Article

A Robust Direction-of-Arrival (DOA) Estimator for Weak Targets

Based on a Dimension-Reduced Matrix Filter with Deep Nulling
and Multiple-Measurement-Vector Orthogonal Matching Pursuit
Shoudong Wang, Haozhong Wang *, Zhaoxiang Bian, Susu Chen, Penghua Song, Bolin Su and Wei Gao

School of Marine Technology, Ocean University of China, Qingdao 266100, China; [email protected] (S.W.);
[email protected] (Z.B.); [email protected] (S.C.); [email protected] (P.S.);
[email protected] (B.S.); [email protected] (W.G.)
* Correspondence: [email protected]

Abstract: In the field of target localization, improving direction-of-arrival (DOA) estima-

tion methods for weak targets in the context of strong interference remains a significant
challenge. This paper presents a robust DOA estimator for localizing weak signals of
interest in an environment with strong interfering sources that improve passive sonar DOA
estimation. The presented estimator combines a multiple-measurement-vector orthogonal
matching pursuit (MOMP) algorithm and a dimension-reduced matrix filter with deep
nulling (DR-MFDN). Strong interfering sources are adaptively suppressed by employing
the DR-MFDN, and the beam-space passband robustness is improved. In addition, Gaus-
sian pre-whitening is used to prevent noise colorization. Simulations and experimental
results demonstrate that the presented estimator outperforms a conventional estimator
based on a dimension-reduced matrix filter with nulling (DR-MFN) and the multiple signal
classification algorithm in terms of interference suppression and localization accuracy.
Moreover, the presented estimator can effectively handle short snapshots, and it exhibits
superior resolution by considering the signal sparsity.
Academic Editor: Xuebo Zhang,
Haiyong Xu, Jingyue Lu, Lei Zhang,
Marc Pinto, Farman Ali
Keywords: matrix filter; DOA estimation; weak target; pre-whitening operation; sparse
representation
Received: 3 December 2024
Revised: 24 January 2025
Accepted: 27 January 2025
Published: 30 January 2025

Citation: Wang, S.; Wang, H.; Bian, Z.;

1. Introduction
Chen, S.; Song, P.; Su, B.; Gao, W. Matrix filters [1–4] have been widely used in sonar and radar applications because
A Robust Direction-of-Arrival (DOA) of their exceptional data filtering and signal separation capabilities in the spatial domain.
Estimator for Weak Targets Based on a A matrix filter projects the observation space onto a subspace by linear transformation,
Dimension-Reduced Matrix Filter
which passes the signals in the sector of interest (passband) while suppressing the out-of-
with Deep Nulling and Multiple-
Measurement-Vector Orthogonal
sector (stopband) interference. Vaccaro and Harrison [1] designed a conventional matrix
Matching Pursuit. Remote Sens. 2025, filter (CMF) in the frequency domain. MacInnes [5] improved passband robustness in
17, 477. https://doi.org/10.3390/ the beam space by applying the least mean square criterion and solving it using con-
rs17030477 vex optimization. However, this method fails to control the stopband attenuation (SA).
Copyright: © 2025 by the authors. Yan et al. [6,7] proposed a second-order cone programming model for CMFs under the
Licensee MDPI, Basel, Switzerland. stopband constraint, a passband least-squares criterion, and passband minimax criterion.
This article is an open access article Han et al. [8] proposed three optimization problems using the Lagrange multiplier method
distributed under the terms and to obtain matrix filters with adjustable SA and passband robustness. Hassanien et al. [9]
conditions of the Creative Commons
proposed a dimension-reduced conventional matrix filter by constraining the stopband
Attribution (CC BY) license
(https://creativecommons.org/
and minimizing the difference between actual and quiescent response matrix filters. These
licenses/by/4.0/). variants of the CMF approach can strictly control the attenuation of the stopband while

Remote Sens. 2025, 17, 477 https://doi.org/10.3390/rs17030477

Remote Sens. 2025, 17, 477 2 of 24

maintaining the passband performance. However, data cannot be effectively filtered when
strong interfering sources are present in the environment. To solve the problem of weak
target direction-of-arrival (DOA) estimation in an environment with strong interfering
sources, adaptive matrix filters that can provide nulls toward the direction of the interfering
sources and attenuate them adaptively were proposed [9–11]. The presence of noise and
perturbation limits the depth of the nulls in the direction of the interfering sources and
degrades the passband’s robustness. Considering the effect of noise after matrix filtering,
Yan [7] limited the power of the noise to reduce its impact on subsequent processing. Nev-
ertheless, non-orthogonal matrix filters transform white noise into colored noise; therefore,
the estimation of weak targets is challenging. Hassanien [9] prevented the transformation
of noise into colored noise by using an orthogonal pre-whitening (OP) operation. How-
ever, this method alters the beam-space characteristics of the matrix filter, particularly by
reducing the SA, leading to insufficient interference suppression. To maintain better spatial
filtering performance and orthogonality of the matrix filter, Wang [12,13] achieved better
performance by synthesizing a set of orthogonal eigenbeams for spatial filtering.
Moreover, the multiple signal classification (MUSIC) algorithm [14], typically em-
ployed for DOA estimation with matrix filters, has exhibited poor performance in scenarios
with coherent sources or short snapshots [7–9,15–17]. Recently, sparse reconstruction tech-
niques have demonstrated great advantages in solving signal recovery problems and have
been widely used in the field of array signal processing [18–25], for example, the sparse spec-
trum fitting (SpSF) algorithm [26] and matching pursuit algorithms [18–20]. Yang [10,27]
proposed estimators based on the SpSF algorithm and matrix filters. However, with these
methods, noise power must be estimated. Moreover, determining the selection of the
regularization parameter is challenging in practice.
This paper aims to introduce an improved DOA estimator for weak targets. First,
considering that white noise at the input may be transformed into colored noise, this study
utilizes a Gaussian matrix for the pre-whitening operation. The objective is to prevent the
transformation of white noise into colored noise while preserving the characteristics of the
beam space. Second, the DOA estimation method is presented based on matrix filters and
the multiple-measurement-vector orthogonal matching pursuit (MOMP) algorithm [28].
The MOMP is an extended form of the orthogonal matching pursuit (OMP) algorithm [29]
under the multiple-measurement-vector (MMV) model. Next, based on the design method
of a dimension-reduced matrix filter with nulling (DR-MFN) proposed in [9], a dimension-
reduced matrix filter with deep nulling (DR-MFDN) is presented for DOA estimation in
this study. With a projection matrix that enhances the interference component in the array
data, the presented matrix filter can provide deep nulls and a lower passband response
error when the array output energy is minimized. Additionally, reducing the beam-space
dimension can simplify the problem dimension, reducing the computational complexity.
An estimator based on the DR-MFDN and MOMP algorithms with the Gaussian pre-
whitening (GP) operations can then be obtained. An estimator based on the DR-MFN
and MUSIC algorithms is used to verify the effectiveness of the pre-whitening operation.
Finally, the DOA estimation performances of four combination estimators, comprising two
DOA estimation algorithms (MUSIC and MOMP) and two matrix filters (DR-MFN and
DR-MFDN), are compared using the preferred pre-whitening operation.
In this study, we extended our preliminary work in the conference paper [30]. Al-
though the conference paper proposed the concept of DR-MFDN and demonstrated its
preliminary performance, due to scope limitations, the paper failed to provide detailed
algorithm descriptions, extensive simulation and experimental verification, and analysis of
computation time. Simulations and experimental results demonstrate that the presented
adaptive matrix filter exhibits greater robustness in the passband and outperforms interfer-
Remote Sens. 2025, 17, 477 3 of 24

ence suppression. Moreover, the GP operation is better adapted for processing beam-space
snapshots than the OP operation. The estimator based on adaptive matrix filters and the
MOMP algorithm with the GP operation can achieve high localization accuracy.

2. Materials and Methods

2.1. Signal Model and Pre-Whitening Operation
In this section, sparse models of element-space and beam-space signals are established,
and a novel pre-whitening operation based on a Gaussian matrix is introduced.

