Robust doa estimation method based on virtual uniform linear array beamforming
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TSINGHUA UNIVERSITY
- Filing Date
- 2026-03-09
- Publication Date
- 2026-06-09
Smart Images
Figure CN122172109A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of array signal processing and relates to a robust multi-signal spatial spectrum direction finding method in scenarios with an unknown number of signal sources, and more particularly to a robust angle of arrival estimation method based on virtual uniform linear array beamforming. Background Technology
[0002] In modern wireless communication environments, the presence of illegal or unknown electromagnetic radiation sources poses a potential threat to critical systems such as communication, navigation, and radar. Especially in cities, border areas, and sensitive regions, the concealment of such signals and the complexity of the environment significantly increase the difficulty of monitoring and identification. Accurately estimating the direction of arrival of signals is a core prerequisite for effective spectrum management and the development of electromagnetic situational awareness capabilities.
[0003] Traditional methods for estimating the angle of arrival (AHA) of signals include amplitude-based direction finding, interferometric direction finding, and time-difference-of-arrival (TDOA) direction finding. The most representative amplitude-based method is the "amplitude comparison method," which compares the AHA direction measured by antennas with different beam directions. This method has low accuracy and heavily relies on the absence of antenna beam pattern distortion and the accuracy of amplitude measurement. Interferometric direction finding calculates the phase difference between different antennas to determine the corresponding AHA direction. A representative method is the "multi-baseline interferometer," which uses multiple antennas to form multiple pairs of baselines of different lengths to compromise between accuracy and ambiguity. This method offers high accuracy and is easy to implement in hardware, making it suitable for practical applications. TDOA direction finding calculates the corresponding AHA by measuring the time difference of signal arrival at different antennas. Theoretically, TDOA can achieve extremely high accuracy, but it heavily depends on the accuracy of the time difference measurement, often requiring nanosecond-level accuracy, which is difficult to achieve in practice. Meanwhile, all the aforementioned traditional methods assume a single-source, single-carrier signal model, making them unsuitable for scenarios with multiple sources operating simultaneously in the same frequency band, and their direction-finding accuracy is highly dependent on the signal modulation method. These methods perform well for pulse radar signals. However, for continuous wave communication signals, the performance of these methods degrades significantly due to factors such as symbol rate.
[0004] Array signal processing, as a scheme that utilizes multi-antenna arrays combined with modern digital signal processing techniques for spatial signal processing, has received widespread attention. Array signal processing methods primarily aimed at measuring the angle of arrival (DOA) of signals are collectively referred to as "spatial spectrum direction finding." Their greatest advantage lies in their ability to simultaneously handle multi-source scenarios and their robustness to narrowband continuous waves. In the 1990s, high-resolution algorithms such as MUSIC and ESPRIT were proposed. These methods and their variants are collectively called subspace methods, named for their use of the eigenspace decomposition of the array data covariance matrix. Subspace decomposition-based methods offer advantages in high resolution and high direction-finding accuracy. However, subspace methods heavily rely on accurate prior information about the number of sources and the array manifold, as well as the model assumption of Gaussian white noise. In real-world scenarios (especially in modern urban environments), there are often non-ideal factors such as unknown source numbers, array system errors, and correlated environmental noise, leading to a significant decrease in the performance or even failure of traditional high-resolution subspace DOA estimation algorithms.
[0005] Since the beginning of the 21st century, a class of direction-finding methods utilizing the sparsity of signal angles of arrival (Angles of Arrival) in the spatial domain has emerged, commonly referred to as sparse reconstruction methods. These methods model the Angle of Arrival estimation problem as a sparse signal recovery problem in the spatial domain, with SPICE and OGSBI being representative examples. These methods avoid reliance on prior information about the number of signal sources, maintaining high resolution and accuracy even in scenarios with unknown source numbers. However, in the absence of information about the number of sources or signal and noise power, sparse reconstruction methods struggle to select appropriate thresholds to distinguish between peak and spurious peaks corresponding to the Angle of Arrival. This limitation is particularly pronounced in scenarios with coexisting strong and weak signals and a large dynamic range of signal-to-noise ratio (SNR). This significantly restricts the practical deployment of sparse reconstruction methods.
