Direction of arrival estimation method, radar system and vehicle

Abstract

The invention provides a direction of arrival estimation method, a radar system and a vehicle, and relates to the field of signal processing. The method comprises the following steps: acquiring an original array signal of a radar and position information of a receiving array element corresponding to the original array signal; and according to the original array signal and the position information, direction-of-arrival estimation is carried out on the detection target. According to the embodiment of the invention, when the direction of arrival estimation is realized, the array signal is processed, the position information of the array elements is embedded into the array signal, and the data features are increased, so that the direction of arrival estimation precision is improved.

Description

Technical Field

[0001] The present application relates to the field of signal processing, and in particular to a direction of arrival estimation method, a radar system and a vehicle. Background Art

[0002] Vehicles perceive the surrounding environment through on-board radar systems. With the development of intelligent driving, the performance requirements of on-board radar systems are getting higher and higher to ensure the personal safety of users and the safety of vehicles. In the on-board radar system, the detection of target position and direction is completed through direction of arrival (DOA) estimation. With the continuous development of artificial intelligence, DOA estimation based on deep learning methods has made progress that is superior to traditional methods, providing faster reasoning and enhanced super-resolution performance under low signal-to-noise ratio (SNR) conditions. However, in related technologies, the direction of arrival is estimated directly through the received array signal. The array signal is not processed and a lot of data will be lost only through the array signal, resulting in low estimation accuracy. Summary of the invention

[0003] The embodiments of the present application provide a direction of arrival estimation method, a radar system, and a vehicle to solve the technical problem of low estimation accuracy existing in the prior art.

[0004] In a first aspect, the present application provides a method for estimating a direction of arrival, the method comprising:

[0005] Obtaining the original array signal of the radar and the position information of the receiving array element corresponding to the original array signal;

[0006] According to the original array signal and position information, the direction of arrival of the detected target is estimated.

[0007] In one embodiment, estimating the direction of arrival of the detected target according to the original array signal and the position information includes:

[0008] The original array signal and the position information are converted into the frequency domain, and the conversion result is embedded into the original array signal to obtain the target array signal;

[0009] According to the target array signal, the direction of arrival of the detected target is estimated.

[0010] In one embodiment, the original array signal and the position information are converted into the frequency domain, and the conversion result is embedded into the original array signal to obtain the target array signal, including:

[0011] Performing frequency domain conversion on the original array signal to obtain a first conversion result, and performing frequency domain conversion on the position information to obtain a second conversion result;

[0012] The first conversion result and the second conversion result are both embedded into the original array signal to obtain a target array signal.

[0013] In one embodiment, the original array signal and the position information are converted into the frequency domain, and the conversion result is embedded into the original array signal to obtain the target array signal, including:

[0014] Merging the original array signal with the position information to obtain a merged result;

[0015] Performing frequency domain conversion on the combined result to obtain a third conversion result;

[0016] The third conversion result is embedded into the original array signal to obtain the target array signal.

[0017] In one embodiment, the frequency domain conversion includes a Fast Fourier Transform.

[0018] In one embodiment, estimating the direction of arrival of the detected target according to the target array signal includes:

[0019] The target array signal is input into the classification model to obtain the angle classification result of the target array signal. The angle classification result is used to characterize the arrival direction of the detected target.

[0020] In one embodiment, the classification model is trained based on target array signals corresponding to historical array signals.

[0021] In one embodiment, the loss function used in training the classification model is a binary cross entropy loss function.

[0022] In one embodiment, obtaining the original array signal of the radar includes:

[0023] Obtain the receiving signal of each array element in the radar array antenna;

[0024] The original array signal is constructed according to the received signal of each array element.

[0025] In one embodiment, constructing an original array signal according to a received signal of each array element includes:

[0026] According to each array element for K signal sources The received signal is used to construct the original array signal.

[0027] In one embodiment, each array element is used to generate K signal sources. The received signal is used to construct the original array signal, including:

[0028] According to K signal sources The direction angle and the relative position relationship between the array elements in the array antenna are used to determine the direction vector of each array element;

[0029] Based on the direction vector and K signal sources , construct the original array signal.

[0030] In one of the embodiments, the original array signal also includes Gaussian white noise.

[0031] In one embodiment, K signal sources Each signal source in is a narrowband far-field signal.

[0032] In one embodiment, the array antenna is a uniform array or a sparse array.

[0033] In one embodiment, the array antenna is a uniform array; constructing an original array signal according to a received signal of each array element includes:

[0034] Obtaining a random position coefficient corresponding to each array element, where the random position coefficient is used to characterize whether the received signal of the array element is valid;

[0035] The original array signal is constructed according to the received signals of the array elements whose random position coefficients are effective in representing the received signals.

[0036] In one embodiment, obtaining position information of a receiving array element corresponding to an original array signal includes:

[0037] Obtaining an initial position matrix of the array antenna, where the initial position matrix is ​​constructed based on an array element structure of the array antenna;

[0038] When the random position coefficient corresponding to the array element represents a valid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to a first value; when the random position coefficient corresponding to the array element represents an invalid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to a second value.