2.1.1. Element-Space Signal Model

Consider K far-field signals with frequency f impinging on a uniform linear array
with M elements distributed along the x direction with spacing d. The received signal at the
mth element of the array can be considered as the sum of the finite plane waves as follows:

L 2π f
pm ( f ) = ∆θ ∑ s ( f , θ j ) ei ( c xm sin θ j )
(1)
l =1

where c is the velocity of sound; s( f , θ j ) is the plane wave amplitude in θ j ; ∆θ = π

J, with J
2 ) ∆θ
1
being the number of angular discretizations, and θ j = ( j − − = 1, · · · , J. π
2 h, j
The wideband signals received by the array were partitioned into N segments, and an
L-point FFT was applied to each segment. The array output in the frequency domain can
be expressed in the discrete form shown in Equation (1) as follows:

pl (n) = A( f l , Θ)sl (n) + el (n), n = 1, 2, · · · , N (2)

where A( f l , Θ) ∈ C M× J is the steering matrix, Θ = [θ1 , · · · , θ J ] T is the vector that comprises

J candidate directions, and the superscript T is the transpose operator. Moreover, pl (n) ∈
C M×1 , sl (n) ∈ C J ×1 , and el (n) ∈ C M×1 represent the FFT coefficients of the received data,
wideband signals, and additive noise at the nth segment, respectively. The steering matrix
is given by
A ( f l , Θ ) = a ( f l , θ1 ), · · · , a ( f l , θ j ), · · · , a ( f l , θ J )
 
(3)

where 2π f l 2π f l
( M−1)d sin θ j ) T
a( f l , θ j ) = [1, ei( c d sin θ j )
, · · · , ei ( c ] (4)

Assuming that el is white noise, the array covariance matrix Rl ∈ M× M of the array
output in the frequency domain is expressed as follows [14,31]:

Rl = E[pl (n)plH (n)] = A( f l , Θ)Rls A H ( f l , Θ) + Rle = A( f l , Θ)Rls A H ( f l , Θ) + σl 2 I M (5)

where E[ · ] denotes the expectation operation; superscript H represents the Hermitian

transpose; Rls = E[sl (n)slH (n)] ∈ CK ×K , Rle = E[el (n)elH (n)] ∈ C M× M ; σl 2 is the variance
of the white noise; and I M is an identity matrix of dimension M.

2.1.2. Beam-Space Signal Model and Pre-Whitening Operation

Conventional array processing methods cannot effectively estimate the DOAs of
weak targets under strong interfering sources. Therefore, transforming the signal into
the designed beam-space is necessary. The beam-space snapshot vector zl (n) ∈ C M×1
in reduced dimensions (M′ < M is the beam-space dimension) is defined as a linear
transformation of the frequency-domain snapshot pl (n) as follows:

zl (n) = GlH pl (n) = GlH A( f l , Θ)sl (n) + GlH el (n) (6)

Remote Sens. 2025, 17, 477 4 of 24

where GlH is the matrix filter corresponding to the frequency f l .

The beam-space array covariance matrix Rz is given by

Rz = E[zl (n)zlH (n)] = GlH Rl Gl = GlH A( f l , Θ)Rls A H ( f l , Θ)Gl + σl 2 GlH Gl (7)

If Gl is not orthogonal, it transforms white noise into colored noise, which is not
convenient for subsequent processing. Thus, an additional noise pre-whitening operation
is required. Under a conventional OP operation, Equation (6) can be rewritten as [16,17,32]

ẑl (n) = (GlH Gl )−1/2 GlH pl (n) (8)

The application of an OP operation alters the characteristics of the beam space. Con-
sidering that the SVD decomposition of the matrix filter is performed with Gl = uεv H and
Gl GlH = uε2 u H , the beam-space response h( f l , θ ) by the matrix filter Gl is

GlH a( f l , θ ) 1 H
h( f l , θ ) = 2
=[ a ( f l , θ )uε2 u H a( f l , θ )]1/2 (9)
∥a( f l , θ )∥2 M

With an OP operation, Equation (9) can be rewritten as

−1/2
(GlH Gl ) GlH a( f l , θ ) 1 H
h( f l , θ ) =
b 2
=[ a ( f l , θ )uu H a( f l , θ )]1/2 (10)
∥a( f l , θ )∥2 M

Considering Equations (9) and (10), the OP operation normalizes the eigenvalues of
the matrix GlH , modifying the beam space. To ensure that the beam-space transformation
preserves the original characteristics and prevents white noise from being transformed
into colored noise, a Gaussian matrix was employed for the pre-whitening operation.
The frequency-domain snapshot is transformed into

z̃l (n) = G a GlH pl (n) (11)

′ ′
where G a = [g a1 , g a2 , · · · , g aM′ ] T ∈ C M × M is a Gaussian matrix, with each vector obey-
ing g ai (m) ∼ N(0, σl2 ), i | m = 1, 2, · · · , M′ and any two arbitrary vectors obeying
cov(g ai , g aj )(i̸= j,1≤i,j≤ M′ ) = 0.
With the GP operation, the beam-space response can be reformulated as follows:

G a GlH a( f l , θ )
 1/2
1 H
ĥ( f l , θ ) = 2
= a ( f l , θ )Gl G aH G a GlH a( f l , θ ) = h( f l , θ ) (12)
∥a( f l , θ )∥2 M

According to Appendix A, noise covariance matrix Rle is transformed as follows [33]:

G a GlH Rle Gl G aH = σl2 I M′ (13)

where I M′ is an identity matrix of dimension M′ .

It is important to note that perfectly uncorrelated Gaussian matrices do not exist in
practice. However, Gaussian matrices statistically approximate this property. Therefore,
extensive experimentation is needed to approach the ideal performance of a Gaussian
matrix, with the average results obtained reflecting the desired outcome.
Remote Sens. 2025, 17, 477 5 of 24

2.2. Robust DOA Estimator Based on a Matrix Filter and MOMP Algorithm
In this section, we present a robust DOA estimator based on a matrix filter and the
MOMP algorithm. Referring to Equation (11), the beam-space snapshot vector with the GP
operation can be further expressed as follows:

z̃l (n) = BlH pl (n) = Ψl e

sl (n) + BlH e(n) (14)
′
where Bl = (G a GlH ) H ; Ψl = BlH A( f l , Θ) ∈ C M × J is the dictionary; e
sl (n) is the amplitude
of the plane wave in the candidate direction after filtering, and its sparsity K ′ (K ′ ≤ K ) is
the number of residual signals after filtering.
In this study, the MOMP algorithm is an extended form of the OMP algorithm under
the MMV model to solve for the column-uniform sparse matrix e sl . The algorithm is
as follows:

Algorithm 1 Multiple-Measurement-Vector Orthogonal Matching Pursuit (MOMP)

ine Step 1 Input: z̃l , Ψl , K ′ .
ine Step 2 Initialization: residual R0 = z̃l , subset Λ0 = ∅, and i = 1.
ine Step 3 At the i-th iteration step:
(a) Choose atom ’i satisfying ’i = arg max’i ΨlT Ri−1 p , and p ≥ 1;
(b) Update Λi = [Λi−1 , ’i ], and b sil = (ΛiH Λi )−1 ΛiH z̃l ;
(c) Update Ri = z̃l − Ψl b i
sl ;
(d) If i < K′, return to step 3; if i ≥ K′, terminate the iteration and switch to step 4.
ine Step 4 Output: The solution e sl = b sil and residual Rz̃ = Riz̃ .
ine

After solving for the matrix e

sl at every frequency, the final estimation of the spatial
spectrum in the beam space can be calculated as follows:

L N
1
s=
N ∑ ∑ esl (n) (15)
l =1 n =1

2.3. Design of Dimension-Reduced Matrix Filter with Deep Nulling

This section describes the approach to achieving high localization accuracy of the
presented estimator in an environment with strong interfering sources. In particular,
a dimension-reduced matrix filter with deep nulling is presented to improve interference
suppression and robustness compared to a dimension-reduced matrix filter with nulling.

2.3.1. Design of Dimension-Reduced Matrix Filter with Nulling

The design of DR-MFNs based on quiescent matrix filters (QMFs) was proposed in [9].
This method uses a matrix filter, designed by the discrete prolate spheroidal sequence-
based approach [34,35] as a quiescent matrix filter, to design the DR-MFN by constraining
the difference between the actual response and QMF and the response of the stopband
while minimizing the output power of the filtered array. The design of the MFN can be
formulated as follows:

min tr GlH Rl Gl

Gl
(
GlH − Glq H ≤ε (16)
s.t. F
Gl a( f l , θs ) 2 ≤ δ, θs ∈ ΘS , s = 1, · · · , S
H

where, ΘS combines a continuum of all out-of-sector directions; θs ∈ ΘS (s = 1, · · · , S) is

the angular grid chosen (uniform or nonuniform) that properly approximates the stopband
Remote Sens. 2025, 17, 477 6 of 24

region by a finite number S of directions; ∥•∥ F is the matrix Frobenius norm; ∥•∥2 is
the vector 2-norm; ε > 0 is the parameter that bounds the passband distortion of the
designed matrix filter Gl with respect to Glq ; δ > 0 is a parameter of the user’s choice that
characterizes the worst acceptable SA; and Glq is the QMF. The beam space of QMFs varies
across different beam-space dimensions, as shown in Figure 1.

0
Amplitude Response(dB)
-20

-40

M'=2 M'=8 M'=14

-60 M'=4 M'=10 M'=16
M'=6 M'=12 M'=18

-90 -60 -30 0 30 60 90

Bearing(º)

Figure 1. Beam-space amplitude response of the QMFs across various dimensions. The two orange
dashed lines represent the left and right limitations of the sector-of-interest area.