[0006] The direction-finding stability and accuracy of linear arrays are highly dependent on the array configuration. Larger array apertures generally result in higher direction-finding accuracy. However, using a uniform linear array with element spacing of half a wavelength leads to a rapid increase in the number of elements with increasing aperture, and the dense element arrangement causes significant mutual coupling, reducing direction-finding stability. Using irregular, sparse, non-uniform arrays can result in excessively high grating lobes in some directions, causing false alarms and reducing direction-finding stability. Minimum redundancy arrays can efficiently construct virtual uniform linear arrays, achieving a balance between direction-finding performance and hardware resources. However, existing methods often only utilize virtual linear arrays for subspace DOA estimation or sparse reconstruction algorithms, leaving these algorithms with the same problems as those used in subspace and sparse reconstruction algorithms. This neglects the advantage of virtual uniform linear arrays in achieving spatial diversity through beamforming.
[0007] Against this backdrop, there is an urgent need to develop a robust DOA algorithm estimation framework that utilizes the spatial diversity capability of a virtual uniform linear array, enabling it to adapt to scenarios such as unknown number of information sources, model mismatch, and power imbalance, and to provide key technical support for passive sensing systems in complex electromagnetic environments. Summary of the Invention
[0008] In view of this, in order to solve the problem of the lack of robustness of array direction finding algorithms in complex scenarios in the existing technology, such as unknown number of signal sources, array systematic errors, environmental noise with spatial coherence, and large differences in signal power, this invention provides a robust direction of arrival (DOA) estimation method based on virtual uniform linear array beamforming. It can adapt to robust array direction finding algorithms with a large dynamic range of signal power, so as to solve the problem of unstable direction finding performance in scenarios with unknown number of signal sources, model mismatch, and large power differences.
[0009] To achieve the above objectives, the present invention provides the following technical solution: A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming includes the following steps: S1. Based on the Minimum Redundancy Array (MRA), the received data model is set up. The covariance matrix of the received data is processed by Khatri-Rao product and spatial smoothing algorithm to generate the data covariance matrix of the equivalent virtual uniform linear array. Then, multi-beam synthesis is performed to generate the virtual uniform linear array. S2. Multiple high-gain narrow beams are generated using a virtual uniform linear array. The main lobe of each beam is evenly spaced within the observation area, and the observation area is divided into several non-overlapping sectors. S3. Within each beam coverage sector, the temporal stability of the beam output power in multiple data block subframes is analyzed, and the presence of a signal source within the sector is determined based on the temporal stability. S4. Select the sector corresponding to the beam with the largest output power among the beams that have passed the stability test, and use a high-precision estimation method without preset parameters to estimate the angle of arrival of the corresponding signal component. S5. Perform time-frequency non-Gaussianity verification on the detected signal angle of arrival. If the signal component passes the non-Gaussianity test, add the corresponding angle of arrival to the list of detected signal angles of arrival; otherwise, mark it as noise. S6. Update the weights using null techniques, generating zeros at the arrival angles of detected signal components to suppress the influence of signal components on subsequent detection; repeat steps S2 to S6 until no new stable beams are detected.
[0010] Furthermore, the virtual uniform linear array synthesis process in step S1 is as follows: Let the number of array elements be... Minimum redundancy array settings for array receiving data , For the number of snapshots; Under the assumption of ideal Gaussian white noise, find the data covariance matrix of the minimum redundancy array MRA. ; The data covariance matrix of the minimum redundancy array Vectorization is performed to obtain the equivalent data reception vector. After deduplication of the equivalent data received vector, the corresponding effective received vector is obtained. ; from Extract in sequence A length of continuous subvectors Each subvector corresponds to a translated subarray. The covariance matrix of each subarray is then subjected to incoherent averaging to achieve the desired result from a subarray with a number of elements. Minimum Redundancy Array to Equivalent synthesis of virtual uniform linear arrays of array elements.
