CN108923869B

CN108923869B - Capon noise power estimation method based on orthogonal operator

Info

Publication number: CN108923869B
Application number: CN201810860790.9A
Authority: CN
Inventors: 艾晓宇; 马俊虎; 王爽; 安建成; 甘露; 廖红舒
Original assignee: University of Electronic Science and Technology of China
Current assignee: University of Electronic Science and Technology of China
Priority date: 2018-08-01
Filing date: 2018-08-01
Publication date: 2020-12-18
Anticipated expiration: 2038-08-01
Also published as: CN108923869A

Abstract

The invention belongs to the technical field of signal processing, and relates to a Capon noise power estimation method based on an orthogonal operator. According to the method, signal distribution directions are obtained according to a low-resolution spatial spectrum direction finding result, orthogonal operators orthogonal to the directions are solved, a classical Capon power estimator is reconstructed, and power estimation is carried out on any direction point of a non-signal area by using a reconstructed modified Capon power estimator to serve as noise power. The relative feature decomposition method does not need to carry out feature decomposition, and compared with a classical Capon power estimator, the method does not need to carry out calculation in all non-signal areas, can effectively reduce the operation amount, and can accurately estimate the noise power even when the signal orientation estimation has deviation due to the orthogonal characteristic.

Description

Capon noise power estimation method based on orthogonal operator

Technical Field

The invention belongs to the technical field of signal processing, and relates to a Capon noise power estimation method based on an orthogonal operator.

Background

With the development of radio technology, applications of mobile communication, broadcasting, television, navigation, remote control and remote measurement, radar and the like gradually cover the aspects of national defense and daily life of people at present, and researchers change time domain sampling into space-time sampling through a sensor array or an antenna array, so that many theoretical achievements of time domain signal processing are popularized to a space domain. When spatial spectrum estimation is performed, spatial noise power is a very important parameter, and the current spatial noise estimation methods mainly include two types: noise power estimation based on classical Capon spatial spectrum and noise power estimation based on eigenvalue decomposition.

When the arrival direction of the spatial signal is estimated, the classical Capon spatial power estimator is applied in a large scale due to the fact that the classical Capon spatial power estimator is easy to implement. However, the method has a problem of power over-estimation when performing noise estimation, and particularly under a plurality of signal scenes, spatial noise is amplified rapidly, so that a large error exists in subsequent estimation of signal power and direction of arrival, and the performance of spatial spectrum estimation is seriously affected.

Although the noise power estimation method based on feature decomposition can estimate the noise power more accurately at a high signal-to-noise ratio, the method has the following two defects compared with the Capon method: firstly, because the method involves characteristic decomposition, the method is large in calculation amount compared with a classical Capon space spectrum estimation method; secondly, the problem of aliasing of the signal subspace and the noise subspace occurs at low signal-to-noise ratio, resulting in errors in the power estimation.

The accuracy of the spatial noise power estimation can greatly affect the array signal processing effect, and in the spatial spectrum estimation, if the noise power estimation is not accurate, an error exists in the estimation of the signal power; in the beam forming process, inaccurate estimation of the noise power of the receiving end can cause great deviation between the weight design of the final self-adaptive beam forming and an ideal weight, so that the output signal-to-noise ratio of the receiving end is greatly reduced relative to a theoretical value, and the overall performance of the system is seriously influenced. Therefore, a new noise power estimator with high precision and wide application range is needed.

Disclosure of Invention

The invention provides a Capon noise power estimation method based on an orthogonal operator, which realizes high-precision spatial noise estimation on the premise of low computation. Compared with classical Capon space noise estimation, the method does not need to calculate the noise power in the whole airspace, only needs to estimate in one square point, does not need to carry out feature decomposition, and greatly reduces the operation amount. In addition, due to the fact that correction of the orthogonal operator is carried out, accurate space noise estimation under an array error scene can be achieved.

For ease of understanding, the techniques employed in the present invention are described as follows:

the conventional Capon power estimator can estimate the spatial noise power, but the noise power is overestimated due to the influence of signal residue, and the reason for overestimating is analyzed as follows. First, the power estimate at the azimuth angle θ can be expressed as

Wherein, R expresses a theoretical data covariance matrix, and is usually replaced by the data covariance matrix in practice. In a white gaussian noise scenario, the presence of only one signal can be expressed as follows:

wherein a is₀(θ₀) Indicating being at angle theta₀The signal of (a) is directed to a vector,

and

respectively representing the power of the signal and the noise, and I is an identity matrix. By matrix inversion lemma, the above formula can be expressed as

The power estimate of the Capon power estimator at azimuth θ at this time can be expressed as

Where M is the number of array elements, that is, when power estimation is performed in the azimuth θ, power over-estimation is caused because the second term in the denominator is always positive.

In order to solve the problems, the technical scheme of the invention is as follows:

a Capon space noise power estimation method based on orthogonal operator correction reconstructs a classic Capon power estimator through an orthogonal operator, and the method is used for accurately estimating space noise and is characterized by comprising the following steps:

s1, firstly, detecting signal distribution by a low-resolution spatial spectrum estimation method, and determining the number and the angle of signals in a space;

s2, assuming that L is located at the angle theta_lL1, 2.., L, and a matrix of steering vectors is a ═ a₁(θ₁),...,a₁(θ_l),...,a_L(θ_L)]；

S3, taking the orthogonality of the matrixes to obtain an orthogonal matrix and taking the first column as delta;

s4, reconstructing a (theta) in the classic Capon power estimator into

Where 0 is an all-zero matrix with dimension M-1.