Figure 1 shows the beam-space amplitude response of the QMFs across various di-
mensions for a 32-element uniform line array with a spacing of 1m when the passband
is set to [33◦ , 55◦ ] with a frequency of 600 Hz. Note that the QMFs have a high passband
response error within the preset passband for M′ = 2. This is because of the low dimen-
sionality, which may lead to an insufficient number of eigenvalues to adequately represent
the passband space. Therefore, selecting such a low number of dimensions is not advisable
when designing the DR-MFN. Moreover, the dimensions should satisfy

M p ≤ M′ ≤ M (17)

where M p is the minimum number of eigenvalues required to represent passband space.

Equation (16) indicates that DR-MFN is derived by imposing additional constraints
based on QMF. Therefore, the beam-space characteristics of the DR-MFN are related to
QMF. Applying energy constraints in noninterfering directions may lead to undesired
distortions within the passband and insufficient suppression of strongly interfering sources.
In particular, obtaining the desired matrix filter when QMF does not satisfy the desired
SA is challenging because the stopband and energy constraints are complicated, and the
parameters are difficult to establish. Therefore, the robustness of this technique for out-of-
sector sources may be insufficient, making it impractical.

2.3.2. Design of Dimension-Reduced Matrix Filter with Deep Nulling

To enhance the robustness of the adaptive matrix filter with nulling, the DR-MFDN
was proposed and used for the DOA estimation of the signals of interest. Assuming that all
out-of-sector sources satisfy INR ≥ 15 dB, the design procedure is as follows:
Step1 Input: f l , Θ P , ΘS , SA, Rl , ϵ.
Step2 Initialization: δ = 10(SA/20) .
Step3 Preprocess with MVDR algorithm [36]:
(a) Search spectral peaks in stopband and record their normalized amplitudes
[ I1 , · · · , IK′′ ];
(b) Record the directions [θ1 , · · · , θK′′ ] corresponding to the peaks.
Step4 Generate the projection matrix:
Remote Sens. 2025, 17, 477 7 of 24

(a) The sector of the main lobe where each peak is located is identified; the k′′ th
sector of the main lobe is denoted as Θk′′ ,

Θk′′ = [θk′′ −lk′′ , · · · , θk′′ , · · · , θk′′ +rk′′ ], k′′ = 1, · · · , K ′′ . (18)

where θk′′ −lk′′ and θk′′ +rk′′ are the left and right boundary angulars of the main lobe
corresponding to the k′′ th peak, respectively.
(b) The projection matrix is constructed as follows: Ψl = [A( f l , Θ1 ), · · · , A( f l , ΘK′′ )]
at frequency f l .
Step5 Formulate the optimization problem:
(a) Conventional form:
n o
min tr Gl H ΨlH Rl Ψl Gl
G l
 Gl H − G H ≤ ε (19)
lc F
s.t.
 Gl H a( f l , θ ) ≤ δ
2

Step6 Transform Equation (19) into SOCP form:

n o n o
tr GlH ΨlH Rl Ψl Gl = tr GlH ΨlH R1/2 1/2
l Rl Ψl Gl
  2
= GlH ΨlH R1/2
l F
n  o 2 (20)
= vec I M′ GlH ΨlH R1/2l 2
 T  2
= ΨlH R1/2
l ⊗ I M′ gl
2

GlH − Glc
H
= ∥gl − glc ∥2 (21)
F
h i
GlH a( f l , θs ) = a T ( f l , θs ) ⊗ I M′ gl (22)
2 2

where vec{·} denotes the vectorization operator stacking the columns of a ma-
trix on top of each other, gl = vec(GlH ), glc = vec(Glc H ), ⊗ denotes the Kro-

necker matrix product and I M′ is an identity matrix of dimension M′ . Substituting

Equations (20)–(22) into Equation (19), Equation (19) can be rewritten as
 T 
min ΨlH Ul ⊗ I M′ gl
gl
2
( (23)
∥gl − glc ∥2 ≤ ε
s.t.
a ( f l , θs ) ⊗ I M′ gl 2 ≤ δ, θs ∈ ΘS , s = 1, · · · , S.
 T 

Step7 Solve the SOCP Problem: Use an SOCP solver (e.g., mosek) to solve for gl .
Step8 Output: Reshape gl back to matrix form Gl = [mat(gl )] H .
Through the above steps, a DR-MFDN Gl at frequency f l can be obtained. The en-
hancement principle of the projection matrix is to project all array data onto the subspace
corresponding to the main lobe.
Remote Sens. 2025, 17, 477 8 of 24

An unknown source with direction θu is projected as

x = Iu A( f l , Θk′′ )A H ( f l , Θk′′ )a( f l , θu )

= Iu A( f l , Θk′′ )[a H ( f l , θk′′ −lk′′ ); · · · ; a H ( f l , θk′′ +rk′′ )]a( f l , θu )
(24)
= l ′′ +Iru ′′ +1 [a( f l , θk′′ −lk′′ ), · · · , a( f l , θk′′ +rk′′ )][∆lk′′ , · · · , 1, · · · , ∆rk′′ ] T
k k
= Iu ∆u a(θu )

π f l Md π f l Md
sin[ c (sin θu −sin θk′′ −l )] sin[ c (sin θu −sin θk′′ +r )]
k′′ k′′
where ∆lk′′ = π fl d ; ∆rk′′ = π fl d ; and ∆u < 1 is
M sin[ c (sin θu −sin θk′′ −l )] M sin[ c (sin θu −sin θk′′ +r )]
k ′′ k′′
the amplitude of the transformation of the unknown source into an interfering source
after projection.
If θu = θk′′ , the spectrum estimation obtained by the MVDR algorithm provides a
narrower main lobe; that is, ∆lk′′ ≈ ∆rk′′ ≈ 1, and Equation (24) can be rewritten as follows:

Ik′′
x≈ [ f , a(θk′′ −lk′′ ), · · · , a( f l , θk′′ +rk′′ )][1, · · · , 1] T ≈ Ik′′ a( f l , θk′′ ) (25)
lk′′ + rk′′ + 1 l

From Equations (24) and (25), it can be seen that the projection matrix retains the
out-of-sector source components while concentrating the signal-of-interest components
from other directions onto these out-of-sector sources, thereby directing energy toward the
strong out-of-sector sources. Additionally, the interfering sources from different directions
reinforce each other.
By enhancing the out-of-sector sources through the projection matrix, the DR-MFDN
algorithm, which is designed based on the minimum energy criterion, effectively reduces
the impact of perturbations and noise within the passband, achieving deeper nulls in the
direction of strong interfering sources. As a result, the robustness of the adaptive matrix
filter with nulling is significantly improved.

3. Discussion
In the simulation, a uniform linear array of 32 isotropic array elements with a spacing
of 1 m was considered to analyze the effectiveness of the GP operation and DOA estimation
performance of the robust estimator based on the MOMP algorithm and DR-MFDN. Three
distinct received-signal scenarios were simulated.

3.1. The Influence of the GP Operation

The first scenario assumed that one weak signal of interest with DOA θ1 = 45◦
was present. Moreover, a strong out-of-sector source (interfering) located at θ2 = 0◦
(90◦ was defined as the end fire direction within the frequency range of 400 to 700 Hz.
The signal-to-noise ratio (SNR) was 10 dB, whereas the interference-to-noise ratio
(INR) of the interference was 40 dB. The data were partitioned into 10 segments,
and a 512-point FFT was performed for each segment. The sector-of-interest area is
◦ ◦ ◦ ◦ ◦ ◦
Θ p = [35 , 55 ], the out-of-sector area is Θs = [−90 , 28.5 ] ∪ [64.5 , 90 ], the beam-space
dimension is M′ = 10, and ε = 0.2, SA = −20 dB. In addition, the amplitude responses
and errors in the beam space were calculated as 20 lg GlH al ( f l , θ ) 2 ∥al ( f l , θ )∥2 and
 
 H 
20 lg Gl al ( f l , θ ) 2 − ∥al ( f l , θ )∥2 ∥al ( f l , θ )∥2 , respectively.
The purpose of this simulation scenario is to verify the necessity of the GP operation
and to analyze the extent to which both the GP operation and the DR-MFDN affect the
DOA estimation.
In the case of DR-MFN, its covariance matrix with no pre-whitening (NP) operation is
shown in Figure 2a. GlH Gl is not a unit matrix, which may result in the transformation of
white noise into colored noise. In Figures 2b and 3, when the OP operation is considered,
Remote Sens. 2025, 17, 477 9 of 24

(GlH Gl )−1/2 GlH Gl (GlH Gl )−1/2 is the unit matrix, which prevents transforming the noise
into colored noise. Meanwhile, the overall attenuation of the out-of-sector area is reduced.
Moreover, changes in the beam space may lead to insufficient noise and interference
suppression, which may affect the final processing results. Figures 2c,d and 3 consider
the case when the GP operation is performed. Note that BlH Bl is a unit matrix, obtained
using numerous Monte Carlo trials, which prevents transforming the noise into colored
noise without significant changes in the beam space. Moreover, a Monte Carlo trial was
necessary because of the randomness of the Gaussian matrix and the absence of a Gaussian
matrix with uncorrelated columns. A large number of Monte Carlo trials is undoubtedly
time-consuming. Therefore, determining an appropriate number of trials is necessary. This
point will be discussed in the third scenario.