[0011] Furthermore, in step S1, under the narrowband far-field assumption, the received data... Represented as:
[0012] in It is an array manifold. For the first The steering vector of each source is represented as: ; The normalized position of each array element; The number of unknown sources , These are the signal and noise matrices, respectively. Under the assumption of ideal Gaussian white noise, the data covariance matrix of the minimum redundancy array (MRA) for:
[0013] in , representing the signal power vector Noise power; Equivalent data receiving vector Represented as:
[0014] in , Represents the Khatri-Rao product; matrix For a manifold corresponding to a virtual array, the element positions are the difference set of the original MRA position set { }, The number of array elements are respectively The elements in the normalized element positions of the minimum redundancy array; The incoherent averaging process of the covariance matrix of each subarray is expressed as follows: .
[0015] Furthermore, in step S2, the sector partitioning process is as follows: Let the observation area of the linear array element be... The coverage sector width of each beam According to the observation area and coverage sector width Determine the number of beams to be generated ; Set the main lobe of adjacent beams to point to beams spaced apart by a width of [value]. The main lobe of the beam points to When no signal is initially detected, calculate the first... Weight of each beam , and thus define the first Each beam covers a sector .
[0016] Furthermore, in step S2, the number of beams The calculation method is as follows:
[0017] No. The weights of each beam are calculated as follows:
[0018] in, The number of elements in the virtual uniform linear array; No. Each beam covers a sector This refers to the first [angle range]. Gain of each beam Larger than other beams, that is: .
[0019] Furthermore, in step S3, the process of determining whether a signal source exists based on the temporal stability of the multi-subframe beam output power is as follows: First, the array receives data according to its snapshot number. Average score The data block subframes are divided into several data block subframes, and the output power of each beam in each data block subframe is calculated. :
[0020] in, It is the virtual uniform linear array. The covariance matrix corresponding to the frame data; Then, a clustering algorithm is used to detect outliers in the beam output power within each subframe, identifying the high-power beam index set whose output power is significantly higher than the average level. If a certain beam index is in all All of them were identified as high-power anomalies in each subframe, that is... If the output power of the beam is stable in the time domain, then the corresponding sector is considered to have a source; otherwise, it is considered not to have a source.
[0021] Furthermore, in step S4, among all stable beam sectors where the information source is determined to exist, the sector with the strongest output power is selected. The Capon filter is used to estimate the angle of arrival of the signal. .
[0022] Furthermore, in step S5, the process of verifying the time-frequency non-Gaussianity of the detected signal arrival angle is as follows: First, at the angle of arrival of the detected signal component Signal extraction using Capon filter ; For the extracted signal Perform a short-time Fourier transform to obtain its time-frequency coefficients. The real parts of the time-frequency coefficients are downsampled to obtain the downsampled sequence. ; Finally, the sample kurtosis of the downsampled sequence is calculated. The difference between the sample kurtosis and the preset kurtosis is used to determine whether it is a real signal source or noise. If it exceeds the preset threshold, it is a real signal source; otherwise, it is noise.
[0023] Furthermore, in step S5, in the already detected direction Extract signal The method is as follows:
[0024] in, These are the Capon filter coefficients corresponding to the actual array. This is the data received by the actual array; Sample kurtosis of downsampled sequences Represented as:
[0025] Set the preset kurtosis to The preset threshold is ,like If the signal is true, its DOA is determined to be a genuine source and included in the final output; otherwise, it is marked as Gaussian noise and not displayed in the final result.
[0026] Furthermore, in step S6, the method for suppressing the influence of the detected signal components using the zero-dip technique is as follows: After each detected signal component's angle of arrival, the beamforming weights are updated, and nulls are formed at the detected angles of arrival.
[0027]
[0028] in, , ; For the first The main lobe of each beam points towards... The angle of arrival of the detected signal component is obtained; the th... The beam weights are .