[0039] In one embodiment, constructing an original array signal according to a received signal of each array element includes:

[0040] Perform single-snap sampling on the received signal of each array element to obtain the sampling signal corresponding to each array element;

[0041] The original array signal is constructed according to the sampling signal corresponding to each array element.

[0042] In a second aspect, the present application further provides an electronic device, wherein the computer comprises: a processor, and a memory communicatively connected to the processor;

[0043] Memory stores computer-executable instructions;

[0044] The above method is implemented when the processor executes the computer-executable instructions stored in the memory.

[0045] In a third aspect, the present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the above method.

[0046] In a fourth aspect, the present application also provides a computer program product, comprising computer execution instructions, which implement the above method when executed by a processor.

[0047] In a fifth aspect, the present application provides a radar system, the system comprising a signal transmitter, an array antenna and a signal processor; the signal transmitter is used to transmit a signal, and the array antenna is used to receive a signal;

[0048] The signal processor is used to determine the direction of arrival estimation result of the detection target according to the received signal, and is specifically used to execute the method steps of any of the above embodiments.

[0049] In one embodiment, the array antenna is a uniform array, and the array element arrangement interval d1 of the uniform array is related to λ, where λ is the wavelength of the highest frequency signal in the received signals received by the array antenna.

[0050] In one embodiment, d1 ≥ 0.5λ.

[0051] In one embodiment, the array antenna is a sparse array.

[0052] In one of the embodiments, the interval between any two array elements in the sparse array is n*d2, where d2 is related to λ, λ is the wavelength of the highest frequency signal in the received signals received by the array antenna; and n is a positive integer.

[0053] In one embodiment, d2 ≥ 0.5λ.

[0054] In a sixth aspect, the present application also provides a vehicle, characterized in that the vehicle includes a radar system according to any one of the above-mentioned embodiments.

[0055] The present application provides a method for estimating a direction of arrival, a radar system, and a vehicle. The method includes: obtaining the original array signal of the radar, and the position information of the receiving array element corresponding to the original array signal; and estimating the direction of arrival of the detected target according to the original array signal and the position information. When implementing the direction of arrival estimation, the embodiment of the present application processes the array signal, embeds the position information of the array element in the array signal, and increases the data features to improve the accuracy of the direction of arrival estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0057] Figure 1 A schematic diagram of an application scenario of a direction of arrival estimation method provided in an embodiment;

[0058] Figure 2 The process of the direction of arrival estimation method provided in one embodiment is as follows Figure 1 ;

[0059] Figure 3 A schematic diagram of the system structure of DOA estimation provided in one embodiment;

[0060] Figure 4 The process of the direction of arrival estimation method provided in one embodiment is as follows Figure 2 ;

[0061] Figure 5 A schematic diagram of a target array signal generation process provided in an embodiment Figure 1 ;

[0062] Figure 6 A schematic diagram of a target array signal generation process provided in an embodiment Figure 2 ;

[0063] Figure 7 A schematic diagram of a network structure of a classification model provided in one embodiment;

[0064] Figure 8 A schematic diagram of the structure of a Mixer Layer in a classification model provided in one embodiment;

[0065] Fig. 9 A schematic diagram of the structure of an MLP layer in a classification model provided in one embodiment;

[0066] Fig.10 The process of the direction of arrival estimation method provided in one embodiment is as follows Figure 3 ;

[0067] Fig.11 A schematic diagram of an array model provided in an embodiment;

[0068] Fig.12 A comparison diagram of angle hit rate algorithms for a uniform line array provided in one embodiment;

[0069] Fig.13 A comparison diagram of angle hit rate algorithms for a sparse line array provided in one embodiment;

[0070] Fig.14A comparison diagram of real data test of a uniform line array provided in an embodiment;

[0071] Fig.15 A comparison diagram of real data test of a sparse line array provided in one embodiment;

[0072] Fig.16 A schematic diagram of the structure of a radar system provided in one embodiment;

[0073] Fig.17 A hardware structure diagram of an electronic device provided in one embodiment.

[0074] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0075] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0076] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art to which the present application belongs. The terms used herein in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. It is understood that the terms "first", "second", etc. used in the present application can be used to describe various information or data in this article, but these elements are not limited by these terms. These terms are only used to distinguish the first information from another information. For example, without departing from the scope of the present application, the first action information can be referred to as the second action information, and similarly, the second action information can be referred to as the first action information. Both the first action information and the second action information are action information, but they are not the same action information.

[0077] First, the terms involved in this application are explained:

[0078] Direction of Arrival (DOA) estimation: also known as angle spectralestimation or angle of arrival estimation. A signal source has many possible propagation paths and angles of arrival.

[0079] IAA: Iterative adaptive approach, adaptive iterative algorithm.

[0080] FFT: fast fourier transform, fast Fourier transform.

[0081] DBF: digital beam forming.