1 1

2 2

Normalized Amplitude

Normalized Amplitude
0.8 0.8

4 4
0.6 0.6

6 6
0.4 0.4

8 8
0.2 0.2

10 10
0 0
2 4 6 8 10 2 4 6 8 10

(a) (b)

1 1

2 2
Normalized Amplitude

Normalized Amplitude
0.8 0.8

4 4
0.6 0.6

6 6
0.4 0.4

8 8
0.2 0.2

10 10
0 0
2 4 6 8 10 2 4 6 8 10

(c) (d)

Figure 2. Covariance matrix distributions of array noise. (a–d) represent the cases of non-pre-
whitening (NP) operation, the OP operation, and the GP operation with 100 and 1000 Monte Carlo
trials, respectively.
Amplitude Response Error(dB)

0 0
Amplitude Response(dB)

-20
-20
-40
-40
-60
-60 NP(MFN) NP(MFDN) NP(MFN) NP(MFDN)
OP(MFN) OP(MFDN) -80 OP(MFN) OP(MFDN)
GP(MFN) GP(MFDN) GP(MFN) GP(MFDN)
-80
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(°) Bearing(°)

(a) (b)

Figure 3. Beam-space characteristics. (a,b) represent the amplitude response and amplitude response
error of DR-MFN with different pre-whitening operations at 600 Hz, respectively. The GP operation
conducted 100 Monte Carlo trials. The two orange dashed lines represent the left and right limitations
of the sector-of-interest area.
Remote Sens. 2025, 17, 477 10 of 24

To clarify the difference in DOA estimation performance under different pre-whitening

operations and matrix filters, Figure 4 shows the estimation results of the DR-MFN-MUSIC
and DR-MFDN-MUSIC algorithms for different pre-whitening operations. The OP oper-
ation does not effectively reduce the effect of the strong interfering source, causing the
energy of the signal of interest to remain lower than that of the interfering source after fil-
tering. In contrast, the case without the pre-whitening operation shows better performance,
indicating that the effect is limited at the current noise levels, even with colored noise.
However, note that this trend may change under low SNR conditions. Furthermore, the GP
operation significantly improves the performance of DOA estimation with matrix filters.

0
Normalized Amplitude(dB)

-20 DR-MFN-MUSIC(NP)
DR-MFN-MUSIC(OP)
DR-MFN-MUSIC(GP)
-40 DR-MFDN-MUSIC(NP)
DR-MFDN-MUSIC(OP)
DR-MFDN-MUSIC(GP)
-60 MUSIC

-80

-100
-90 -60 -30 0 30 60 90
Bearing(°)

Figure 4. Estimation results of MUSIC without filtering and the DR-MFN-MUSIC and the DR-MFDN-
MUSIC algorithms for different pre-whitening operations. The green dashed line represents the DOA
of a weak signal of interest.

To analyze the impact of noise on the algorithms, Figure 5 shows the root-mean-square
errors (RMSEs) of the DR-MFN-MUSIC and the DR-MFDN-MUSIC algorithms with three
pre-whitening operations versus the INR and SNR. Specifically, the INR varies with a fixed
SNR of 10 dB, and the SNR varies with a fixed INR of 20 dB. The RMSE curves were
averaged over the signal sources and 500 independent simulation runs.

6 6
DR-MFN-MUSIC(NP) DR-MFN-MUSIC(NP)
5 DR-MFN-MUSIC(OP) 5 DR-MFN-MUSIC(OP)
DR-MFN-MUSIC(GP) DR-MFN-MUSIC(GP)
DR-MFDN-MUSIC(NP) DR-MFDN-MUSIC(NP)
4 4
RMSE(°)

RMSE(°)

DR-MFDN-MUSIC(OP) DR-MFDN-MUSIC(OP)
DR-MFDN-MUSIC(GP) DR-MFDN-MUSIC(GP)
3 3

2 2

1 1

0 0
15 20 25 30 35 40 -20 -10 0 10 20
INR(dB) SNR(dB)

(a) (b)

Figure 5. DOA estimation of RMSEs of the DR-MFN-MUSIC algorithm versus (a) INR and (b) SNR
with different pre-whitening operations.

In Figure 5a, the following observations are made in the context of using DR-MFN:
(1) RMSE increases with INR across all three pre-whitening methods. (2) The GP operation
yields the lowest RMSE, while the OP operation yields the highest. This difference is likely
due to the OP operation altering the beamspace characteristics (as shown in Figure 3),
which weakens DR-MFN’s interference suppression. (3) RMSE without pre-whitening is
generally lower than with OP. This may be because, in high-SNR conditions, changes in
beamspace characteristics have a greater impact on DOA estimation than colored noise does.
In the context of using DR-MFDN with sufficient suppression, noise effects in high-SNR
Remote Sens. 2025, 17, 477 11 of 24

conditions become negligible, leading to no significant performance differences among the

three methods.
In Figure 5b, further analysis shows the following results. (1) For MFN, the OP
operation outperforms the NP operation in the SNR range of [−20 dB, −14 dB], which
is likely due to the dominant impact of colored noise in low-SNR environments. (2) For
DR-MFDN, OP outperforms NP in the range of [−20 dB, −18.5 dB], and with identical
pre-whitening, DR-MFDN performs better than DR-MFN. This may be attributed to DR-
MFDN’s lower passband distortion, which makes it less susceptible to noise compared to
DR-MFN. (3) Regardless of the filtering method, GP consistently has the smallest impact
on DOA estimation, and as SNR increases, the performance differences between the three
methods decrease.
In conclusion, based on Figures 4 and 5, DR-MFDN’s strong interference suppression
and low passband distortion significantly enhance DOA estimation performance for weak
signals, while the benefits of the GP operation are relatively limited.

3.2. The Performance of the DR-MFDN-MOMP with the GP Operation

The second scenario assumed that two weak signals of interest with the DOAs θ3 = 44◦
and θ4 = 46◦ were present. Moreover, two strong out-of-sector sources (interfering) were
located at θ5 = 0◦ and θ6 = 15◦ within a frequency range of 600 to 700 Hz. The SNR of
both signals was 0 dB, whereas the INR of all the interfering sources was 40 dB. All the
other parameters of the tested scenario were the same as those of the previous scenario.
In addition, a GP operation was applied to this scenario in all the cases for simplicity.
In Figure 6, the DR-MFDN provides deeper nulls in the direction of the interfering
sources and a lower passband response error in the sector of interest than the DR-MFN.
Figure 7a plots estimation results of the MUSIC and MOMP without filtering, and the direc-
tions of the weak signals-of-interest cannot be distinguished. Subsequently, robust estima-
tors based on the MUSIC or MOMP algorithm without filtering and DR-MFN or DR-MFDN,
namely, DR-MFN-MUSIC, DR-MFDN-MUSIC, DR-MFN-MOMP, and DR-MFDN-MOMP
estimators, with GP operations, were applied for DOA estimation, as shown in Figure 7b.
The DR-MFN-MUSIC and DR-MFN-MOMP estimators with the GP operation exhibit poor
estimation performance for weak targets, which can be due to the insufficient suppression
of interfering sources by the DR-MFN. Comparatively, both algorithms coupled with the
DR-MFDN demonstrated effective estimation of weak targets, and the DR-MFDN-MOMP
estimator with the GP operation demonstrated a high resolution.

0 0
Amplitude Response(dB)

Amplitude Response(dB)

-10
-20

-20

-40
-30
DR-MFN DR-MFN
DR-MFDN DR-MFDN
-60 -40
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(º) Bearing(º)

(a) (b)

Figure 6. Beam-space characteristics. (a,b) represent amplitude responses and amplitude response
errors of the DR-MFN and DR-MFDN at 600 Hz. The two orange dashed lines represent the left and
right limitations of the sector-of-interest area.