[0029] The beneficial effects of this invention are as follows: This invention achieves spatial partition detection and guided estimation through virtual uniform linear array beamforming, effectively avoiding the reliance on prior information source numbers in traditional methods. It utilizes beam output power stability and time-frequency non-Gaussianity to verify and improve robustness under non-ideal conditions such as array errors, colored noise, and signal power imbalance. The array configuration based on a minimum redundancy array maintains a large effective aperture while reducing hardware costs. Simultaneously, it utilizes the spatial diversity of the virtual uniform linear array to achieve multi-signal isolation, improving multi-signal direction finding stability. The algorithm can automatically remain silent when there is no information source, requiring no manual intervention. This method is suitable for DOA estimation scenarios with high requirements for robustness and practicality, such as passive radar and wireless sensing. Specifically, it has the following significant characteristics: (1) Completely get rid of the dependence on prior information of the number of sources: Spatial partition detection and serialization estimation are realized by virtual uniform linear array beamforming, which transforms the complex source number estimation problem into multiple independent binary hypothesis testing problems, fundamentally avoiding the strong dependence of traditional subspace methods and sparse reconstruction methods on accurate prior information of the number of sources.
[0030] (2) Robustness under non-ideal conditions: Robust to array error - by utilizing the spatial diversity characteristics of beamforming, the sensitivity to array manifold accuracy is reduced; Robust to environmental noise - by passing the time-frequency non-Gaussianity test, the real signal is effectively distinguished from Gaussian noise; Robust to power imbalance - by adopting a serialization detection mechanism, the strongest signal is captured first and its influence is suppressed, ensuring that weak signals are not overwhelmed.
[0031] (3) It has the ability to operate automatically and resist false alarms: the method automatically shuts down when there is no information source through time-domain stability testing and statistical verification mechanism, effectively suppressing false alarms; it achieves reliable separation and tracking in multi-source scenarios without the need for manual intervention in threshold setting or result interpretation.
[0032] (4) Applicable to a variety of practical scenarios: This invention has good adaptability to various signal forms such as narrowband continuous wave and pulse signal, and can be widely used in fields with high practicality and robustness requirements such as passive radar, electromagnetic spectrum monitoring, and wireless communication sensing.
[0033] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description
[0034] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a schematic diagram of the overall process of the robust angle of arrival estimation method based on virtual uniform linear array beamforming according to an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the equivalent synthesis process from a 6-element MRA to a 14-element virtual uniform linear array in an embodiment of the present invention. Figure 3 This is a schematic diagram of the antenna radiation pattern and array configuration of the linear array elements in an embodiment of the present invention, and a synthesized multi-beam distribution; wherein, Figure 3 (a) shows the antenna radiation pattern. Figure 3 (b) shows the array structure and the distribution of the synthetic multibeams; Figure 4 This is an example diagram showing the tracking results of a dynamic information source according to an embodiment of the present invention, wherein, Figure 4 (a) shows the trajectory of the actual angle of arrival of the dynamic source. Figure 4 (b) shows the direction finding results of the method of the present invention. Detailed Implementation
[0035] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0036] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.
[0037] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.
[0038] Please see Figures 1-4 This is a robust angle-of-arrival estimation method based on virtual uniform linear array beamforming.
[0039] Example This embodiment details the implementation process of a robust angle-of-arrival estimation method based on virtual uniform linear array beamforming, such as... Figure 1 As shown, it includes at least the following steps: S1: Based on the Minimum Redundancy Array (MRA), the data covariance matrix of an equivalent virtual uniform linear array is generated by expanding the Khatri-Rao product and spatial smoothing algorithm, and multi-beam synthesis is performed simultaneously using this virtual linear array. S2: Multiple high-gain narrow beams are generated simultaneously using a virtual uniform linear array. The main lobe of each beam is evenly distributed in the observation area, achieving complete coverage of the entire observation area and dividing the observation area into several non-overlapping sectors. S3: Within each beam coverage sector, by analyzing the temporal stability of the beam output power in multiple data block subframes, determine whether there is a signal source within that sector; S4: Select the sector corresponding to the beam with the largest output power among the beams that have passed the stability test, and use a high-precision estimation method without preset parameters in this sector to achieve accurate estimation of the angle of arrival (DOA) of the corresponding signal component. S5: Perform time-frequency non-Gaussianity verification on the detected signal angle of arrival (DOA). If the signal component passes the non-Gaussianity test, add the corresponding angle of arrival (DOA) to the list of detected signal angles of arrival; otherwise, mark it as noise and do not include it in the final output result. S6: Update the weights using null techniques, generating zeros at the arrival angles of detected signal components to suppress their influence on subsequent detection. Repeat steps S2 to S6 until no new stable beams are detected.
