A new method for fast estimation of direction of arrival and diffusion angle of multipath signals

By constructing a distributed source model and singular value decomposition, combined with one-dimensional spectral peak search, the problems of accuracy and complexity in direction-of-arrival estimation under multipath effects are solved, realizing fast and accurate estimation of direction-of-arrival and diffusion angle, which is suitable for radio monitoring and positioning in complex electromagnetic environments.

CN113625220BActive Publication Date: 2024-11-15西岗区富强技研汽车服务中心
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202110720072.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-06-28
Publication Date
2024-11-15
Estimated Expiration
2041-06-28

AI Technical Summary

Technical Problem

Existing direction-of-arrival (DOA) estimation algorithms suffer from low accuracy and high computational complexity under multipath effects, making it difficult to meet real-time requirements.

Method used

A distributed source model is constructed. By constructing the covariance matrix singular value decomposition and two-dimensional spatial spectrum, combined with the single-peak conjugate symmetric function and small-angle approximation theory, the generalized array manifold is reconstructed to achieve separation of the direction of arrival and the diffusion angle. The angle of arrival is estimated by one-dimensional spectral peak search.

Benefits of technology

While ensuring accuracy, the computational complexity has been reduced, and the real-time performance and reliability of the algorithm have been improved, making it suitable for radio monitoring and positioning in highly complex electromagnetic environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113625220B_ABST
    Figure CN113625220B_ABST
Patent Text Reader

Abstract

The present invention discloses a new method for fast estimation of the direction of arrival and diffusion angle of a multipath signal, which specifically includes: constructing a distributed source model of array output signals to effectively solve the problem of low accuracy of the direction of arrival estimation under the point source model; constructing a two-dimensional spatial spectrum through a noise subspace, and estimating the direction of arrival and diffusion angle of the distributed source in the two-dimensional spatial spectrum; using a single-peak conjugate symmetric function to describe the deterministic angular signal density function of the distributed source, and combining with the small-angle approximation theory, reconstructing the generalized array manifold of the uniform linear array to achieve separation of the direction of arrival and the diffusion angle; constructing a quadratic optimization function to obtain a one-dimensional spatial spectrum of the direction of arrival, and obtaining the direction of arrival estimation through the one-dimensional spatial spectrum; obtaining a one-dimensional spatial spectrum of the diffusion angle, and obtaining the diffusion angle estimation through the one-dimensional spatial spectrum. The present invention improves the reliability of the algorithm, reduces the computational complexity of the algorithm, and can be adapted to radio monitoring and positioning scenarios with high real-time requirements.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of passive radio monitoring and positioning, and in particular to a new method for quickly estimating the arrival direction and diffusion angle of a multipath signal. Background Art

[0002] The origin of array signal processing technology can be traced back to the adaptive antenna technology that used phase-locked loops for antenna tracking in the 1940s. In 1965, the adaptive notch sidelobe canceller proposed by Howells opened up a new research direction for array signal processing. Compared with traditional single-sensor signal processing, array signals have the advantages of strong anti-interference ability, flexible beam control, large signal gain and high spatial resolution. Therefore, since the 1970s, array signal processing has been widely used in civil fields such as wireless communications, intelligent conference systems, seismic exploration, wireless astronomy, machine status monitoring and fault diagnosis, as well as military fields such as sonar and radar. As the core technology and method of array signal processing, the direction of arrival (DOA) estimation method based on the characteristic subspace of the array output signal can achieve super-resolution estimation of the direction of arrival of the target source and is widely used.

[0003] In the study of super-resolution DOA estimation, it is generally assumed that the array received signal is radiated from a point target source. When the radiation surface of the target is smaller than the resolution of the array, the physical size of the target source in space can be ignored and abstracted into a point in a geometric sense. Therefore, it can be considered as a point target source (abbreviated as point source). The point source model of the signal is as follows: Figure 1 Most of the classic DOA estimation algorithms assume that there is line-of-sight propagation between the source and the observation array and that the direct wave signal strength is large. Therefore, the point source model can effectively approximate the actual propagation environment of the signal. At this time, the point source is a reasonable assumption, which greatly simplifies the complexity of DOA estimation.