To further investigate the performance under varying conditions, the RMSEs of the
DR-MFN-MUSIC(GP), DR-MFDN-MUSIC(GP), DR-MFN-MOMP(GP), and DR-MFDN-
Remote Sens. 2025, 17, 477 12 of 24

MOMP(GP) estimator are plotted with respect to INR and SNR in Figure 8. Specifically,
the INR was varied for a fixed SNR of 0 dB, and the SNR was varied for a fixed INR
of 25 dB. The RMSE curves were averaged over the signal sources and 500 independent
simulation runs.
0 0
MUSIC DR-MFN-MUSIC(GP)
MOMP DR-MFDN-MUSIC(GP)

Normalized Amplitude(dB)

Normalized Amplitude(dB)
DR-MFN-MOMP(GP)
-20 -10 DR-MFDN-MOMP(GP)

-40 -20

-60 -30

-80 -40

-30 0 30 60 -30 0 30 60
Bearing(°) Bearing(°)

(a) (b)

Figure 7. Estimation results of (a) the MUSIC and MOMP; (b) the DR-MFN-MUSIC(GP), DR-MFDN-
MUSIC(GP), DR-MFN-MOMP(GP), and DR-MFDN-MOMP(GP) estimators. The two green dashed
lines represent the DOAs of two weak signals of interest.

5
6 DR-MFN-MUSIC(GP) DR-MFN-MUSIC(GP)
DR-MFDN-MUSIC(GP) DR-MFDN-MUSIC(GP)
5 DR-MFN-MOMP(GP) 4 DR-MFN-MOMP(GP)
DR-MFDN-MOMP(GP) DR-MFDN-MOMP(GP)
RMSE(°)

RMSE(°)
4 3
3
2
2
1
1

0 0
15 20 25 30 35 40 -20 -15 -10 -5 0 5 10 15
INR(dB) SNR(dB)

(a) (b)

Figure 8. DOA estimation RMSEs with respect to (a) INR and (b) SNR.

Figure 9 shows the probabilities of the source resolution with respect to INR and SNR
for the same scenario and methods tested. Similar to Figure 8, all curves were averaged
over the signal sources and 500 simulation runs. The signal sources are considered to be
resolved in the jth run if [17]

2
∑ |θbi − θi | < |θ1 − θ2 |
j
(26)
i =1
j
where θbi is the DOA estimation for the i-th source in the j-th run.
Probability of Source Resolution

Probability of Source Resolution

1 1

0.8 0.8

0.6 0.6

0.4 0.4
DR-MFN-MUSIC(GP) DR-MFN-MUSIC(GP)
DR-MFDN-MUSIC(GP) DR-MFDN-MUSIC(GP)
0.2 DR-MFDN-MOMP(GP)
0.2 DR-MFN-MOMP(GP)
DR-MFDN-MOMP(GP) DR-MFDN-MOMP(GP)
0 0
15 20 25 30 35 40 -20 -10 0 10 20
INR(dB) SNR(dB)

(a) (b)

Figure 9. Probabilities of source resolution with respect to (a) INR and (b) SNR.
Remote Sens. 2025, 17, 477 13 of 24

Figures 8a and 9a show that the RMSEs of the DR-MFN-MUSIC(GP) and DR-MFN-
MOMP(GP) algorithms increase with increasing INR because of the insufficient suppression
of strong out-of-sector sources by the DR-MFN. Conversely, the DR-MFDN provides
sufficient suppression, leading to superior DOA estimation capabilities for both the DR-
MFN-MUSIC(GP) and DR-MFN-MOMP(GP) estimators. Figures 8b and 9b show that both
the DR-MFN and DR-MFDN can sufficiently suppress strong out-of-sector sources for a
fixed INR of 25 dB. However, the DR-MFDN improved the performance with the same
DOA estimation algorithm, indicating superior signal preservation with a lower passband
response error. Figures 8 and 9 show that the MOMP algorithm outperforms the MUSIC
algorithm when utilizing the same matrix filter.

3.3. The Efficiency of the Presented Algorithm

In the third simulation scenario, we explore the impact of the number of GP operations
on DOA estimation accuracy and algorithm runtime. Given that the GP operation process-
ing is primarily employed to mitigate the nonlinear effects induced by colored noise, we set
the SNR to −20 dB, building on the conditions established in Scenario 1. The analysis from
the first two scenarios has already demonstrated the necessity of GP processing and the
superior performance of the DR-MFDN-MOMP(GP) algorithm for weak-signal DOA esti-
mation. However, the effect of GP on the runtime of different algorithms remains unclear,
which is crucial for practical applications. First, we conduct a comparative study of DOA
estimation performance for both the DR-MFDN-MOMP(GP) and DR-MFDN-MUSIC(GP)
algorithms under varying levels of GP, as shown in Figure 10. Additionally, Figure 10 also
illustrates the relationship between the degree of the beam-space response distortion and
the number of Monte Carlo trials, where the tolerance is calculated as follows:

J
Tolerance = ∑ ||GlH a( f l , θ j )−BlH a( f l , θ j )||2 (θ j ∈ Θ) (27)
j =1

0.2 5.5
Tolerance
DR-MFDN-MOMP(GP)
DR-MFDN-MUSIC(GP)
0.15
5
Tolerance

RMSE(°)

0.1

4.5
0.05

0 4
100 200 400 600 800 1000
Times

Figure 10. Effect on the beam-space preservation and the DOA estimation RMSEs of the DR-MFDN-
MUSIC(GP) and the DR-MFDN-MOMP(GP) algorithms with different numbers of GP operations.

Figure 10 illustrates that as the number of the GP operations increases, the tolerance
decreases. The rate of reduction progressively slows down, with convergence not being
achieved until after 1000 operations. While an increased number of GP operations may yield
a smaller tolerance, it also introduces a substantial computational burden. Upon evaluating
the DOA estimation performance of both algorithms across varying GP operation counts,
it is evident that the RMSEs for both algorithms reach convergence after approximately
100 GP operations. This phenomenon occurs because, although further increasing the GP
operation count reduces the discrepancy between the pre-whitened matrix filter and the
original matrix filter, the original matrix filter inherently exhibits a degree of passband
Remote Sens. 2025, 17, 477 14 of 24

distortion. After approximately 100 GP operations, the passband distortion stabilizes

and ceases to exhibit significant changes. Consequently, selecting 100 GP operations
effectively maximizes the performance gains from the GP processing while minimizing
computational overhead.
Figure 11 shows that the RMSE decreases as the number of snapshots increases for
both algorithms. This indicates that the accuracy of DOA estimation improves with more
data points, leading to a reduction in estimation error. Throughout the range of snapshot
numbers tested, the DR-MFDN-MOMP(GP) algorithm consistently exhibits lower RMSE
values compared to the DR-MFDN-MUSIC(GP) algorithm. This suggests that DR-MFDN-
MOMP(GP) provides more accurate DOA estimations under the same conditions. The rate
at which RMSE decreases is relatively steep when the number of snapshots is low (e.g.,
between 0 and 10). As the number of snapshots continues to increase, the rate of decrease
in RMSE begins to slow down, particularly noticeable after surpassing 20 snapshots. This
suggests diminishing returns on increasing the number of snapshots beyond a certain point.

6.5
DR-MFDN-MOMP(GP)
6 DR-MFDN-MUSIC(GP)

5.5
RMSE(°)

4.5

3.5
0 5 10 15 20 25 30
Snapshot Number
Figure 11. The DOA estimation RMSEs of the DR-MFDN-MUSIC(GP) and the DR-MFDN-MOMP(GP)
algorithms with A different number of snapshots.

Based on the above conclusion, we computed the runtime of the two algorithms
under different numbers of the GP operations, as shown in Table 1. These simulations
were performed on a workstation equipped with an Intel Core i9-13900HX processor
(24 cores), 16 GB of DDR5 memory running at 5600 MHz, and an NVIDIA GeForce RTX
4060 Laptop GPU with 8 GB of dedicated memory. The primary storage device was a 1 TB
NVMe SSD (WD PC SN810), and the system was running on a Windows operating system.
The algorithms were implemented in Matlab R2022b.

Table 1. Runtime (ms) of the DR-MFDN-MUSIC(GP) and DR-MFDN-MOMP(GP) algorithms with

different numbers of GP operations.

Times
1 5 10 20 50 100
Method

MUSIC 55.20 \ \ \ \ \
DR-MFDN-MUSIC(GP) 8.40 39.80 79.40 158.70 392.10 779.5
MOMP 0.54 \ \ \ \ \
DR-MFDN-MOMP(GP) 0.50 2.50 4.60 9 22.8 44.60

From Table 1, we can conclude the following. (1) The runtime of MUSIC without di-
mensionality reduction is significantly higher than that of DR-MFDN-MUSIC(GP), resulting
in greater computational complexity. (2) DR-MFDN-MOMP(GP) with 100 GP operations
Remote Sens. 2025, 17, 477 15 of 24

runs in 44.60 ms, much more efficiently than DR-MFDN-MUSIC(GP) at 779.5 ms. While the
runtime of DR-MFDN-MOMP(GP) increases with more GP operations, its computational
complexity remains low, and the increase is gradual. (3) The runtime of DR-MFDN-
MOMP(GP) after 100 GP operations is lower than the single execution time of MUSIC
(55.20 ms), meaning it does not introduce a significant computational burden and is suitable
for real-time applications.
Based on Figure 10 and Table 1, the DR-MFDN demonstrates a high cost-effectiveness
ratio by effectively reducing computational burden while maintaining accuracy through
dimensionality reduction. With 100 GP operations, the maximum computational accu-
racy benefits are achieved. Additionally, the runtime of the DR-MFDN-MOMP(GP) algo-
rithm is significantly lower than that of DR-MFDN-MUSIC(GP), highlighting its superior
performance in high GP operation counts. The lower computational time of DR-MFDN-
MOMP(GP) makes it well suited for practical, real-time applications. In summary, the DR-
MFDN-MOMP(GP) estimator can accurately and quickly estimate DOA for weak targets in
an environment with strong interfering sources.