[0082] DNN: deep neural networks.

[0083] FCN: fully convolution networks, fully convolutional neural networks.

[0084] Fully Connected Layer: Each neuron in the fully connected layer is connected to all neurons in the previous layer. Each connection has a weight to adjust the strength of information transmission, and each neuron also has a bias term.

[0085] Vehicles perceive the surrounding environment through on-board radar systems. With the development of intelligent driving, the performance requirements of on-board radar systems are getting higher and higher to ensure the personal safety of users and the safety of vehicles. In the on-board radar system, the detection of target position and direction is completed through direction of arrival (DOA) estimation. With the continuous development of artificial intelligence, DOA estimation based on deep learning methods has made progress that is superior to traditional methods, providing faster reasoning, enhanced super-resolution and robust performance under low signal-to-noise ratio (SNR) conditions. However, in related technologies, the direction of arrival is estimated directly through the received array signal. The array signal is not processed and a lot of data will be lost only through the array signal, resulting in low estimation accuracy.

[0086] In view of the above technical problems, the embodiment of the present application provides a method for estimating the direction of arrival, which can be applied to the scenario of estimating the direction of arrival after real-time data collection, and can also be applied to the scenario of estimating the direction of arrival after data collection. The embodiment of the present application does not specifically limit this. For ease of understanding, the embodiment of the present application takes the application in the scenario of real-time data acquisition and direction of arrival estimation as an example. The method provided in the embodiment of the present application can realize accurate direction of arrival estimation of the detected target.

[0087] The method provided in the embodiments of the present application can be applied to Figure 1In the application environment shown. Among them, the acquisition device 102 is used to collect the echo signal reflected by the detection target in real time. The acquisition device 102 communicates with the signal processing unit 104 through the network, and sends the collected echo signal to the signal processing unit 104, which processes and analyzes the echo signal to achieve the direction of arrival estimation. The data storage system can store the data that the signal processing unit 104 needs to process. The data storage system can be integrated on the signal processing unit 104, or it can be placed on the cloud or other network servers.

[0088] The acquisition device 102 refers to a terminal device including a sensor or a signal receiver, which may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc.

[0089] In one embodiment, Figure 2 As shown, a method for estimating the direction of arrival is provided, and the method is applied to Figure 1 The signal processing unit 104 in FIG. 1 is used as an example to illustrate, and the following steps are included:

[0090] Step 102, obtaining the original array signal of the radar and the position information of the receiving array element corresponding to the original array signal;

[0091] Step 104, estimating the direction of arrival of the detected target based on the original array signal and the position information.

[0092] Multiple sensors or receivers are set at different positions in space to form an array. The array signal can be obtained by receiving the spatial signal field through this array. More accurate and reliable information can be obtained by simultaneously receiving and processing the array signal. In a radar system, multiple identical single antennas (such as symmetrical antennas) are arranged according to a certain pattern to form an array antenna, also known as an antenna array. The independent units that make up the array antenna are called array elements or antenna units (the array elements are used as an example in the embodiment of the present application), and the array signal is obtained through the array antenna.

[0093] According to the arrangement rules of array elements, array antennas can be divided into uniform arrays and sparse arrays. It can be understood that a uniform array refers to an array antenna in which the spacing between any two adjacent array elements is equal and the amplitude patterns of all array elements are the same. A sparse array refers to an antenna array design that reduces the number of array elements according to a certain rule. The array signal in the embodiment of the present application is obtained by an array antenna in a radar system, and the structure of the array antenna is not specifically limited.

[0094] The original array signal refers to the array signal obtained by the array antenna and has not yet been processed to realize the wave direction estimation. When in use, the array antenna is usually arranged in a fixed position to detect the detection target within the detection range. After each array element in the array antenna receives the signal, it sends the received signal to the signal processing unit through the corresponding channel. After receiving the received signal of each array element, the signal processing unit needs to process the received signals of all array elements to obtain the mathematical model of the array signal corresponding to the array antenna, and then analyze it. The original array signal in the embodiment of the present application is the array element signal constructed by the received signal of each array element.

[0095] Exemplarily, the system structure of DOA estimation of the radar system is as follows: Figure 3 As shown. Among them, the target space 301 is the space that the array antenna can detect, and the detection target reflects the transmitted signal, which is received by the array antenna. Therefore, for the array antenna, each detection target in the target space can be regarded as a signal source. Each channel in the observation space 302 corresponds to an array element. After each array element obtains the received signal, it sends the received signal to the processor 3031 in the estimation space 303, and the processor 3031 processes the received signal to achieve DOA estimation. The processor 3031 will first process the received signals of all array elements to obtain array signals. For example, the processor 3031 includes an adaptive array processor, and the signals received by each array element enter the adaptive array processor for weighted addition to obtain the original array signal.

[0096] The position information of the receiving array element corresponding to the original array signal is used to characterize the position information of the receiving array element corresponding to the receiving signal for constructing the original array signal in space, and the position information is a characterization of the relative position in the array antenna. For example, when acquiring the original array signal, all array elements in the array antenna participate in the reception of the signal, and a position matrix can be constructed in combination with the structure of the array antenna to represent the position information.