[0040] In step S1 of this embodiment, the physical array is a minimum redundancy array (MRA). By performing Khatri-Rao product and spatial smoothing on the received data covariance matrix of the MRA, an equivalent virtual uniform linear array data covariance matrix is generated, and this virtual linear array is used for simultaneous multi-beam synthesis. Figure 2 As shown, taking a 6-element minimum redundancy array (MRA) as an example, its normalized element positions are {0,1,6,9,11,13}. Let the array receive data as... ,in Let be the number of snapshots. Under the narrowband far-field assumption, the receiving model is:
[0041] in It is an array manifold. For the first The steering vector of each information source. The normalized position of each array element. The number of unknown information sources. , These are the signal and noise matrices, respectively. The minimum redundancy array is listed based on the number of array elements. For example, a 4-element array is {0,1,4,6}, and a 5-element array is commonly {0,1,4,7,9}.
[0042] Under the assumption of ideal Gaussian white noise, the data covariance matrix of the minimum redundancy array (MRA) is:
[0043] in , which represents the signal power vector. This represents noise power.
[0044] For array data covariance matrix Vectorization yields the equivalent data reception vector:
[0045] in , Represents the Khatri-Rao product. Matrix For a manifold corresponding to a virtual array, the element positions are the difference set of the original MRA position set { }, . Each element corresponds to the equivalent data received by the virtual array element at its respective position. It was observed that... There are duplicate values, such as This indicates the existence of duplicate virtual array elements. After deduplication (removal)... After corresponding elements in the middle, a continuous aperture is obtained { 13, There are 27 virtual array elements, 12,…,13}, corresponding to the effective receiving vectors. .
[0046] Then, from Extract 14 consecutive subvectors of length 14 from the given data. Each subvector corresponds to a translated subarray, such as position {0:13}, { 1:12},…,{ 13:0}.
[0047] Incoherent averaging of the subarray covariance matrix:
[0048] In actual direction finding, the sample data covariance matrix is used. Alternative ,Right now:
[0049] Thus, the equivalent synthesis from a 6-element MRA to a 14-element virtual uniform linear array was achieved. The term "virtual uniform array synthesis" as used thereafter refers to the steps obtained through this method. It is a theoretical quantity, requiring an infinite number of sampling points. In practice, it is impossible to obtain an accurate one. Only can be used Approximate substitution This is a common and widespread practice in array signal processing.
[0050] In step S2 of this embodiment, multiple high-gain, narrow main lobe beams are generated simultaneously, with the main lobes pointing at uniform intervals to achieve complete coverage of the entire observation area, dividing the observation area into several non-overlapping sectors. Several high-gain, narrow beams with different main lobe directions are generated based on a virtual linear array, covering the entire observation area. In this embodiment, the linear array elements are directional antennas, and the observation area... Set the coverage sector width for each beam. Given a 3dB beamwidth, the number of beams required to completely cover the entire observation area is... for:
[0051] Adjacent beam main lobes point 3 dB apart. Let the beam main lobe direction be... If no signal is detected initially, then the first... The weights of each beam are set as follows:
[0052] in The number of elements in the virtual uniform linear array.
[0053] Definition of the first Each beam covers a sector For: within this angular range, the first Gain of each beam Larger than other beams, that is:
[0054] The antenna pattern and array configuration of the linear array elements, and the synthesized multi-beam distribution are as follows: Figure 3 As shown. Among them. Figure 3 (a) represents the antenna radiation pattern. Figure 3 (b) Represents the array structure (solid elements are real array elements, hollow elements are extended virtual array elements) and the distribution of synthetic multibeams.
[0055] In step S3 of this embodiment, the number of snapshots (sampling points) of the array receiving data is: Divide it into equal parts Subframe data. For each subframe of data, calculate the output power of each beam, i.e.:
[0056] in It is the virtual uniform linear array. The covariance matrix corresponding to the frame data. The DBSCAN clustering algorithm is used to detect outliers in the beam output power within each subframe, identifying the set of high-power beam indices whose output power is significantly higher than the average level. If a certain beam index is in all All of them were identified as high-power anomalies in each subframe, that is... If the output power of the beam is considered to have time-domain stability, the corresponding sector is determined to have a signal source.