[0004] However, the point source is only a mathematical assumption and is impossible to exist in practical applications. The radiation surface of the real target source always has a certain size. In addition, with the continuous development of science and technology, the electromagnetic environment in space has become more and more complex. Coupled with the rapid progress of human society, the number of high-rise buildings has increased exponentially. Figure 2 As shown in the figure, the multipath effect caused by scattering, reflection or diffraction during the propagation of radio signals in this environment makes the energy of the signal received by the sensor array present a certain dispersion in a certain range of space. The signal source at this time is called a distributed signal source (abbreviated as: distributed source), and the signal radiated by it is also called a distributed source signal. The distributed source model is as follows Figure 3As shown in the figure, the point source signal model is not applicable to such situations, which will cause the super-resolution DOA estimation algorithm based on the point source assumption to fail. In order to solve the original signal diffusion problem caused by the multipath propagation of the above signal, the distributed source model usually uses two parameters, the central DOA and the diffusion angle, to describe the spatial distribution of the signal source: the energy of the distributed source is concentrated in the spatial region with the central DOA as the center and the diffusion angle as the radius.

[0005] With the continuous advancement of science and technology, the methods of DOA estimation have been improved day by day, and a large number of high-precision and high-resolution estimation methods have been proposed. However, most of these methods are based on the classical point source model, that is, they often assume that the signal to be estimated is narrowband and satisfies the inverse of the signal bandwidth is greater than the time required for the wavefront to pass through the array space. This assumption is applicable in most cases. However, in recent years, studies have found that due to the increasingly complex electromagnetic environment in space and the increasingly tall and dense buildings on the ground, there is often no direct signal from the signal source to be estimated to the receiving sensor array, but it reaches the receiving array in the form of multipath after reflection and refraction. Therefore, in this case, ignoring the physical size of the target source to be estimated and only abstracting it as a point in a geometric sense will greatly reduce the accuracy of parameter estimation. On this basis, the concept of distributed source was proposed and successfully applied to DOA estimation.

[0006] For the distributed source situation in array signal processing, although the method based on characteristic subspace technology can realize the estimation of wave arrival direction and diffusion angle, it requires a two-dimensional spectrum peak search. Therefore, under the premise of ensuring the accuracy of parameter estimation, the computational complexity increases exponentially, which seriously affects the real-time performance of the algorithm and is therefore greatly limited in practical applications. Summary of the invention

[0007] In order to overcome the problem that the existing DOA estimation method has high computational complexity in estimating the DOA and diffusion angle of a target source due to multipath effects caused by shielding, reflection or even refraction during radio wave propagation, the present invention provides a new method for fast estimation of the DOA and diffusion angle of a multipath signal, which specifically includes:

[0008] Construct a distributed source model of the array output signal to effectively solve the problem of low accuracy of direction of arrival estimation under the point source model;

[0009] Obtaining a covariance matrix of an array output signal, performing singular value decomposition on the covariance matrix to obtain a signal subspace and a noise subspace, constructing a two-dimensional spatial spectrum through the noise subspace, and estimating a distribution source arrival direction and a diffusion angle in the two-dimensional spatial spectrum;

[0010] The deterministic angular signal density function of the distributed source is described by using a single-peak conjugate symmetric function, and combined with the small-angle approximation theory, the generalized array manifold of the uniform linear array is reconstructed to achieve the separation of the wave arrival direction and the diffusion angle.

[0011] A quadratic optimization function is constructed to obtain a one-dimensional spatial spectrum of the direction of arrival, and an estimation of the direction of arrival is obtained through the one-dimensional spatial spectrum;

[0012] A one-dimensional spatial spectrum of the diffusion angle is obtained, and a diffusion angle estimation is obtained through the one-dimensional spatial spectrum.