4. Results
In September 2019, a horizontal array detection experiment was conducted in the sea
area of Laoshan Bay, Qingdao. During the experiment, the relative planar position between
the horizontal array and the surface target vessel is shown in Figure 12a. The horizontal array
was deployed in a region with a water depth of 18 m, with 32 hydrophones uniformly spaced
at 1 m. The deployment depth of the horizontal array was 17.8 m, and the relative positions
of the array elements are illustrated in Figure 12b. The sound speed remained approximately
constant at around 1530 m/s, as indicated by the sound speed profile shown in Figure 12c.
The seabed in the experimental area was relatively flat, characterized by fine sand.
The sampling frequency was set to 6 kHz, and the duration of the processed data was
2 min. A short-time Fourier transform (STFT) was applied to the received data of a specific
array element, resulting in the time-frequency spectrogram shown in Figure 13a. The STFT
used a Hamming window of 4096 samples, with a 12.5% overlap. The time-domain and
frequency-domain information of a particular segment are illustrated in Figure 13b and
Figure 13c, respectively. Figure 13b shows that the received signal appears disordered.
However, multiple line spectra can be observed in Figures 13c. In the 400–700 Hz frequency
band, background noise is relatively low compared to the spectral line intensity. To better
identify the DOAs of weak targets, this frequency band was selected for the broadband
DOA estimation of weak signals.

0°
strong interfering source
weak target
hydrophone

X:30.9m
Y:1.29m

x
X:0m
Y:0m

(a)

Figure 12. Cont.

Remote Sens. 2025, 17, 477 16 of 24

0 30

3 25

6 20
Depth(m)

x(m)
9 15

12 10

15
5

0
18
1510 1530 1550 -5 0 5
Sound Speed(m/s) y(m)
(b) (c)

Figure 12. (a) Schematic of the relative planar position between the horizontal array and the target
vessel located on the water surface. The arrows indicate the arrival directions of the target signals.
(b) Sound speed profile. (c) Relative positions of the elements of the horizontal array.

dB
2 25
20
1.5 15
Time(Min)

10
1 5
0
0.5 -5
-10
0 -15
0 0.25 0.5 0.75
Frequency(kHz)
(d)
600 400
400
300
Amplitude

200
Amplitude

0 200

-200
100
-400

-600 0
0 10 20 30 0.25 0.5 0.75
Time(s) Frequency(kHz)

(e) (f)

Figure 13. (a) Time-frequency spectrogram of the processed data. The information of the data segment
in (b) time domain and (c) frequency domain.
Remote Sens. 2025, 17, 477 17 of 24

The observed space, [−90◦ , 90◦ ], was uniformly discretized into 181 candidate direc-
tions. The Bearing-Time Records (BTRs) of the MVDR, MUSIC and MOMP, as shown in
Figure 14, reveal information about the interference source and the sector where the weak
targets are located. The two interfering sources were moved to the far field of the array.
One interference was in the direction of approximately −17◦ , and the other was around
0◦ . A few weak targets were located at −47◦ , 42◦ , and 54◦ , and the first target was clearly
tracked. The DR-MFN and DR-MFDN were designed with the passband bearing sector Θ p
at [35◦ , 65◦ ] and the stopband bearing sector Θs at [−90◦ , 25◦ ] ∪ [80◦ , 90◦ ]. The SAs were
−15 dB. The received data sampled were transformed into the frequency domain by using
a 512-point FFT. The sample covariance matrix was obtained by averaging 22 snapshots
with a 50% overlap. The duration of each segment was 1s, and the total analysis time
was 2 min. The DR-MFN-MUSIC, DR-MFDN-MUSIC, DR-MFN-MOMP, and DR-MFDN-
MOMP estimators with orthogonal or GP operations were applied to localize weak targets.
The results are presented in Figures 15 and 16.
dB dB
2 0 2 0

-5 -5
1.5 1.5
Time(Min)

-10 -10

Time/Min
1 -15
1 -15

-20
-20
0.5 0.5
-25
-25

0 -30
0 -30
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(°) Bearing(°)
(a) (b)
dB
2 0 0
Normalized Amplitude(dB)

MVDR
-5 MUSIC
MOMP
1.5
-10 -5
Time/Min

1 -15

-20 -10
0.5
-25

0 -30 -15
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(°) Bearing(°)

(c) (d)

Figure 14. BTRs of the (a) MVDR, (b) MUSIC, (c) MOMP. (d) Profile of the results of (a–c) at t = 57 s.
The two green dashed lines in (d) represent the directions of weak targets.

Figure 15 illustrates the DOA estimation results of the four algorithms using the OP
operation. Figure 15a,b,e show that the DR-MFN-MUSIC(OP) and DR-MFDN-MUSIC(OP)
estimators could not obtain the DOA of a weak target with a broad main lobe. Moreover,
many spurious peaks appeared outside these sectors. Figure 15c–e show that the DR-MFN-
MOMP(OP) and DR-MFDN-MOMP(OP) estimators could determine the DOAs of the two
weak signals. In contrast, occasional spurious peaks were present near the target. Thus,
these two algorithms using the OP operation are not robust. Overall, combined with matrix
filters, the MOMP algorithm demonstrated superior DOA estimation performance to the
MUSIC algorithm, as shown in Figure 15. Figure 16 illustrates the DOA estimation results
of the four algorithms with the GP operation. A comparison of Figures 15 and 16 shows that
the algorithms with the GP operation outperform those with the OP operation. Figures 15a
and 16b,e show that the DR-MFDN-MUSIC(GP) estimator has a smaller response than
Remote Sens. 2025, 17, 477 18 of 24

the DR-MFN-MUSIC(GP) estimator in all non-target sectors and reduces the width of the
main lobe. This indicates that the DR-MFDN shows improved interference suppression
performance and localization accuracy. In Figure 16c–e, the track obtained using the DR-
MFDN-MOMP(GP) algorithm is clearer than that obtained using the DR-MFN-MOMP(GP)
estimator. In Figure 16e, the DR-MFDN-MOMP(GP) estimator localizes two weak targets.
Owing to the high passband response error of the DR-MFN, the result of the DR-MFN-
MOMP(GP) estimator shows a slight but potentially misleading spurious peak. Therefore,
the DR-MFDN-MOMP(GP) estimator achieves superior performance in an environment
with strong interfering sources.

dB dB
2 0 2 0

1.5 1.5
Time(Min)

Time(Min)
-5 -5

1 1

-10 -10
0.5 0.5

0 -15 0 -15
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(°) Bearing(°)

(a) (b)

dB dB
2 0 2 0

1.5 1.5
Time(Min)

Time(Min)

-5 -5

1 1

-10 -10
0.5 0.5

0 -15 0 -15
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(°) Bearing(°)

(c) (d)

0
Normalized Amplitude(dB)

-5

-10 DR-MFN-MUSIC(OP)
DR-MFDN-MUSIC(OP)
DR-MFN-MOMP(OP)
DR-MFDN-MOMP(OP)
-15
-90 -60 -30 0 30 42 54 90
Bearing(°)

(e)

Figure 15. BTRs of the (a) DR-MFN-MUSIC(OP), (b) DR-MFDN-MUSIC(OP), (c) DR-MFN-
MOMP(OP), and (d) DR-MFDN-MOMP(OP) estimators. (e) Profile of the results of (a–d) at t = 57 s.
The two green dashed lines in (e) represent the directions of weak targets.
Remote Sens. 2025, 17, 477 19 of 24

dB dB
2 0 2 0

1.5 1.5

Time(Min)

Time(Min)
-5 -5

1 1

-10 -10
0.5 0.5

0 -15 0 -15
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(°) Bearing(°)

(a) (b)
dB dB
2 0 2 0

1.5 1.5
Time(Min)

Time(Min)
-5 -5

1 1

-10 -10
0.5 0.5

0 -15 0 -15
-90 -60 -30 0 30 60 90 -90 -60 -30 0 30 60 90
Bearing(°) Bearing(°)

(c) (d)
0
Normalized Amplitude(dB)

DR-MFN-MUSIC(GP)
DR-MFDN-MUSIC(GP)
DR-MFN-MOMP(GP)
-5 DR-MFDN-MOMP(GP)

-10

-15
-90 -60 -30 0 30 42 54 90
Bearing(°)

(e)

Figure 16. BTRs of the (a) DR-MFN-MUSIC(GP), (b) DR-MFDN-MUSIC(GP), (c) DR-MFN-
MOMP(GP), (d) DR-MFDN-MOMP(GP) estimators. (e) Profile of the results of (a–d) at t = 57 s.
The two green dashed lines in (e) represent the directions of weak targets.