[0057] In step S4 of this embodiment, among all stable beam sectors where the information source is determined to exist, the sector with the strongest output power is selected. This method employs a high-precision estimation technique that requires no preset parameters to accurately estimate the angle of arrival (DOA) of a signal. In this example, a Capon filter is used, namely:
[0058] In step S5 of this embodiment, the time-frequency non-Gaussianity of the signal component detected in step S4 is verified to distinguish the real signal from environmental Gaussian noise. First, a Capon filter is used in the detected direction... Extracting the signal, i.e.:
[0059] in These are the Capon filter coefficients corresponding to the actual array (a 6-element minimum redundancy array). It is the array receiving data corresponding to the actual array.
[0060] For the extracted signal Perform a short-time Fourier transform to obtain its time-frequency coefficients. To meet the statistical independence requirement, the real part of the time-frequency coefficients is downsampled to obtain... .
[0061] Calculate the sample kurtosis of the downsampled sequence ,Right now
[0062] like If the signal is true, its DOA is determined to be a genuine source and included in the final output; otherwise, it is marked as Gaussian noise and not displayed in the final result.
[0063] In step S6 of this embodiment, the angle of arrival of the signal component corresponding to the strongest beam output power is detected. Then, the contribution of this signal component is suppressed using null techniques. In this embodiment, all beams except the detected beams are weighted using the following method:
[0064]
[0065] in , . For the first The main lobe of each beam points towards... Let DOA be the angle of arrival (DOA) of the detected signal component. Solve for the... The beam weights are Repeat steps S2 to S5 until all signal components have been detected. This process constitutes a sequential detection mechanism that detects signal components sequentially based on the beam output power. When all signals have been detected and suppressed, the beam output power contains only the contribution of noise. Because the noise is not concentrated in the spatial domain, the beam power does not have temporal stability. The algorithm will terminate automatically.
[0066] like Figure 4 The figure shown is an example of the tracking results of the robust angle of arrival estimation method for unknown source numbers based on virtual uniform linear array beamforming. The algorithm framework proposed in this invention not only avoids the prior requirement for the number of sources, but also significantly improves the robustness of the sources under non-ideal conditions such as power imbalance, colored noise, and array system errors.
[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming, characterized in that: The method includes the following steps: S1. Based on the minimum redundancy array, the receiving data model is set up. The covariance matrix of the received data is processed by Khatri-Rao product and spatial smoothing algorithm to generate the data covariance matrix of the equivalent virtual uniform linear array. Then, multi-beam synthesis is performed to generate the virtual uniform linear array. S2. Multiple high-gain narrow beams are generated using a virtual uniform linear array. The main lobe of each beam is evenly spaced within the observation area, and the observation area is divided into several non-overlapping sectors. S3. Within each beam coverage sector, the temporal stability of the beam output power in multiple data block subframes is analyzed, and the presence of a signal source within the sector is determined based on the temporal stability. S4. Select the sector corresponding to the beam with the largest output power among the beams that have passed the stability test, and use a high-precision estimation method without preset parameters to estimate the angle of arrival of the corresponding signal component. S5. Perform time-frequency non-Gaussianity verification on the detected signal angle of arrival. If the signal component passes the non-Gaussianity test, add the corresponding angle of arrival to the list of detected signal angles of arrival; otherwise, mark it as noise. S6. Update the weights using null techniques, generating zeros at the arrival angles of detected signal components to suppress the influence of signal components on subsequent detection; repeat steps S2 to S6 until no new stable beams are detected.
2. The robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 1, characterized in that: The virtual uniform linear array synthesis process in step S1 is as follows: Let the number of array elements be... Minimum redundancy array settings for array receiving data , For the number of snapshots; Under the assumption of ideal Gaussian white noise, find the data covariance matrix of the minimum redundancy array MRA. ; The data covariance matrix of the minimum redundancy array Vectorization is performed to obtain the equivalent data reception vector. After deduplication of the equivalent data received vector, the corresponding effective received vector is obtained. ; from Extract in sequence A length of continuous subvectors Each subvector corresponds to a translated subarray. The covariance matrix of each subarray is then subjected to incoherent averaging to achieve the desired result from a subarray with a number of elements. Minimum Redundancy Array to Equivalent synthesis of virtual uniform linear arrays of array elements.