[0013] By adopting the above technical scheme, the present invention can achieve the following technical effects: the present invention accurately realizes the rapid estimation of the arrival direction and diffusion angle of the target source, and transforms the two-dimensional spectrum peak search with high computational complexity into a one-dimensional spectrum peak search with low complexity while ensuring the accuracy of parameter estimation, thereby improving the reliability of the algorithm, reducing the computational complexity of the algorithm, and being adaptable to radio monitoring and positioning scenarios with high real-time requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a point source model diagram;

[0015] Figure 2 It is a distributed source model diagram;

[0016] Figure 3 It is the electromagnetic wave multipath propagation diagram;

[0017] Figure 4 It is a comparison diagram of the present invention with the FLOM algorithm, the PFLOM algorithm, the CRCO algorithm and the DSPE algorithm under different generalized signal-to-noise ratio conditions;

[0018] Figure 5 It is a comparison chart of the present invention with the FLOM algorithm, the PFLOM algorithm, the CRCO algorithm and the DSPE algorithm under different snapshot number conditions. DETAILED DESCRIPTION

[0019] The embodiments of the present invention are implemented on the premise of the technical solution of the present invention, and detailed implementation methods and specific operation processes are given, but the protection scope of the present invention is not limited to the following embodiments.

[0020] Example 1

[0021] This embodiment provides a new method for quickly estimating the direction of arrival and diffusion angle of a multipath signal, which may include the following steps:

[0022] The first step is to construct a distributed source model of the array output signal;

[0023] Specifically, assume that L distributed source signals are incident on a uniform linear array, which includes M signal receiving sensors. Using the complex envelope representation, the signal vector x output by the sensor can be expressed as:

[0024]

[0025] Among them, for the i-th distribution source, s i (θ,ψ i ) is the angular signal density, Contains the direction of arrival of the target source and the diffusion angle σ i , a(θ) is the steering vector of the uniform linear array, n is the noise vector, and Θ = [-π / 2,π / 2] is the observation range of the distributed source.

[0026] Considering that the signal and noise are uncorrelated, the correlation matrix of the signal vector x can be expressed as:

[0027] R xx =E(xx H )=R ss (ψ)+R nn (2)

[0028] Among them, R ss (ψ) is the correlation matrix of the pure signal, R nn is the correlation matrix of the noise.

[0029] The angular cross-correlation kernel is defined as:

[0030]

[0031] Then R ss (ψ) can be rewritten as:

[0032]

[0033] Usually, multiple target sources are uncorrelated, so the angular cross-correlation kernel can be simplified as:

[0034] ρ ij (θ,θ′,ψ i ,ψ j )=δ ij ·μ i (θ,θ′,ψ i ) (5)

[0035] Among them, δ ij is the Kroneckerdelta function, and the angular autocorrelation kernel of the i-th target source is:

[0036]

[0037] At this time, the angular signal density of the distributed source is:

[0038] s i (θ,ψ i )=χ(θ,ψ i ) i (7)

[0039] Among them, χ(θ,ψ i ) is the deterministic angular signal density function, ζ i is a complex-valued random variable.

[0040] The second step is to obtain the covariance matrix of the array output signal, perform singular value decomposition on the covariance matrix to obtain a signal subspace and a noise subspace, construct a two-dimensional spatial spectrum through the noise subspace, and estimate the distribution source arrival direction and diffusion angle in the two-dimensional spatial spectrum, which may specifically include:

[0041] S2.1 construct the covariance matrix of the array output signal x;

[0042] S2.2 performs singular value decomposition on the covariance matrix, that is, sort the M eigenvalues ​​from large to small, divide the eigenvectors corresponding to the L largest eigenvalues ​​into one group, and divide the eigenvectors corresponding to the remaining ML smaller eigenvalues ​​into another group. Then the eigenvectors corresponding to the L largest eigenvalues ​​are the signal subspace U s , the eigenvectors corresponding to the ML smaller eigenvalues ​​are the noise subspace U n ;

[0043] S2.3 constructs a two-dimensional spatial spectrum using the noise subspace:

[0044]

[0045] S2.4 constructs an estimation function of the distribution source arrival direction and diffusion angle in the two-dimensional spatial spectrum:

[0046]

[0047] in, is the estimated value of ψ, and b(ψ) is the generalized steering vector of the uniform linear array.

[0048] The third step is to use a single-peak conjugate symmetric function to describe the deterministic angular signal density function of the distributed source, and combine it with the small-angle approximation theory to reconstruct the generalized array manifold of the uniform linear array to achieve the separation of the wave arrival direction and the diffusion angle.

