Method and device for estimating number of information sources and storage medium

文档序号:789290 发布日期:2021-04-09 浏览:10次 中文

阅读说明:本技术 一种信源数量估计的方法、装置和存储介质 (Method and device for estimating number of information sources and storage medium ) 是由 张瑞齐 张峰 于 2020-12-01 设计创作,主要内容包括:本申请的实施例提供了一种信源数量估计的方法、装置和存储介质。所述方法应用于毫米波雷达的接收系统中,包括:获取雷达天线阵列的阵列响应,所述雷达天线阵列包括至少一个天线单元;对所述阵列响应进行自相关计算得到相关矩阵;对所述相关矩阵做特征值分解,提取特征向量;设定预估计的信源数量,对所述预估计的信源数量分类,获得至少一个分类项;将所述特征向量输入与所述至少一个分类项对应的支持向量机,输出对应所述至少一个分类项的评估值;比较所述至少一个分类项的评估值,获得其中最大的评估值;根据所述最大的评估值对应的分类项确定信源数量。所述方法能够解决低信噪比区间性能和高信噪比区间性能存在偏差较大的问题。(Embodiments of the present application provide a method, apparatus, and storage medium for source number estimation. The method is applied to a receiving system of the millimeter wave radar, and comprises the following steps: acquiring an array response of a radar antenna array, wherein the radar antenna array comprises at least one antenna unit; performing autocorrelation calculation on the array response to obtain a correlation matrix; performing eigenvalue decomposition on the correlation matrix, and extracting eigenvectors; setting the number of pre-estimated information sources, and classifying the number of the pre-estimated information sources to obtain at least one classification item; inputting the feature vector into a support vector machine corresponding to the at least one classification item, and outputting an evaluation value corresponding to the at least one classification item; comparing the evaluation values of the at least one classification item to obtain the largest evaluation value; and determining the number of the information sources according to the classification item corresponding to the maximum evaluation value. The method can solve the problem that the performance of a low signal-to-noise ratio interval and the performance of a high signal-to-noise ratio interval have larger deviation.)

1. A method for estimating the number of information sources is applied to a receiving system of millimeter wave radar, and comprises the following steps:

acquiring an array response of a radar antenna array, wherein the radar antenna array comprises at least one antenna unit;

performing autocorrelation calculation on the array response to obtain a correlation matrix;

performing eigenvalue decomposition on the correlation matrix, and extracting eigenvectors;

setting the number of pre-estimated information sources, and classifying the number of the pre-estimated information sources to obtain at least one classification item;

inputting the feature vector into a support vector machine corresponding to the at least one classification item, and outputting an evaluation value corresponding to the at least one classification item; comparing the evaluation values of the at least one classification item to obtain the largest evaluation value; and determining the number of the information sources according to the classification item corresponding to the maximum evaluation value.

2. The method for source number estimation according to claim 1, the method further comprising a step of training a support vector machine corresponding to the at least one classification item, specifically comprising:

grouping training vectors in a training set; the training vector is a characteristic vector with information source quantity labels;

selecting training vectors of the corresponding group of the at least one classification item according to the grouping;

taking the training vector of the corresponding group as positive; forming a truth set by taking the training vectors outside the corresponding groups in the training set as negative values;

inputting the training set and the truth set into a support vector machine, training the support vector, the parameter vector and the variable of the support vector machine, and obtaining the trained support vector machine corresponding to the at least one classification item.

3. The method of source number estimation according to claim 1 or 2, the acquiring an array response of a radar antenna array, comprising:

acquiring an echo signal received by a radar antenna linear array; the radar antenna linear array comprises the M radar antenna units which are arranged at equal intervals; m is any natural number; the echo signals are reflection signals of targets with the same speed in the same detection distance;

and responding the linear array to form a row/column vector.

4. The method of source number estimation according to claim 1 or 2, the acquiring an array response of a radar antenna array, comprising:

acquiring an echo signal received by a radar antenna area array, wherein the radar antenna area array comprises W rows and M rows of radar antenna units which are arranged at equal intervals; the echo signals received by the radar antenna area array comprise echo signals received by the W-row and M-column radar antenna units; the echo signals are reflection signals of targets with the same speed in the same detection distance, and W, M is any natural number;

splicing the W line vectors of the echo signals into 1 linearly arranged line vector, wherein each line vector in the W line vectors comprises M echo signals; and obtaining the array response of the radar antenna area array.