[0097] The spatial spectrum represents the energy distribution of the signal in each space, which is actually the direction of arrival of the array signal in space. Therefore, DOA estimation can be completed by analyzing the spatial spectrum of the array signal. The position information of the receiving array element corresponding to the original received signal directly contains the position information of the receiving array element in space. Therefore, by analyzing the original array signal and the corresponding position information of the receiving array element at the same time, the number of features based on DOA estimation can be increased, which can effectively improve the accuracy of DOA estimation results.

[0098] In the method provided in the above embodiment, the original array signal of the radar and the position information of the receiving array element corresponding to the original array signal are obtained; the direction of arrival of the detected target is estimated based on the original array signal and the position information. When the embodiment of the present application realizes the direction of arrival estimation, the array signal is processed, the position information of the array element is embedded in the array signal, and the data features are increased to improve the accuracy of the direction of arrival estimation.

[0099] In one embodiment, if Figure 4 As shown, based on the original array signal and position information, the direction of arrival of the detected target is estimated, including:

[0100] Step 402, performing frequency domain conversion on the original array signal and the position information, and embedding the conversion result into the original array signal to obtain the target array signal;

[0101] Step 404: estimate the direction of arrival of the detected target according to the target array signal.

[0102] By embedding the position information in the original array signal, the target array signal is obtained, and the DOA estimation is completed based on the target array signal. It should be noted that the position information is embedded in the original array signal to increase the data dimension of the signal, and the original array signal is not processed by the position information, such as directly merging the original array signal with the position information to obtain the target array information.

[0103] According to the above description, DOA estimation is actually the spatial spectrum estimation of the array signal. The research object of DOA estimation processing is actually the spatial domain signal, so it is necessary to convert the original array signal from the time domain to the frequency domain. In one embodiment, the frequency domain conversion of the original array signal and the position information can be achieved by fast Fourier transform.

[0104] The target array signal includes the original array signal, the frequency domain conversion result of the original array signal, and the frequency domain conversion result of the position information. In one embodiment, the conversion result is embedded into the original array signal to obtain the target array signal, including:

[0105] Performing frequency domain conversion on the original array signal to obtain a first conversion result, and performing frequency domain conversion on the position information to obtain a second conversion result;

[0106] The first conversion result and the second conversion result are both embedded into the original array signal to obtain a target array signal.

[0107] The first conversion result is the frequency domain conversion result of the original array signal, and the second conversion result is the frequency domain conversion result of the position information. The original array signal is firstly converted in the frequency domain, and then the first conversion result and the second conversion result are embedded into the original received signal to obtain the target array signal. The process of generating the target array signal is as follows: Figure 5 shown.

[0108] By converting the original array signal and the position information respectively and then embedding them, it can be ensured that the position information does not affect the original array signal, thereby ensuring the validity of the original array signal in the target array signal.

[0109] In one embodiment, the original array signal and the position information are converted into the frequency domain, and the conversion result is embedded into the original array signal to obtain the target array signal, which further includes:

[0110] Merging the original array signal with the position information to obtain a merged result;

[0111] Performing frequency domain conversion on the combined result to obtain a third conversion result;

[0112] The third conversion result is embedded into the original array signal to obtain the target array signal.

[0113] Among them, merging the original array signal with the position information means directly merging the original array information in the data dimension through the position information. During merging, the position information will not affect the data itself in the original array signal. The process of generating the target array signal is as follows: Figure 6 shown.

[0114] When the original array signal is combined with the position information, the original array signal needs to be stored first to ensure that the accurate original array signal can be obtained when the third conversion result is embedded in the original array signal, rather than the processed original array signal. The third conversion result includes the frequency domain conversion result of the original array signal and the frequency domain conversion result of the position information.

[0115] By first merging the position information into the original array signal, it is possible to perform frequency domain conversion only once, thus reducing the amount of calculation.

[0116] After processing the original array signal and position information, the obtained target array signal will eventually include a large number of data features, realize data enhancement, and complete DOA estimation based on the target array signal.

[0117] In one embodiment, estimating the direction of arrival of the detected target according to the target array signal includes:

[0118] The target array signal is input into the classification model to obtain the angle classification result of the target array signal. The angle classification result is used to characterize the arrival direction of the detected target.

[0119] The classification model performs a series of processing on the target array information, including but not limited to feature extraction and classification prediction, performs classification prediction through the extracted features, and outputs the angle classification result of the target array signal. The classification model can be constructed based on a deep learning algorithm, such as using DNN and FCN to construct a deep learning network to process the target array signal, and also, for example, by inputting the target array signal into a CNN model to obtain the DOA estimation result.

[0120] The present application embodiment provides a classification model based on MLP-Mixer. The specific structure of the network model is shown in Figure 7 In the MLP-Mixer network structure, the fully-connected layer is first used to convert the N×1 input into a high-dimensional space, and then these high-dimensional features are sent to N Mixer Layers for feature extraction and mixing, and finally a fully-connected layer is used for classification prediction.