[0049] Specifically, the deterministic angular density function is described using a single-peak conjugate function:

[0050]

[0051] At this point, combined with the small angle approximation theory, the generalized array manifold of the reconstructed uniform linear array is:

[0052]

[0053] Define the function h(σ) as:

[0054]

[0055] Therefore, the generalized array manifold of a uniform linear array can be further expressed as:

[0056]

[0057] Formula (13) realizes the separation of the direction of arrival and the diffusion angle.

[0058] The fourth step is to construct a quadratic optimization function to obtain a one-dimensional spatial spectrum of the direction of arrival, and obtain the direction of arrival estimate through the one-dimensional spatial spectrum;

[0059] Specifically, the two-dimensional spatial spectrum of the distribution source is reconstructed using formula (13):

[0060]

[0061] Defining functions for:

[0062]

[0063] but

[0064]

[0065] Constructing a quadratic optimization function for:

[0066]

[0067] To ensure that the solution of the quadratic optimization function in equation (17) is not zero, the quadratic optimization function can be rewritten as the following constraint form:

[0068]

[0069] Defining functions for:

[0070]

[0071] Solve for the partial derivative with respect to h(σ):

[0072]

[0073] but The solution is:

[0074]

[0075] The one-dimensional spatial spectrum of the direction of arrival estimate is:

[0076]

[0077] Performing a one-dimensional peak search on equation (22) yields the direction of arrival estimate:

[0078] The fifth step is to obtain a one-dimensional spatial spectrum of the diffusion angle, and obtain an estimation of the diffusion angle through the one-dimensional spatial spectrum;

[0079] Specifically, Substituting into equation (9), we can get the one-dimensional spatial spectrum of the diffusion angle. By searching the peak of the one-dimensional spatial spectrum, we can get the diffusion angle estimation.

[0080] Algorithm comparison: The experimental conditions set the incident angles of the two distributed source targets to be and The diffusion angles are σ1=1.5° and σ2=2.1° respectively. The present invention is compared with the FLOM algorithm, the PFLOM algorithm, the CRCO algorithm and the DSPE algorithm. Figure 4 As shown in the figure, as the generalized signal-to-noise ratio increases from 0 dB to 20 dB, although the root mean square error and the resolvable probability of various algorithms are improved, the performance of the algorithm of the present invention is always better than other comparison algorithms, and in the entire range of generalized signal-to-noise ratio changes, the two incident target signal sources can always be completely separated.

[0081] ② If Figure 5 As shown in the figure, as the number of snapshots of the array output signal sampling increases from 100 to 1000, the performance of the present invention and the FLOM algorithm, the PFLOM algorithm, and the CRCO algorithm are all improved, but excluding the influence of statistical errors, the performance of the DPSE algorithm has no significant change. However, when the number of snapshots of the present invention changes, the performance of the present invention is always the best.

[0082] The foregoing description of specific exemplary embodiments of the present invention is for the purpose of illustration and demonstration. These descriptions are not intended to limit the present invention to the precise form disclosed, and it is clear that many changes and variations can be made based on the above teachings. The purpose of selecting and describing the exemplary embodiments is to explain the specific principles of the present invention and its practical application, so that those skilled in the art can realize and utilize various different exemplary embodiments of the present invention and various different selections and changes. The scope of the present invention is intended to be limited by the claims and their equivalents.