5. The method of source number estimation according to claim 1 or 2, the acquiring an array response of a radar antenna array, comprising:

acquiring an echo signal received by a radar antenna circular array, wherein the radar antenna circular array comprises M radar antenna units which are uniformly arranged along the circumference, and M is any natural number; the echo signals received by the radar antenna circular array comprise M echo signals which are uniformly arranged along the circumference;

and stretching the M echo signals uniformly arranged along the circumference into linearly arranged row vectors at equal intervals to obtain the array response of the radar antenna circular array.

6. The method of source number estimation according to claim 1 or 2, said determining evaluation values of respective classification terms corresponding to a pre-estimated source number, comprising: and multiplying the support vector in the support vector machine by the kernel function, and summing the multiplication result and the coefficient to obtain the evaluation value of each classification item of the information source quantity to be pre-estimated.

7. The method of source number estimation as claimed in claim 6, the kernel function being a linear kernel function.

8. The method of source number estimation as claimed in claim 6, the kernel function being a polynomial kernel function.

9. The method of source number estimation as claimed in claim 6, the kernel function being a gaussian kernel function.

10. The method of source number estimation according to claim 6, the kernel function being a sigmiod kernel function.

11. The method of source number estimation according to claim 1, said auto-correlating said array responses to obtain a correlation matrix, comprising: and performing J times of snapshots on the array response, multiplying the array response at the jth snapshot time by a conjugate transpose vector of the jth snapshot time to obtain an autocorrelation matrix at the jth snapshot time, adding the autocorrelation matrices of the J snapshots, and averaging to obtain a correlation matrix, wherein J and J are both natural numbers.

12. The method of source number estimation according to claim 1, said auto-correlating said array responses to obtain a correlation matrix, comprising: and carrying out single snapshot on the array response, grouping the array response at the snapshot time to obtain at least one group of row vectors, multiplying the at least one group of row vectors by a conjugate transpose vector of the at least one group of row vectors to obtain an autocorrelation matrix of the at least one group of row vectors, and adding the autocorrelation matrices of the at least one group of row vectors and then averaging to obtain a correlation matrix.

13. The method of source number estimation according to claim 1, further comprising mapping the eigenvector by a functional process to obtain a second eigenvector.

14. The method of source number estimation according to claim 13, wherein the mapping the eigenvector by a function process to obtain the second eigenvector with the eigenvector as a first eigenvector comprises:

calculating the value of a logarithmic function by taking the characteristic value in the first characteristic vector as an independent variable of the logarithmic function;

and forming the second feature vector by taking the value of the logarithmic function as an element.

15. The method of source number estimation according to claim 13, wherein the mapping the eigenvector by a function process to obtain the second eigenvector with the eigenvector as a first eigenvector comprises:

comparing the characteristic values in the first characteristic vector to obtain the maximum value;

calculating a ratio of the characteristic value to the maximum value;

and calculating the value of the logarithmic function by taking the ratio as an independent variable of the logarithmic function, and forming the second feature vector by taking the value of the logarithmic function as an element.

16. The method of source number estimation according to claim 13, wherein the mapping the eigenvector by a function process to obtain the second eigenvector with the eigenvector as a first eigenvector comprises:

sorting the eigenvalues in the first eigenvector;

and calculating the value of the logarithmic function by taking the sorted characteristic values as the independent variables of the logarithmic function, and forming the second characteristic vector by taking the value of the logarithmic function as an element.

17. An apparatus for source number estimation, the apparatus comprising:

the data acquisition module is used for acquiring array response of a radar antenna array, and the radar antenna array comprises at least one antenna unit;

the correlation matrix calculation module is used for carrying out autocorrelation calculation on the array response to obtain a correlation matrix;

the characteristic extraction module is used for decomposing the characteristic value of the correlation matrix and extracting a characteristic vector;

the evaluation value calculation module is used for setting the number of pre-estimated information sources, classifying the number of the pre-estimated information sources and obtaining at least one classification item; and

the information source quantity determining module is used for inputting the feature vector into a support vector machine corresponding to the at least one classification item and outputting an evaluation value corresponding to the at least one classification item; comparing the evaluation values of the at least one classification item to obtain the largest evaluation value; and determining the number of the information sources according to the classification item corresponding to the maximum evaluation value.