[0121] Among them, the structure of Mixer Layer is as follows Figure 8 As shown in the figure, the network consists of two Layer-Norms (layer normalization) and MLP (multi-layer perceptron). The input first passes through the first Layer-Norm, changes the input distribution in the channel dimension, and then transposes the data dimension and sends it to the MLP layer; after the MLP layer completes feature extraction, it is transposed again, and the original input is added and fused through jump connections. After completing the information exchange and fusion in the data dimension, the operation of the LN layer and the MLP layer is repeated to obtain the output of the Mixer Layer.

[0122] The structure of the MLP layer is as follows Fig. 9As shown in the figure, it consists of two fully-connected layers and a GELU activation function. The GELU activation function does not directly clip to zero when processing negative numbers, which helps to avoid the gradient vanishing problem during network training and accelerates convergence. The classification output of MLP-Mixer is a binary probability prediction vector, which is used as the angle classification result to represent the probability prediction of whether there is a signal within the angle grid range. Among them, the angle grid range is determined according to the detectable range of the array antenna. For example, the angle is -60°~60°. The angle grid range is divided to determine the number of classification labels of the classification model. For example, if every 2° is a label range, there are 60 classification labels. Each element in the binary probability prediction vector represents the probability that the direction angle of the detection target corresponding to the array signal is within the corresponding classification label.

[0123] The classification model is obtained by training the original array signal obtained in the historical stage and the corresponding angle true value. When training the classification model, the original array signal can be used directly for training, or the original array signal can be processed, and the position information can be embedded in it, and then the target array signal can be used for training. By embedding the position information, the trained classification model can learn more features and effectively improve the classification effect of the classification model.

[0124] The training loss function of the classification model can adopt the binary cross entropy loss function, and the specific formula is as follows:

[0125] ;

[0126] Among them, the base of log is e; y represents the label of the sample, the positive class is 1, and the negative class is 0; y' represents the probability of the sample being predicted as the positive class.

[0127] In the method provided in the above embodiment, a classification model is constructed by a deep learning algorithm, and DOA estimation is completed by the classification model, which provides faster reasoning, enhanced super-resolution and robust performance under low signal-to-noise ratio (SNR) conditions.

[0128] In one embodiment, if Fig.10 As shown, the original array signal of the radar is obtained, including:

[0129] Step 1002, obtaining a received signal of each array element in the radar array antenna;

[0130] Step 1004: construct an original array signal according to the received signal of each array element.

[0131] The received signal of the array element is directly obtained by the array element. Combined with the arrangement rules of all the array elements in the array antenna, the received signals of all the array elements are processed to construct the array signal belonging to the array antenna. The array signal is essentially a collection of a bunch of received signals, a data matrix. The construction of the array signal is explained based on the following example:

[0132] like Fig.11 As shown, the array antenna is a uniform linear array with M array elements and an array element spacing of d. The scene is set with K narrowband far-field signal sources. to be incident on the array antenna, where the directions of the K narrowband far-field signals are , It represents the angle between the normal line of the uniform linear array and the incoming wave signal in space. Taking array element 1 as the reference array element, the received signal of array element 1 is:

[0133] ;

[0134] Then the received signal of array element m is:

[0135] ;

[0136] Where t represents the number of snapshots, and λ represents the wavelength of the carrier frequency of the signal source.

[0137] make: ,

[0138] Then the vector matrix form of constructing the array signal is:

[0139] ;

[0140] in, is the array manifold matrix, is the source matrix, is the noise matrix.

[0141] The above example is based on a uniform linear array. If the array antenna is a sparse array, the corresponding received signal of element m is:

[0142] ;

[0143] in, Indicates the interval between array element m and array element 1.

[0144] In actual application scenarios, the hardware structure of the array antenna of the radar system is known, that is, the model of the array antenna is known (uniform array or sparse array), and the original array signal is constructed by the received signals of all array elements in the array antenna, and the mathematical model of the original array signal is obtained, and then DOA estimation is performed. At this time, the position information of the receiving array element corresponding to the original array signal should be known and fixed. The method provided in the embodiment of the present application can be applied to uniform arrays and sparse arrays.

[0145] In actual application scenarios, such as in a vehicle-mounted radar system, the vehicle-mounted radar system often has a limited number of signal snapshots, and sometimes even performs single-snap sampling on the received signal. Therefore, when constructing the original array signal, the single-snap signal can also be collected by the number of snapshots mentioned in the above embodiment, specifically including:

[0146] Perform single-snap sampling on the received signal of each array element to obtain the sampling signal corresponding to each array element;

[0147] The original array signal is constructed according to the sampling signal corresponding to each array element.