And S5, carrying out power estimation on the reconstructed Capon power estimator in any direction of a non-signal area, wherein the obtained value is the space noise power.

The method has the advantages that the Capon power estimator based on the orthogonal operator can remove signal residues in the space, accurate noise power estimation is achieved, the result is irrelevant to the angle, accurate noise estimation can be obtained only by calculating a space angle point, the calculated amount is greatly reduced, and reliable priori knowledge is provided for subsequent direction finding and signal power estimation.

Drawings

FIG. 1 is a flow chart of a process for implementing the present invention;

FIG. 2 is a comparison graph of noise power estimates at different fast beat numbers;

FIG. 3 is a comparison graph of noise power estimates at different signal-to-noise ratios;

Detailed Description

The technical solution of the present invention will be further explained with reference to the accompanying drawings and examples.

Example 1

The purpose of this embodiment is to compare different noise power estimation methods under different snapshot number scenes, and verify that the method of the present invention can achieve accurate noise power estimation. In this example, the number of snapshots is 50 to 500, and noise power estimation is performed by the classical Capon method, the eigen decomposition method, and the modified Capon method, respectively, and 200 times of experiments are repeated for each number of snapshots.

The noise power estimation implementation method of the embodiment is shown in the attached figure 1. Four signals at-35, 0, 40 and 70 degrees after low resolution spatial spectrum estimation are known to have signal-to-noise ratios of about 20dB, 10dB, 20dB and a noise power of 0 dB. In the test, a 10-array element half-wavelength array-distributed uniform linear array is selected, and uncertainty of an incoming wave direction is assumed to exist, namely, errors exist in the low-resolution estimated direction, and the errors obey normal distribution with a zero mean value and a variance of 2. And taking 50 snapshots of the received data, performing autocorrelation to replace a Capon power estimator and a theoretical autocorrelation matrix in a characteristic decomposition method, and then performing orthogonality on the signal orientation guide vector after spatial spectrum estimation to construct an orthogonal operator modified Capon power estimator. It should be noted that, the classical Capon method needs to estimate and average the noise power in all regions except the signal region; the characteristic decomposition method is that after characteristic decomposition is carried out on a data autocorrelation matrix, the characteristics corresponding to signals are removed according to the sequence from big to small, and then the average value of the remaining characteristic values is taken as the estimated value of noise power; the modified Capon method only requires the calculation of power values for any point outside the signal region to be an accurate noise estimate. In the embodiment, the power estimation of the three methods is stored after each test, and after 200 tests are finished, the power estimation is respectively summed and averaged to be used as the power estimation in the snapshot scene. And then data is fetched in steps of 50 until the snapshot of 500 is taken. The comparison results of the power estimation of the three methods are shown in fig. 2, and the results show that the method provided by the invention can still accurately estimate the noise power when the signal estimation direction has deviation, the estimated noise power is not multiplied as the classic Capon method, especially when the snapshot number is more than 200, the noise power is closely attached to the estimated value of the characteristic decomposition method and approaches to the theoretical noise power by 0dB, and the method can accurately estimate the noise under the premise of low computation amount.

Example 2

The purpose of this embodiment is to compare different noise power estimation methods under different signal-to-noise ratio scenarios, and verify that the method of the present invention can achieve accurate noise power estimation. In this example, the fast beat number is 200, and noise power estimation is performed by the classical Capon method, the eigen decomposition method and the modified Capon method, respectively, and 200 times of experiments are repeated at each signal-to-noise ratio.

The noise power estimation implementation method of the embodiment is shown in the attached figure 1. Signals at-35, 0, 40 and 70 degrees are generated, the signal-to-noise ratio of the second signal ranges from-20 dB to 20dB and takes 5dB as a step size, the signal-to-noise ratios of other signals are 20dB, and the noise power is 0 dB. In the test, a 10-array element half-wavelength array-distributed uniform linear array is selected, and uncertainty of an incoming wave direction is assumed to exist, namely, errors exist in the low-resolution estimated direction, and the errors obey normal distribution with a zero mean value and a variance of 2. And performing autocorrelation on the received data to replace a Capon power estimator and a theoretical autocorrelation matrix in a characteristic decomposition method, and then performing orthogonality on the signal orientation guide vector after spatial spectrum estimation to construct an orthogonal operator modified Capon power estimator. The comparison results of the power estimation of the three methods are shown in fig. 3, and it can be seen that at 200 snapshots, along with the improvement of the signal-to-noise ratio, the noise power estimated by the method of the present invention and the eigen decomposition method closely fit and approach to the theoretical noise power of 0dB, while the noise power estimation of the classical Capon method is always around 3dB, and there is severe over-estimation on the noise, which indicates that the method still has good noise estimation performance when there is a deviation in the signal azimuth estimation.

Claims

1. A Capon noise power estimation method based on an orthogonal operator is characterized by comprising the following steps:

s1, detecting signal distribution by a low-resolution spatial spectrum estimation method, and determining the number and the angle of signals in a space;

s2, setting L position at angle theta_lL1, 2.., L, and a matrix of steering vectors is obtained as a ═ a₁(θ₁),...,a₁(θ_l),...,a_L(θ_L)]；

S3, taking the orthogonality of the matrix A to obtain an orthogonal matrix and taking the first column as delta;

s4, in the classical Capon power estimator, the power estimate at azimuth angle θ is expressed as:

r is a theoretical data covariance matrix; reconstructing a (theta) in the formula as

Where 0 is an all-zero matrix with dimension M-1;