18. The apparatus for source number estimation as recited in claim 17, the apparatus further comprising a training support vector machine module for grouping training vectors in a training set according to different source numbers; the training vectors are characteristic vectors with information source quantity labels, the training vectors of the corresponding groups of each classification item are selected according to the grouping, and the training vectors of the corresponding groups are taken as positive; and forming a truth set by taking training vectors in the training set except for the training vector corresponding to the corresponding group as negative, inputting the training set and the truth set into the support vector machine corresponding to each classification item, training the support vector, the parameter vector and the variable of the support vector machine, and obtaining the trained support vector machine corresponding to each classification item.

19. An electronic device comprising a memory and a processor; the processor is configured to execute computer-executable instructions stored in the memory, the processor executing the computer-executable instructions to perform the method of source number estimation as claimed in any one of claims 1 to 18.

20. A storage medium comprising a readable storage medium and a computer program stored in the readable storage medium for implementing the method of source number estimation as claimed in any one of claims 1 to 18.

Technical Field

The present application relates to the field of millimeter wave radar sensing, and in particular, to a method and an apparatus for estimating a number of signal sources, and a storage medium.

Background

The millimeter wave is an electromagnetic wave with the wavelength of 1-10 mm, the corresponding frequency range is 30-300 GHz, and the millimeter wave radar plays an important role in the fields of automatic driving, roadside perception and the like. Fig. 1 is a schematic diagram of a millimeter-wave radar, and as shown in fig. 1, a vehicle-mounted millimeter-wave radar system generally includes an oscillator, a transmitting antenna, a receiving antenna, a mixer, a processor, a controller, and other devices. The oscillator generates a radar signal with a frequency that increases linearly with time, typically a Frequency Modulated Continuous Wave (FMCW), which is an electromagnetic wave with a frequency that varies linearly with time. The linear variation is generally a linear variation within one transmission period. In particular, the waveform of the chirped continuous wave is typically a sawtooth or triangular wave, or other possible waveforms are possible, such as a chirped step frequency waveform, etc. A part of the radar signal is output to the frequency mixer through the directional coupler to serve as a local oscillation signal, a part of the radar signal is transmitted through the transmitting antenna, the receiving antenna receives the radar signal transmitted out, the radar signal is reflected back after encountering a target in front of a vehicle, and the frequency mixer mixes the received radar signal with the local oscillation signal to obtain an intermediate frequency signal. The intermediate frequency signal contains information such as the relative distance, velocity, and angle of the target to the radar system. The intermediate frequency signal is amplified by the low pass filter and then transmitted to the processor, and the processor processes the received signal, generally, performs fast fourier transform, spectrum analysis and the like on the received signal, so as to obtain information such as the distance, speed, angle and the like of the target object relative to the radar system.

Generally, the target detected by the radar may be a car, a ship, an airplane, or a billboard, or the like. If the target volume detected by the radar is relatively large, the target is also called an extended target, such as a vehicle, a ship, an airplane or a billboard, the target has a plurality of scattering points, each scattering point is a signal source, and the number of the signal sources is the number of the scattering points. When the target detected by the radar is small and is approximately a point target, one target is a scattering point, and the number of the sources is 1.

In order to accurately locate the position of a radar-detected target, in addition to the distance of the target to the radar, it is necessary to know the angular information of the target in the radar coordinate system. Methods for estimating the angle of a target in a radar coordinate system include super-resolution algorithms such as a direction of arrival (DOA) method, multiple signal classification (MUSIC), and a rotation invariant parameter estimation technique (ESPRIT).

In the super-resolution algorithms such as MUSIC and ESPRIT, the number of sources is often needed as a basis, for example, in the super-resolution algorithms based on MUSIC and ESPRIT, the noise subspace V is considerednBy selecting, requiring as input, the number of sources, e.g. Vn=V(:,Ns+1: end), wherein NsFor the number of the information sources, the estimation error of the number of the information sources greatly influences the precision of the super-resolution algorithm of MUSIC and ESPRIT, thereby greatly influencing the precision of the radar system for positioning the detected target position.

Disclosure of Invention

In order to solve the above problem, embodiments of the present application provide a method, an apparatus, and a storage medium for source number estimation.