[0148] Among them, single-snap sampling means sampling the received signal only once, sampling the received signals of all array elements at the same time to obtain a single-snap original array signal. When the number of snapshots T is 1, the single-snap original array signal is:

[0149] ;

[0150] The method provided in the embodiment of the present application can be better applied to single-snapshot scenarios by performing data enhancement processing on array signals. Through data enhancement, some data lost in a single snapshot can be compensated to ensure the accuracy of DOA estimation.

[0151] In one embodiment, the array antenna is a uniform array; constructing an original array signal according to a received signal of each array element includes:

[0152] Obtaining a random position coefficient corresponding to each array element, where the random position coefficient is used to characterize whether the received signal of the array element is valid;

[0153] The original array signal is constructed according to the received signals of the array elements whose random position coefficients are effective in representing the received signals.

[0154] The random position coefficient is randomly generated during the construction of the original array signal, and is used to randomly select the received signals of several array elements from the uniform array to construct the original array signal, so as to achieve the effect of a sparse array. There is a one-to-one correspondence between the random position coefficient and the array element, and usually, the random position coefficient matrix of all array elements is constructed according to the spatial arrangement structure of all array elements in the array antenna. The random position coefficient has two values, where the first value represents that the received signal is valid, and the second value represents that the receiver of the array element is inactivated, the received signal is invalid, and does not participate in the construction of the original array signal.

[0155] In the method provided in the above embodiment, the effect of a sparse array is achieved in a uniform array through random position coefficients, which effectively reduces the amount of data processing and improves the efficiency of DOA estimation.

[0156] In one embodiment, obtaining position information of a receiving array element corresponding to an original array signal includes:

[0157] Obtaining an initial position matrix of the array antenna, where the initial position matrix is ​​constructed based on an array element structure of the array antenna;

[0158] When the random position coefficient corresponding to the array element represents a valid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to a first value; when the random position coefficient corresponding to the array element represents an invalid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to a second value.

[0159] The initial position matrix is ​​used to characterize the spatial structure of the array elements in the array antenna. For example, for a uniform linear array with 10 elements, the initial position matrix can be in the form of ,in, and its value indicates the position of the first array element and whether it is a receiving array element. In some embodiments, the initial position matrix may be composed of random position coefficients of each array element. The initial position matrix is ​​determined according to the initial hardware state of the array antenna, and when constructing the original array signal, if random position coefficients are generated, the initial position matrix is ​​updated by the random position coefficients.

[0160] Specifically, when the random position coefficient corresponding to the array element represents a valid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to the first value, and when the random position coefficient corresponding to the array element represents an invalid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to the second value. For example, for a uniform linear array of 10 array elements, the initial position matrix can be in the form of , indicating that the received signals of the 10 array elements are all valid. During use, if the random position coefficients of the 2nd, 4th, 5th, 7th, 8th, and 9th array elements are invalid, the initial position matrix becomes , which has the effect of a sparse array.

[0161] In the method provided in the above embodiment, the initial position matrix is ​​updated by random position coefficients, so as to achieve the effect of a sparse array on the basis of a uniform array, thereby reducing the amount of data processing.

[0162] In combination with all the methods provided in the above embodiments, the embodiment of the present application further provides a single snapshot signal DOA estimation method based on a deep neural network, the method comprising:

[0163] Step 1: Obtain a single snapshot original array signal;

[0164] The single-snapshot original array signal is obtained through the array antenna of the radar system. The array antenna can be a uniform array or a sparse array. The single-snapshot original array signal is constructed through the received signal of at least one array element in the array antenna.

[0165] Step 2: Perform frequency domain conversion on the position information of the receiving array element corresponding to the single snapshot original array signal and the single snapshot original array signal, and then embed them into the single snapshot original array signal through a channel to obtain a single snapshot target array signal;

[0166] The position information may be initially determined based on the structure of the array antenna, or may be determined by modifying a binary mask that determines a sparse array.

[0167] Step 3: Build a classification model based on MLP-Mixer and complete the training. Use the trained classification model to process the single-snap target array signal to obtain the angle classification result. The label with the highest predicted probability in the angle classification result is used as the DOA estimation result of the single-snap original array signal.

[0168] The method provided in the above embodiment is applicable to single-snapshot received data of uniform arrays and sparse arrays; and the method performs frequency domain conversion on the array signal and the receiving element position and embeds them into the array signal. After data enhancement, the robustness of the network model can be improved and the overfitting of the network can be prevented. In addition, the trained network model adopts the improved MLP-Mixer, which makes full use of the powerful nonlinear fitting ability of the deep neural network, can better converge during training, and improves the accuracy of DOA estimation.

[0169] The following combination Figure 12-15 And an embodiment of the present application further describes in detail the above-mentioned single snapshot signal DOA estimation method based on deep neural network.

[0170] like Figure 12-15 , which are respectively the DOA estimation simulation results of the present application on the uniform line array and sparse array models, wherein OURS (Ours) represents the data of the embodiment.

[0171] The parameters in the simulation process are set as follows: the number of elements of the uniform linear array is 10, the number of elements of the sparse linear array is 7, there are 3 signal sources, the field of view (FOV) of the radar system is [-30°, 30°], the angle interval is 1°, and a DOA angle grid is obtained.