5. Conclusions
This study presents a robust DOA estimator for weak targets based on a combina-
tion of Gaussian pre-whitening (GP), dimension-reduced matrix filter with deep nulling
(DR-MFDN), and multiple-measurement-vector orthogonal matching pursuit (MOMP).
The presented method demonstrates superior performance in localizing weak signals in
environments with strong interfering sources. The main conclusions can be summarized
as follows:
(1) The DR-MFDN effectively suppresses strong interfering sources by forming deep
nulls in their directions, thereby significantly improving interference suppression
and localization accuracy compared to conventional methods like DR-MFN. This
enhancement is crucial for weak target detection.
(2) The Gaussian pre-whitening operation prevents the transformation of white noise
into colored noise, preserving the beam-space characteristics and ensuring robustness
Remote Sens. 2025, 17, 477 20 of 24

in DOA estimation. This is particularly important for maintaining accuracy in high-

noise environments.
(3) The MOMP algorithm, combined with the DR-MFDN, provides higher resolution and
better performance in handling short snapshots compared to traditional algorithms
such as MUSIC. This makes the proposed estimator more suitable for real-time appli-
cations.
(4) Experimental results from both simulations and sea trials demonstrate that the pre-
sented DR-MFDN-MOMP(GP) estimator outperforms existing methods in terms
of interference suppression, localization accuracy, and computational efficiency.
The method is highly cost-effective and suitable for practical implementation.
Future work may focus on further optimizing computational efficiency and address-
ing limitations in scenarios with small angular separations between targets and interfer-
ence sources.

Author Contributions: Conceptualization, S.W. and H.W.; methodology, S.W.; software, H.W.;
validation, Z.B., H.W. and S.C.; formal analysis, Z.B.; investigation, H.W.; resources, H.W.; data
curation, H.W.; writing—original draft preparation, S.W.; writing—review and editing, W.G. and
H.W.; visualization, Z.B.; supervision, S.C.; project administration, S.C., P.S. and B.S. All authors have
read and agreed to the published version of the manuscript.

Funding: This study was supported by the National Natural Science Foundation of China under
grant nos. 52071309 and 52001296.

Data Availability Statement: The original contributions presented in the study are included in the
article; further inquiries can be directed to the corresponding author.

Acknowledgments: The authors thank Jie Pang for the insightful discussions and instructive guidance.

Conflicts of Interest: The authors declare no conflicts of interest.

Abbreviations
The following abbreviations are used in this manuscript:
BTRs Bearing-Time Records
CMF Conventional matrix filter
DOA Direction-of-arrival
DR-MFDN Dimension reduced matrix filter with deep nulling
DR-MFN Dimension-reduced matrix filter with nulling
GP Gaussian pre-whitening
INR Interference-to-noise ratio
MMV Multiple-measurement-vector
MOMP Multiple-measurement-vector orthogonal matching pursuit
MUSIC Multiple signal classification
MVDR Minimum variance distortionless response
NP None pre-whitening
OMP Orthogonal matching pursuit
OP Orthogonal pre-whitening
QMFs Quiescent matrix filters
RMSE Root-mean-square errors
SA Stopband attenuation
SNR Signal-to-noise ratio
SpSF Sparse spectrum fitting
STFT Short-time Fourier transform
Remote Sens. 2025, 17, 477 21 of 24

Appendix A
The application of the GP operation serves two purposes: preserving the character-
istics of beam space and preventing the transformation of white noise into colored noise.
The former is evident from Equation (12). Therefore, Appendix A focuses on proving the
latter, namely, the covariance matrix of array noise with the GP operation.
The matrix GlH can be rewritten as a combination of column vectors,
′
GlH = [g1 , g2 , · · · , g M ] ∈ C M × M (A1)
′
where gi ∈ C M ×1 , i = 1, · · · , M.
′
By defining Ω = ∑iM H M′ × M′ , there is
=1 g i g i ∈ C

H Ωg H Ωg H Ωg
 
g a1 a1 g a1 a2 ··· g a1 aM′
 H Ωg
g a2 H Ωg
g a2 ··· H Ωg
g a2

a1 a2 aM′
BlH Bl
 
= .. .. .. ..  (A2)
. . . .
 
 
g aM′ Ωg a1
H ··· ··· g aM′ Ωg aM′
H

As Ω is a symmetric matrix, Ω can be rewritten as

H
Ω = QΩ ΛΩ QΩ (A3)
′ ′
where QΩ ∈ C M × M is the eigenvector matrix and also an orthogonal matrix; ΛΩ is the
eigenvalue matrix and is given by
 
λ1 0 ··· 0
 .. 
 0 λ2 . 
ΛΩ = (A4)
 
.. .. .. 
. . .
 
 
0 ··· ··· λ M′

′ ′
Define G′a = G a QΩ = [g′a1 , g′a2 , · · · , g′aM′ ] T ∈ C M × M . Based on the fact that an
orthogonal matrix multiplied by a Gaussian matrix does not change its characteristics and
therefore g′ai (m) and g ai (m) are identically distributed, Equation (A2) can be rewritten as
M′ ′2 M′ ′ ′ M′ ′ ′
 
∑m =1 λm g a1 ( m ) ∑m =1 λm g a1 ( m ) g a2 ( m ) · · · ∑m=1 λm g a1 ( m ) g aM′ ( m )
 M′ M′ M′
 ∑m=1 λm g′a2 (m)g′a1 (m) ∑m ′2
=1 λm g a2 ( m ) · · · ∑m ′ ′
=1 λm g a2 ( m ) g aM ( m ) 

H
Bl Bl = 
 .. .. .. ..  (A5)
. . . .

 
M′ ′ ′ M′ ′2
∑m =1 λm g aM′ ( m ) g a1 ( m ) ··· ··· ∑m =1 λm g aM′ ( m )

The characteristic function of the product sum of two n-dimensional Gaussian vectors
with zero means is denoted as [33]
!n/2
1
Ψ X (ω ) = (A6)
1 − 2jωρσ1 σ2 + σ12 σ22 ω 2 (1 − ρ2 )

where X is the inner product of two n-dimensional Gaussian vectors; ω is the conjugate
of the Fourier transform of X; ρ is the correlation coefficient of the two Gaussian vectors;
a σ1 dn and σ2 are the variances obeyed by the two Gaussian vectors, respectively. On the
basis of the relationship between the eigenfunctions and the origin moments,

kΨ  
−k d X (ω ) k
j = nE X (A7)
dω k
ω =0
Remote Sens. 2025, 17, 477 22 of 24

where k is the order of moment.

After an algebraic simplification, there is

E[ X ] = ρσ1 σ2 (A8)

Based on the relationship between g′ai (m) and g ai (m), as well as Equation (A8), it can
be deduced that
′ ′
E[g aiH g ai ] = E[g ai
H
g ai ] = σl2 (A9)
′ M′
1 ′ ′
M σ2
H
E[g ai Ωg ai ] = ′
E[g aiH g ai ] ∑ λm = l ′ ∑ λm (A10)
M m =1
M m =1

When the vectors of the columns of the matrix G a satisfy ∥g ai ∥2 = 1(i = 1, · · · , M′ ),

M′
there is ∑ λm = M′ . Equation (A10) can be further written as
m =1

H
E[g ai Ωg ai ] = σl2 (A11)

By Equations (A5) and (A11),

BlH Bl = σl2 I M′ × M′ (A12)

The array covariance matrix obtained after linear transformation with matrix Bl on
the array data is given by

Rz̃ = BlH Rl Bl
= BlH (A( f l , Θ)Rls A H ( f l , Θ) + Rle )Bl (A13)
= BlH A( f l , Θ)Rls A H ( f l , Θ)Bl + σl2 I M′ × M′

Thus, Equation (13) is proved.