In a first aspect, an embodiment of the present application provides a method for estimating a number of signal sources, which is applied to a receiving system of a millimeter wave radar, and the method includes: acquiring an array response of a radar antenna array, wherein the radar antenna array comprises at least one antenna unit; performing autocorrelation calculation on the array response to obtain a correlation matrix; performing eigenvalue decomposition on the correlation matrix, and extracting eigenvectors; setting the number of pre-estimated information sources, and classifying the number of the pre-estimated information sources to obtain at least one classification item; inputting the feature vector into a support vector machine corresponding to the at least one classification item, and outputting an evaluation value corresponding to the at least one classification item; comparing the evaluation values of the at least one classification item to obtain the largest evaluation value; and determining the number of the information sources according to the classification item corresponding to the maximum evaluation value.

The method for estimating the number of the information sources provided by the embodiment of the application is used for estimating the number of the information sources, the result has higher confidence coefficient, and the problem that the deviation between the performance of a low signal-to-noise ratio interval and the performance of a high signal-to-noise ratio interval is larger in the traditional method for estimating the number of the information sources can be solved.

In an embodiment, the method further includes a step of training a support vector machine corresponding to the at least one classification item, specifically including: grouping training vectors in a training set; the training vector is a characteristic vector with information source quantity labels; selecting training vectors of the corresponding group of the at least one classification item according to the grouping; taking the training vector of the corresponding group as positive; forming a truth set by taking the training vectors outside the corresponding groups in the training set as negative values; inputting the training set and the truth set into a support vector machine, training the support vector, the parameter vector and the variable of the support vector machine, and obtaining the trained support vector machine corresponding to the at least one classification item.

This embodiment has good performance with large scale data training, both at high and low signal-to-noise ratios. When the support vector machine adopts the linear kernel function, the calculation complexity of the method is very low.

In one embodiment, the acquiring an array response of the radar antenna array includes: acquiring an echo signal received by a radar antenna linear array; the radar antenna linear array comprises the M radar antenna units which are arranged at equal intervals; m is any natural number; the echo signals are reflection signals of targets with the same speed in the same detection distance; and forming a row/column vector by the array response of the linear array.

The method for estimating the number of the signal sources provided by the embodiment can estimate the number of the signal sources according to the array response of the antenna linear array on the same distance and speed unit, and can also estimate the number of the signal sources of other dimensions, such as the number of the signal sources of the speed dimension in the same distance and azimuth direction, or the total number of the signal sources of the speed dimension and the angle dimension on the same distance unit, and the like.

In one embodiment, the acquiring an array response of the radar antenna array includes: acquiring an echo signal received by a radar antenna area array, wherein the radar antenna area array comprises W rows and M rows of radar antenna units which are arranged at equal intervals; the echo signals received by the radar antenna area array comprise echo signals received by the W-row and M-column radar antenna units; the echo signals are reflection signals of targets with the same speed in the same detection distance, and W, M is any natural number; splicing the W line vectors of the echo signals into 1 linearly arranged line vector, wherein each line vector in the W line vectors comprises M echo signals; and obtaining the array response of the radar antenna area array.

The method for estimating the number of the information sources splices the area arrays into the linear arrays for calculation, can estimate the number of the information sources according to the array response of the antenna linear arrays on the same distance and speed unit, and simplifies an algorithm model for estimating the number of the information sources in a multi-dimensional mode.

In one embodiment, the method includes acquiring an array response of a radar antenna array, including: acquiring an echo signal received by a radar antenna circular array, wherein the radar antenna circular array comprises M radar antenna units which are uniformly arranged along the circumference, and M is any natural number; the echo signals received by the radar antenna circular array comprise M echo signals received by the radar antennas which are uniformly arranged along the circumference; and stretching the M echo signals uniformly arranged along the circumference into linearly arranged row vectors at equal intervals to obtain the array response of the radar antenna circular array.

The method for estimating the number of the information sources converts the circular arrays into the linear arrays for calculation, can estimate the number of the information sources according to the array response of the antenna linear arrays on the same distance and speed unit, and simplifies an algorithm model for estimating the number of the information sources in a multi-dimensional mode.

In one embodiment, the determining the evaluation value of each classification item corresponding to the pre-estimated source number includes: and multiplying the support vector in the support vector machine by the kernel function, and summing the multiplication result and the coefficient to obtain the evaluation value of each classification item of the information source quantity to be pre-estimated.