[0172] In order to make the network more robust, multiple targets of 1-3 are randomly selected during the training data generation process, and 100,000 signal data with various SNR levels from 0dB to 30dB are generated with 5dB enhancement. The validation set uses the same configuration and generates 1,000 signals for each SNR level for testing.

[0173] The network was trained for 300 epochs with a batch-size of 1024 and end-to-end training using the Adam optimizer with a learning rate of 0.0001. Fig.12 and Fig.13 In order to evaluate the ability of the DOA estimation method of this application to distinguish close-range targets, the test experiment placed two targets symmetrically around the origin, with an angular interval of 2° between the targets. If the deviation between the estimated DOA and the actual position is within 1°, the test is classified as a hit, and the hit rate of 5000 Monte Carlo tests is calculated, and the SNR level is set to 40dB. At the same time, the IAA and FFT algorithms that can be used for single-snapshot DOA estimation are compared to evaluate the DOA estimation effect of this application.

[0174] according to Fig.12 It can be seen that in the configuration of uniform line arrays, the method of the present application is significantly better than the model-based method in terms of angle hit rate. Fig.13 In the present invention, the configuration of the sparse array reduces the hit rate of all methods to a certain extent, but compared with other methods, the method of the present application still maintains better performance. There is currently no open-source real data set available for DOA estimation, and simulation data is generally used to train and evaluate existing models. In order to fill this gap, a DOA estimation data set was collected in a parking scenario, using a radar-equipped vehicle to collect data from a corner reflector located 15 meters away, capturing signals from all possible directions. This process generates 195 high signal-to-noise ratio signals from different angles, each representing a target. By superimposing these vectors, a scene with multiple targets is simulated.

[0175] Fig.14 and 15 The DOA estimation results of the targets at 0° and 7° are shown in Figure 2. Fig.14It represents a uniform line array configuration. The first row and the first column are the array element position diagrams. The second column is the spatial spectrum diagrams of the comparison algorithms IAA and DBF. The third column is the spatial spectrum diagram of the method of the present invention. It can be seen that the DBF algorithm has only one spectrum peak and cannot distinguish the two targets. The method of the present invention has a smaller angle estimation error than the IAA method. Fig.15 Under the sparse line array configuration, both the DBF and IAA algorithms cannot distinguish the two targets, while the proposed method can also have good angular resolution under the sparse array.

[0176] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0177] Based on the same inventive concept, the embodiment of the present application also provides a radar system, such as Fig.16 As shown, the system includes a signal transmitter 1601, an array antenna 1602, and a signal processor 1603; the signal transmitter 1601 is used to transmit signals, and the array antenna 1602 is used to receive signals;

[0178] The signal processor 1603 is used to determine the direction of arrival estimation result of the detection target according to the received signal, and is specifically used to execute the method steps in any of the above embodiments.

[0179] In one embodiment, the array antenna is a uniform array, and the array element arrangement interval d1 of the uniform array is related to λ, where λ is the wavelength of the highest frequency signal in the received signals received by the array antenna.

[0180] In one embodiment, d1 ≥ 0.5λ.

[0181] In one embodiment, the array antenna is a sparse array.

[0182] In one of the embodiments, the interval between any two array elements in the sparse array is n*d2, where d2 is related to λ, λ is the wavelength of the highest frequency signal in the received signals received by the array antenna; and n is a positive integer.

[0183] In one embodiment, d2 ≥ 0.5λ.

[0184] In a specific application scenario, if the array antenna is a uniform array, the array element spacing in the uniform array can be 0.5λ, and the antenna aperture of the array antenna is 0.5(N-1)λ, where N is the number of array elements; if the array antenna is a sparse array, the antenna aperture of the sparse array can refer to the antenna aperture of the uniform array and be set to 0.5(N-1)λ, and then the interval between any two array elements can be 0.5n*λ, where n is a positive integer.

[0185] An embodiment of the present application also provides a vehicle, which includes the radar system in any one of the above embodiments.

[0186] like Fig.17 As shown, an embodiment of the present application provides an electronic device 1700 , which includes a memory 1701 and a processor 1702 .

[0187] The memory 1701 is used to store computer-executable instructions executable by the processor;

[0188] The processor 1702 implements each step of the method in the above embodiment when executing the computer execution instruction. For details, please refer to the relevant description in the above method embodiment.

[0189] Optionally, the memory 1701 can be independent or integrated with the processor 1702. When the memory 1701 is independently provided, the electronic device further includes a bus for connecting the memory 1701 and the processor 1702.

[0190] The embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, each step of the method in the above embodiment is implemented.

[0191] The embodiment of the present application also provides a computer program product, including computer-executable instructions, which implement the various steps of the method in the above embodiment when executed by a processor.