Claims

1. A method for fast estimation of direction of arrival and diffusion angle of multipath signals, characterized in that: include: Construct a distributed source model of the array output signal, specifically: Assume that L distributed source signals are incident on a uniform linear array, which includes M signal receiving sensors. Using the complex envelope representation, the signal vector x output by the sensor is expressed as: Among them, for the i-th distribution source, s i (θ,ψ i ) is the angular signal density of the ith distributed source signal, Contains the direction of arrival of the target source and the diffusion angle σ i , a(θ) is the steering vector of the uniform linear array, n is the noise vector, Θ = [-π / 2,π / 2] is the observation range of the distributed source; Since the signal and noise are uncorrelated, the correlation matrix of the signal vector x is expressed as: R xx =E(xx H )=R ss (ψ)+R nn (2) Among them, R ss (ψ) is the correlation matrix of the pure signal, R nn is the correlation matrix of the noise; The angular cross-correlation kernel is defined as: Then R ss (ψ) can be rewritten as: Since multiple target sources are uncorrelated, the angular cross-correlation kernel is simplified to: r ij (θ,θ′,ψ i ,ψ j )=d ij ·m i (θ,θ′,ψ i ) (5) Among them, δ ij is the Kronecker delta function, and the angular autocorrelation kernel of the i-th target source is: At this time, the angular signal density of the distributed source is: s i (θ,ψ i )=χ(θ,ψ i )g i (7) Among them, χ(θ,ψ i ) is the deterministic angular signal density function, ζ i is a complex-valued random variable; The covariance matrix of the array output signal is obtained, and the covariance matrix is ​​subjected to singular value decomposition to obtain a signal subspace and a noise subspace. A two-dimensional spatial spectrum is constructed through the noise subspace, and the direction of arrival and diffusion angle of the distribution source are estimated in the two-dimensional spatial spectrum, specifically: Sort the M eigenvalues ​​from large to small, divide the eigenvectors corresponding to the L largest eigenvalues ​​into one group, and divide the eigenvectors corresponding to the remaining ML smaller eigenvalues ​​into another group. Then the eigenvectors corresponding to the L largest eigenvalues ​​are the signal subspace U s , the eigenvectors corresponding to the ML smaller eigenvalues ​​are the noise subspace U n ; The noise subspace constructs a two-dimensional spatial spectrum: The distribution source arrival direction and diffusion angle are estimated in the two-dimensional spatial spectrum, specifically: in, is the estimated value of ψ, b(ψ) is the generalized steering vector of the uniform linear array; The deterministic angular signal density function of the distributed source is described by a single-peak conjugate symmetric function, and combined with the small-angle approximation theory, the generalized array manifold of the uniform linear array is reconstructed to achieve the separation of the wave arrival direction and the diffusion angle. Specifically: At this point, combined with the small angle approximation theory, the generalized array manifold of the reconstructed uniform linear array is: Define the function h(σ) as: Combined with formula (12), the generalized array manifold of the uniform linear array is further expressed as: Formula (13) realizes the separation of the direction of arrival and the diffusion angle; A quadratic optimization function is constructed to obtain a one-dimensional spatial spectrum of the direction of arrival, and an estimation of the direction of arrival is obtained through the one-dimensional spatial spectrum; Obtain a one-dimensional spatial spectrum of the diffusion angle, and obtain the diffusion angle estimation through the one-dimensional spatial spectrum; construct a quadratic optimization function, specifically: The two-dimensional spatial spectrum of the distribution source is reconstructed using formula (13): Defining functions for: but Constructing a quadratic optimization function for: in, is the wave arrival direction; σ is the diffusion angle; U n is the noise subspace; The one-dimensional spatial spectrum of the direction of arrival is obtained, and the direction of arrival estimation is obtained through the one-dimensional spatial spectrum, specifically: To ensure that the solution of the quadratic optimization function in equation (17) is not zero, the quadratic optimization function is rewritten into the following constraint form: Defining functions for: Solve for the partial derivative with respect to h(σ): but The solution is: The one-dimensional spatial spectrum of the direction of arrival estimate is: Perform a one-dimensional spectrum peak search on equation (22) to obtain the direction of arrival estimate.

2. According to claim 1, a method for fast estimating direction of arrival and diffusion angle of multipath signals, characterized in that: Will Substitute into equation (9) to obtain the one-dimensional spatial spectrum of the diffusion angle, perform a spectrum peak search on the one-dimensional spatial spectrum, and obtain the diffusion angle estimation.

Citation Information

Patent Citations

  • Method for estimating high-accuracy arrival direction without knowing information source number

    CN104076324A

  • Uniform circular array differential phase based single distributed source DOA estimation method

    CN107907853A

  • Large-diffusion-angle multi-path signal direction-of-arrival estimation method

    CN110231589A

  • Radio wave arrival direction estimation method under condition of no direct signal

    CN110320490A

  • Two-dimensional direction finding estimation method based on dimensionality reduction Capon rooting in co-prime area array

    CN112016037A