Further, the kernel function is a linear kernel function.

Further, the kernel function is a polynomial kernel function.

Further, the kernel function is a gaussian kernel function.

Further, the kernel function is a sigmiod kernel function.

The method for estimating the number of the information sources provided by the embodiment has higher confidence coefficient of the estimation result of the number of the information sources through the support vector machine, and can solve the problem that the traditional method for estimating the number of the information sources has larger deviation between the performance of a low signal-to-noise ratio interval and the performance of a high signal-to-noise ratio interval. When the support vector machine adopts the linear kernel function, the calculation complexity of the method is very low.

In one embodiment, the performing autocorrelation calculations on the array responses to obtain a correlation matrix includes: and performing J times of snapshots on the array response, multiplying the array response at the jth snapshot time by a conjugate transpose vector of the jth snapshot time to obtain an autocorrelation matrix at the jth snapshot time, adding the autocorrelation matrices of the J snapshots, and averaging to obtain a correlation matrix, wherein J and J are both natural numbers.

The method for estimating the number of the information sources provided by the embodiment has the advantages that the dimensionality of the correlation matrix calculated by taking N times of pictures is large, and the accuracy of the extracted feature vector is high. The method is suitable for calculating the correlation matrix according to the array response of the radar linear array, the area array and the circular array.

In one embodiment, the performing autocorrelation calculations on the array responses to obtain a correlation matrix includes: and carrying out single snapshot on the array response, grouping the array response at the snapshot time to obtain at least one group of row vectors, multiplying the at least one group of row vectors by a conjugate transpose vector of the at least one group of row vectors to obtain an autocorrelation matrix of the at least one group of row vectors, and adding the autocorrelation matrices of the at least one group of row vectors and then averaging to obtain a correlation matrix.

The method for estimating the number of the information sources provided by the embodiment adopts single snapshot and smooth grouping to obtain the spatial correlation matrix, only one sampling is needed, the obtained correlation matrix R has low dimensionality, the calculated amount is reduced, and the spatial correlation matrix R is evaluated in a smooth grouping mode, so that the orientation can be accurately realized. The method is suitable for calculating the correlation matrix R according to the array response y of the radar linear array and the area array, but is not suitable for the radar circular array.

In one embodiment, the method further includes mapping the feature vector by a functional process to obtain a second feature vector.

The method for estimating the number of the information sources reduces the difference between the maximum value and the minimum value, so that the value range of the characteristic function is concentrated, the discreteness is small, and the calculation amount of a subsequent Support Vector Machine (SVM) is reduced.

In one embodiment, the mapping the feature vector by a function process to obtain the second feature vector, with the feature vector being a first feature vector, includes: calculating the value of a logarithmic function by taking the characteristic value in the first characteristic vector as an independent variable of the logarithmic function; and forming the second feature vector by taking the value of the logarithmic function as an element.

The method for estimating the number of the information sources provided by the embodiment can convert the characteristic parameters from a linear domain to a log domain by adopting the logarithmic function, reduce the difference between the maximum value and the minimum value, centralize the value domain of the characteristic function, have small discreteness and reduce the calculation amount of a subsequent Support Vector Machine (SVM).

In one embodiment, the mapping the feature vector by a function process to obtain the second feature vector, with the feature vector being a first feature vector, includes: comparing the characteristic values in the first characteristic vector to obtain the maximum value; calculating a ratio of the characteristic value to the maximum value; and calculating the value of the logarithmic function by taking the ratio as an independent variable of the logarithmic function, and forming the second feature vector by taking the value of the logarithmic function as an element.

The method for estimating the number of the information sources provided by the embodiment has the advantages that through the normalization processing and the logarithmic function mapping of the feature vectors, the distribution domains of the feature parameters are concentrated, the discreteness is small, and the calculation amount of a subsequent Support Vector Machine (SVM) is reduced.

In one embodiment, the mapping the feature vector by a function process to obtain the second feature vector, with the feature vector being a first feature vector, includes: sorting the eigenvalues in the first eigenvector; and calculating the value of the logarithmic function by taking the sorted characteristic values as the independent variables of the logarithmic function, and forming the second characteristic vector by taking the value of the logarithmic function as an element.