References
1. Vaccaro, R.; Harrison, B. Optimal matrix-filter design. IEEE Trans. Signal Process. 1996, 44, 705–709. https://doi.org/10.1109/78.4
89044.
2. Vaccaro, R.J.; Harrison, B.F. Matrix filters for short data records. In Proceedings of the 28th Conference on Information Sciences and
Systems; Princeton University: Princeton, NJ, USA, 1994.
3. Vaccaro, R.; Harrison, B. Matrix filters for passive sonar. In Proceedings of the 2001 IEEE International Conference on
Acoustics, Speech, and Signal Processing, Proceedings (Cat. No.01CH37221), Salt Lake City, UT, USA, 7–11 May 2001; Volume 5,
pp. 2921–2924. https://doi.org/10.1109/ICASSP.2001.940258.
4. Vaccaro, R.J.; Chhetri, A.; Harrison, B.F. Matrix filter design for passive sonar interference suppression. J. Acoust. Soc. Am. 2004,
115, 3010–3020. https://doi.org/10.1121/1.1736653.
5. MacInnes, C. Source localization using subspace estimation and spatial filtering. IEEE J. Ocean Eng. 2004, 29, 488–497.
https://doi.org/10.1109/JOE.2004.827290.
6. Yan, S.; Ma, Y. Matched field noise suppression: A generalized spatial filtering approach. Chin. Sci. Bull. 2004, 49, 2220–2223.
7. Yan, S.; Hou, C.; Ma, X. Matrix spatial prefiltering approach for direction-of-arrival estimation. Chin. J. Acoust. 2007, 32, 151–157.
8. Han, D.; Zhang, X.H. Optimal Matrix Filter Design With Application to Filtering Short Data Records. IEEE Signal Process. Lett.
2010, 17, 521–524. https://doi.org/10.1109/LSP.2010.2044850.
9. Hassanien, A.; Elkader, S.; Gershman, A.; Wong, K. Convex optimization based beam-space preprocessing with improved
robustness against out-of-sector sources. IEEE Trans. Signal Process. 2006, 54, 1587–1595. https://doi.org/10.1109/TSP.2006.870564.
10. Yang, Y.; Zhang, Y.; Yang, L. Wideband sparse spatial spectrum estimation using matrix filter with nulling in a strong interference
environment. J. Acoust. Soc. Am. 2018, 143, 3891–3898. https://doi.org/10.1121/1.5042406.
11. Feng, J.; Yang, Y.; Sun, C. Adaptive spatial matrix filter design with application to DOA estimation. J. Syst. Simul. 2007,
19, 4798–802.
12. Wang, X.; Amin, M.; Wang, X.; Cao, X. Sparse Array Quiescent Beamformer Design Combining Adaptive and Deterministic
Constraints. IEEE Trans. Antennas Propag. 2017, 65, 5808–5818. https://doi.org/10.1109/TAP.2017.2751672.
Remote Sens. 2025, 17, 477 23 of 24

13. Wang, X.; Aboutanios, E. Adaptive Reduced-Dimensional Beamspace Beamformer Design by Analogue Beam Selection. In
Proceedings of the ICASSP 2019—2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP),
Brighton, UK, 12–17 May 2019; pp. 4350–4354. https://doi.org/10.1109/ICASSP.2019.8683360.
14. Schmidt, R. Multiple emitter location and signal parameter estimation. IEEE Trans. Antennas Propag. 1986, 34, 276–280.
https://doi.org/10.1109/TAP.1986.1143830.
15. Yan, S. Optimal Array Signal Processing: Modal Array Processing and Direction-of-Arrival Estimation; Science Press: Beijing, China,
2018; pp. 214–223.
16. Zoltowski, M.; Kautz, G.; Silverstein, S. Beamspace Root-MUSIC. IEEE Trans. Signal Process. 1993, 41, 344–364. https:
//doi.org/10.1109/TSP.1993.193151.
17. Gershman, A. Direction finding using beamspace root estimator banks. IEEE Trans. Signal Process. 1998, 46, 3131–3135.
https://doi.org/10.1109/78.726831.
18. Karabulut, G.; Kurt, T.; Yongacoglu, A. Angle of arrival detection by matching pursuit algorithm. In Proceedings of the IEEE
60th Vehicular Technology Conference, 2004, VTC2004-Fall. 2004, Los Angeles, CA, USA, 26–29 September 2004; Volume 1,
pp. 324–328. https://doi.org/10.1109/VETECF.2004.1400011.
19. Karabulut, G.Z.; Kurt, T.; Yongaçoglu, A. Estimation of directions of arrival by matching pursuit (EDAMP). EURASIP J. Wirel.
Commun. Netw. 2005, 2005, 197–205. https://doi.org/10.1155/WCN.2005.197.
20. Cotter, S.F. Multiple snapshot matching pursuit for direction of arrival (DOA) estimation. In Proceedings of the 2007 15th
European Signal Processing Conference, Poznan, Poland, 3–7 September 2007; pp. 247–251.
21. Malioutov, D.; Cetin, M.; Willsky, A. A sparse signal reconstruction perspective for source localization with sensor arrays. IEEE
Trans. Signal Process. 2005, 53, 3010–3022. https://doi.org/10.1109/TSP.2005.850882.
22. Fuchs, J.J. On the application of the global matched filter to DOA estimation with uniform circular arrays. IEEE Trans. Signal
Process. 2001, 49, 702–709. https://doi.org/10.1109/78.912914.
23. Model, D.; Zibulevsky, M. Signal reconstruction in sensor arrays using sparse representations. Signal Process. 2006, 86, 624–638.
https://doi.org/10.1016/j.sigpro.2005.05.033.
24. Gorodnitsky, I.; Rao, B. Sparse signal reconstruction from limited data using FOCUSS: A re-weighted minimum norm algorithm.
IEEE Trans. Signal Process. 1997, 45, 600–616. https://doi.org/10.1109/78.558475.
25. Zdunek, R.; Cichocki, A. Improved M-FOCUSS Algorithm With Overlapping Blocks for Locally Smooth Sparse Signals. IEEE
Trans. Signal Process. 2008, 56, 4752–4761. https://doi.org/10.1109/TSP.2008.928160.
26. Zheng, J.; Kaveh, M. Sparse Spatial Spectral Estimation: A Covariance Fitting Algorithm, Performance and Regularization. IEEE
Trans. Signal Process. 2013, 61, 2767–2777. https://doi.org/10.1109/TSP.2013.2256903.
27. Yang, L.; Yang, Y.; Zhu, J. Source localization based on sparse spectral fitting and spatial filtering. In Proceedings of the OCEANS
2016 MTS/IEEE Monterey, Monterey, CA, USA, 19–23 September 2016; pp. 1–4. https://doi.org/10.1109/OCEANS.2016.7761271.
28. Chen, J.; Huo, X. Theoretical Results on Sparse Representations of Multiple-Measurement Vectors. IEEE Trans. Signal Process.
2006, 54, 4634–4643. https://doi.org/10.1109/TSP.2006.881263.
29. Tropp, J.A.; Gilbert, A.C. Signal Recovery From Random Measurements Via Orthogonal Matching Pursuit. IEEE Trans. Inf. Theory
2007, 53, 4655–4666. https://doi.org/10.1109/TIT.2007.909108.
30. Bian, Z.; Wang, H.; Gao, W.; Song, P. DOA Estimation of Broadband Sources Using Dimension-Reduced Matrix Filter with Deep
Nulling in a Strong Interference Environment. J. Phys. Conf. Ser. 2024, 2718, 012094. https://doi.org/10.1088/1742-6596/2718/1
/012094.
31. Trees, H.L.V. Adaptive Beamformers. In Optimum Array Processing; John Wiley & Sons, Ltd.: Hoboken, NJ, USA, 2002; Chapter 7,
pp. 710–916. https://doi.org/10.1002/0471221104.ch7.
32. Lee, H.; Wengrovitz, M. Resolution threshold of beamspace MUSIC for two closely spaced emitters. IEEE Trans. Acoust. Speech
Signal Process. 1990, 38, 1545–1559. https://doi.org/10.1109/29.60074.
33. Simon, M.K. Probability Distributions Involving Gaussian Random Variables: A Handbook for Engineers and Scientists; Springer: Boston,
MA, USA, 2002; pp. 49–60. https://doi.org/10.1007/978-0-387-47694-0_7.
34. Forster, P.; Vezzosi, G. Application of spheroidal sequences to array processing. In Proceedings of the ICASSP ’87 IEEE
International Conference on Acoustics, Speech, and Signal Processing, Dallas, TX, USA, 6–9 April 1987; Volume 12, pp. 2268–2271.
https://doi.org/10.1109/ICASSP.1987.1169421.
Remote Sens. 2025, 17, 477 24 of 24

35. Van Veen, B.; Williams, B. Structured covariance matrices and dimensionality reduction in array processing. In Proceedings
of the Fourth Annual ASSP Workshop on Spectrum Estimation and Modeling, Minneapolis, MN, USA, 3–5 August 1988;
pp. 168–171. https://doi.org/10.1109/SPECT.1988.206184.
36. Capon, J. High-resolution frequency-wavenumber spectrum analysis. Proc. IEEE 1969, 57, 1408–1418. https://doi.org/10.1109/
PROC.1969.7278.

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.