position parameter. Therefore, to confirm the convergence of the MM estimate, it is necessary to prove the convergence of the M estimate.

3 is a graph showing a MUSIC result for a Student-t (degree of freedom = 3) distribution according to an embodiment.

As shown in FIG. 3 , it can be confirmed that both the existing method and the proposed method find the angle of arrival well. Since this is a distribution where heavy-tailness is not too severe, it can be seen that the existing algorithm finds the angle of arrival relatively well.

4 is a graph illustrating a MUSIC result in SS noise (=1) according to an embodiment.

As shown in FIG. 4 , it can be seen that the azimuth estimation performance is severely degraded when the conventional method is used. However, it can be confirmed that the angle of arrival can be accurately found when the proposed method (the present invention) is used.

5 is a graph illustrating an RMSE estimation result of an angle of arrival in heavy-tailness according to an embodiment.

As shown in FIG. 5 , it was confirmed that the RMSE of the proposed method was smaller as the heavy-tailness was severe compared to the conventional method. This phenomenon is due to the robustness of the MM estimate and the skip algorithm estimate.

6 is a graph illustrating an RMSE estimation result of an angle of arrival according to the number of receivers according to an embodiment.

As shown in FIG. 6 , it can be seen that the RMSE of the MUSIC-MM-sub and MUSIC-MM-skip algorithms is smaller than that of the existing algorithms. It can be seen that the RMSE decreases as the number of receivers increases.

7 is a graph illustrating the number of samples and an RMSE estimation result according to an embodiment.

As shown in FIG. 7 , the RMSE of the MUSIC-MM-sub and MUSIC-MM-skip algorithms were the smallest in this experimental result as well.

As described above, according to the embodiments, it can be confirmed that the performance is improved with respect to a noise distribution with a thicker tail than that of the existing technology.

The device described above may be implemented as a hardware component, a software component, and/or a combination of the hardware component and the software component. For example, devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), It may be implemented using one or more general purpose or special purpose computers, such as a programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that may include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

The software may comprise a computer program, code, instructions, or a combination of one or more thereof, which configures a processing device to operate as desired or is independently or collectively processed You can command the device. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or device, to be interpreted by or to provide instructions or data to the processing device. may be embodied in The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

As described above, although the embodiments have been described with reference to the limited embodiments and drawings, various modifications and variations are possible by those skilled in the art from the above description. For example, the described techniques are performed in a different order than the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (10)

A method for providing a strong weight MUSIC algorithm implemented through a computer device, the method comprising:
determining whether an observation value is an outlier by a statistical test based on at least one of an MM estimation method and a skip estimation method; and
When it is determined as an outlier by the statistical test, the observation value determined as an outlier is replaced with an MM algorithm estimate or a skip algorithm estimate and updated;
A method of providing a robust weight MUSIC algorithm, including. According to claim 1,
When it is determined that the value is normal by the statistical test, using the observation value determined as the normal value as it is
A method of providing a strong weight MUSIC algorithm further comprising a. According to claim 1,
The updating step is
Student-t distribution or

The updated observations are used to determine a sample covariance matrix (SCM) to reduce performance degradation on the estimation performance of impact noise modeled as a distribution.
A method for providing a strong weight MUSIC algorithm, characterized in that. 4. The method of claim 3,
coupling the sample covariance matrix (SCM) to a multiple signal classification (MUSIC) algorithm.
A method of providing a strong weight MUSIC algorithm further comprising a. According to claim 1,
The updating step is
After the observation value determined as the outlier is removed, the average value of the remaining observation values is obtained, and the observation value determined as the outlier value is replaced with the average value of the remaining observation values and updated
A method for providing a strong weight MUSIC algorithm, characterized in that. an outlier determination unit for determining whether an observation value is an outlier by a statistical test based on at least one of an MM estimation method and a skip estimation method; and
When it is determined as an outlier by the statistical test, the observation value determined as an outlier is replaced with an MM algorithm estimate or a skip algorithm estimate and updated by an observation value update unit
Including, strong weight MUSIC algorithm providing apparatus. 7. The method of claim 6,
When it is determined as a normal value by the statistical test, using the observed value determined as a normal value as it is
A device for providing a strong weight MUSIC algorithm, characterized in that 8. The method of claim 7,
The observation value update unit,
Student-t distribution or

The updated observations are used to determine a sample covariance matrix (SCM) to reduce performance degradation on the estimation performance of impact noise modeled as a distribution.
A device for providing a strong weight MUSIC algorithm, characterized in that 9. The method of claim 8,
MUSIC connector where the sample covariance matrix (SCM) is connected to a multiple signal classification (MUSIC) algorithm
Further comprising, a strong weight MUSIC algorithm providing apparatus. 7. The method of claim 6,
The observation value update unit,
After the observation value determined as the outlier is removed, the average value of the remaining observation values is obtained, and the observation value determined as the outlier value is replaced with the average value of the remaining observation values and updated
A device for providing a strong weight MUSIC algorithm, characterized in that

Images