[0192] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer execution instructions in the computer program are executed, they can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0193] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0194] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for estimating a direction of arrival, characterized in that: The method comprises: Acquire the original array signal of the radar and the position information of the receiving array element corresponding to the original array signal; The direction of arrival of the detected target is estimated according to the original array signal and the position information.

2. The method according to claim 1, characterized in that The step of estimating the direction of arrival of the detected target according to the original array signal and the position information includes: Performing frequency domain conversion on the original array signal and the position information, and embedding the conversion result into the original array signal to obtain a target array signal; The direction of arrival of the detected target is estimated according to the target array signal.

3. The method according to claim 2, characterized in that The converting the original array signal and the position information into the frequency domain, and embedding the conversion result into the original array signal to obtain the target array signal includes: Performing frequency domain conversion on the original array signal to obtain a first conversion result, and performing frequency domain conversion on the position information to obtain a second conversion result; The first conversion result and the second conversion result are both embedded into the original array signal to obtain a target array signal.

4. The method according to claim 2, characterized in that: The converting the original array signal and the position information into the frequency domain, and embedding the conversion result into the original array signal to obtain the target array signal includes: Combining the original array signal with the position information to obtain a combined result; Performing frequency domain conversion on the combined result to obtain a third conversion result; The third conversion result is embedded into the original array signal to obtain a target array signal.

5. The method according to any one of claims 2 to 4, characterized in that: The frequency domain conversion includes a Fast Fourier Transform.

6. The method according to claim 2, characterized in that The step of estimating the direction of arrival of the detected target according to the target array signal comprises: The target array signal is input into a classification model to obtain an angle classification result of the target array signal, and the angle classification result is used to characterize the arrival direction of the detected target.

7. The method according to claim 6, characterized in that The classification model is obtained by training the target array signal corresponding to the historical array signal.

8. The method according to claim 7, characterized in that The loss function used in the training of the classification model is a binary cross entropy loss function.

9. The method according to claim 1, characterized in that: The obtaining of the original array signal of the radar comprises: Acquire a received signal of each array element in the array antenna of the radar; The original array signal is constructed according to the received signal of each array element.

10. The method according to claim 9, characterized in that The step of constructing the original array signal according to the received signal of each array element includes: According to each array element for K signal sources The received signal is used to construct the original array signal.

11. The method according to claim 10, characterized in that According to each array element, K signal sources The received signal is used to construct the original array signal, including: According to the K signal sources The direction angle and the relative position relationship between the array elements in the array antenna are used to determine the direction vector of each array element; Based on the direction vector and K signal sources , construct the original array signal.

12. The method according to claim 11, characterized in that The original array signal also includes Gaussian white noise.

13. The method according to claims 10-12, characterized in that The K signal sources Each signal source in is a narrowband far-field signal.

14. The method according to claim 9, characterized in that The array antenna is a uniform array or a sparse array.

15. The method according to claim 14, characterized in that The array antenna is a uniform array; and constructing an original array signal according to a received signal of each array element includes: Obtaining a random position coefficient corresponding to each array element, wherein the random position coefficient is used to indicate whether a received signal of the array element is valid; The original array signal is constructed according to the received signals of the array elements whose random position coefficients are effective in representing the received signals.

16. The method according to claim 15, characterized in that The obtaining the position information of the receiving array element corresponding to the original array signal includes: Acquire an initial position matrix of the array antenna, where the initial position matrix is ​​constructed based on an array element structure of the array antenna; When the random position coefficient corresponding to the array element represents a valid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to a first value; when the random position coefficient corresponding to the array element represents an invalid received signal, the element value corresponding to the array element in the initial position matrix is ​​set to a second value.

17. The method according to claim 14, characterized in that The step of constructing the original array signal according to the received signal of each array element includes: Perform single-snap sampling on the received signal of each array element to obtain the sampling signal corresponding to each array element; The original array signal is constructed according to the sampling signal corresponding to each array element.

18. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 17.

19. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 17 when executed by a processor.

20. A computer program product, characterized in that The method comprises a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 17.

21. A radar system, characterized in that: The system comprises a signal transmitter, an array antenna and a signal processor; the signal transmitter is used to transmit signals, and the array antenna is used to receive signals; The signal processor is used to determine the direction of arrival estimation result of the detected target based on the received signal, and is specifically used to execute the method steps described in any one of claims 1-17.

22. The radar system according to claim 21, characterized in that The array antenna is a uniform array, and the array element arrangement interval d1 of the uniform array is related to λ, where λ is the wavelength of the highest frequency signal in the received signals received by the array antenna.

23. The radar system according to claim 22, characterized in that d1≥0.5λ.

24. The radar system according to claim 21, characterized in that The array antenna is a sparse array.

25. The radar system according to claim 24, characterized in that The interval between any two array elements in the sparse array is n*d2, where d2 is related to λ, λ is the wavelength of the highest frequency signal in the received signals received by the array antenna; and n is a positive integer.

26. The radar system according to claim 25, characterized in that d2≥0.5λ.

27. A vehicle, characterized in that: The vehicle comprises the radar system of any one of claims 21-26.

Images