3. The robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 2, characterized in that: In step S1, under the narrowband far-field assumption, the received data Represented as: in It is an array manifold. For the first The steering vector of each source is represented as: ; The normalized position of each array element; The number of unknown sources , These are the signal and noise matrices, respectively. Under the assumption of ideal Gaussian white noise, the data covariance matrix of the minimum redundancy array (MRA) for: in , representing the signal power vector Noise power; Equivalent data receiving vector Represented as: in , Represents the Khatri-Rao product; matrix For a manifold corresponding to a virtual array, the element positions are the difference set of the original MRA position set { }, The number of array elements are respectively The elements in the normalized element positions of the minimum redundancy array; The incoherent averaging process of the covariance matrix of each subarray is expressed as follows: .
4. The robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 2, characterized in that: In step S2, the sector partitioning process is as follows: Let the observation area of the linear array element be... The coverage sector width of each beam According to the observation area and coverage sector width Determine the number of beams to be generated ; Set the main lobe of adjacent beams to point to beams spaced apart by a width of [value]. The main lobe of the beam points to When no signal is initially detected, calculate the first... Weight of each beam , and thus define the first Each beam covers a sector .
5. A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 4, characterized in that: In step S2, the number of beams The calculation method is as follows: No. The weights of each beam are calculated as follows: in, The number of elements in the virtual uniform linear array; No. Each beam covers a sector This refers to the first [angle range]. Gain of each beam Larger than other beams, that is: .
6. A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 4, characterized in that: In step S3, the process of determining whether a signal source exists based on the temporal stability of the multi-subframe beam output power is as follows: First, the array receives data according to its snapshot number. Average score The data block subframes are divided into several data block subframes, and the output power of each beam in each data block subframe is calculated. : in, It is the virtual uniform linear array. The covariance matrix corresponding to the frame data; Then, a clustering algorithm is used to detect outliers in the beam output power within each subframe, identifying the high-power beam index set whose output power is significantly higher than the average level. If a certain beam index is in all All of them were identified as high-power anomalies in each subframe, that is... If the output power of the beam is stable in the time domain, then the corresponding sector is considered to have a source; otherwise, it is considered not to have a source.
7. A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 6, characterized in that: In step S4, among all stable beam sectors where the information source is determined to exist, the sector with the strongest output power is selected. The Capon filter is used to estimate the angle of arrival of the signal. .
8. A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 7, characterized in that: In step S5, the process of verifying the time-frequency non-Gaussianity of the detected signal angle of arrival is as follows: First, at the angle of arrival of the detected signal component Signal extraction using Capon filter ; For the extracted signal Perform a short-time Fourier transform to obtain its time-frequency coefficients. The real parts of the time-frequency coefficients are downsampled to obtain the downsampled sequence. ; Finally, the sample kurtosis of the downsampled sequence is calculated. The difference between the sample kurtosis and the preset kurtosis is used to determine whether it is a real signal source or noise. If it exceeds the preset threshold, it is a real signal source; otherwise, it is noise.
9. A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 8, characterized in that: In step S5, in the detected direction Extract signal The method is as follows: in, These are the Capon filter coefficients corresponding to the actual array. This is the data received by the actual array; Sample kurtosis of downsampled sequences Represented as: Set the preset kurtosis to The preset threshold is ,like If the signal is true, its DOA is determined to be a genuine source and included in the final output; otherwise, it is marked as Gaussian noise and not displayed in the final result.
10. A robust angle-of-arrival estimation method based on virtual uniform linear array beamforming according to claim 8, characterized in that: In step S6, the method for suppressing the influence of the detected signal components using the null technique is as follows: After each detected signal component's angle of arrival, the beamforming weights are updated, and nulls are formed at the detected angles of arrival. in, , ; For the first The main lobe of each beam points towards... The angle of arrival of the detected signal component is obtained; the th... The beam weights are .