After the method for estimating the number of the information sources is sequenced, the characteristic function is adopted to convert the characteristic parameters from a linear domain to a log domain, and the difference between the maximum value and the minimum value can be further reduced through normalization processing of the characteristic vector, so that the distribution domain of the characteristic function is concentrated, the discreteness is small, and the calculation is facilitated.

In a second aspect, an embodiment of the present application provides an apparatus for source number estimation, where the apparatus includes: the data acquisition module is used for acquiring array response of a radar antenna array, and the radar antenna array comprises at least one antenna unit; the correlation matrix calculation module is used for carrying out autocorrelation calculation on the array response to obtain a correlation matrix; the characteristic extraction module is used for decomposing the characteristic value of the correlation matrix and extracting a characteristic vector; the evaluation value calculation module is used for setting the number of pre-estimated information sources, classifying the number of the pre-estimated information sources and obtaining at least one classification item; and a source quantity determining module, which is used for inputting the feature vector into a support vector machine corresponding to the at least one classification item and outputting an evaluation value corresponding to the at least one classification item; comparing the evaluation values of the at least one classification item to obtain the largest evaluation value; and determining the number of the information sources according to the classification item corresponding to the maximum evaluation value.

In one embodiment, the apparatus further comprises a training support vector machine module for grouping training vectors in a training set according to different source numbers; the training vectors are characteristic vectors with information source quantity labels, the training vectors of the corresponding groups of each classification item are selected according to the grouping, and the training vectors of the corresponding groups are taken as positive; and forming a truth set by taking training vectors in the training set except for the training vector corresponding to the corresponding group as negative, inputting the training set and the truth set into the support vector machine corresponding to each classification item, training the support vector, the parameter vector and the variable of the support vector machine, and obtaining the trained support vector machine corresponding to each classification item.

In a third aspect, an embodiment of the present application provides an electronic device, including a memory and a processor; the processor is configured to execute the computer executable instructions stored in the memory, and the processor executes the computer executable instructions to perform the method for estimating the number of signal sources according to any one of the above embodiments.

In a fourth aspect, a storage medium of this embodiment includes a readable storage medium and a computer program stored in the readable storage medium, where the computer program is used to implement the method for source number estimation described in any one of the above embodiments.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments disclosed in the present specification, the drawings needed to be used in the description of the embodiments will be briefly introduced below, it is obvious that the drawings in the following description are only the embodiments disclosed in the present specification, and it is also possible for those skilled in the art to obtain other drawings based on the drawings without creative efforts.

FIG. 1 is a diagram of the operation of a millimeter-wave radar in the background art;

FIG. 2a is a schematic coordinate diagram of a linear array millimeter wave radar;

FIG. 2b is a schematic diagram showing an angle between a connecting line between a target and a radar in a linear array and a normal line;

FIG. 3 is a schematic diagram of an application scenario of a radar receiving system;

fig. 4 is a flowchart of a method for source number estimation according to an embodiment of the present application;

FIG. 5 is a schematic diagram of a radar array receiving;

fig. 6 is a flowchart illustrating that a correlation matrix R is subjected to eigenvalue decomposition to extract eigenvectors of the correlation matrix R in the method for estimating the number of information sources according to the embodiment of the present application;

fig. 7 is a flowchart of inputting feature vectors into a support vector machine to determine the number of signal sources in a method for estimating the number of signal sources according to an embodiment of the present application;

FIG. 8 is a diagram illustrating source number estimation using a trained support vector machine according to an embodiment of the present disclosure;

FIG. 9a is a confidence curve graph of source number estimation using the method for source number estimation provided by the embodiment of the present application;

FIG. 9b is a confidence plot of source number estimation using the modified MDL;

fig. 10 is a functional block diagram of an apparatus for source number estimation according to an embodiment of the present application;

fig. 11 is a schematic diagram of an electronic device for source number estimation according to an embodiment of the present application.

Detailed Description

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

In the following description, references to the terms "first \ second \ third, etc. or module a, module B, module C, etc. are used solely to distinguish between similar objects and do not denote a particular order or importance to the objects, but rather the specific order or sequence may be interchanged as appropriate to enable embodiments of the application described herein to be practiced in an order other than that shown or described herein.

In the following description, reference to reference numerals indicating steps, such as S110, S120 … …, etc., does not necessarily indicate that the steps are performed in this order, and the order of the preceding and following steps may be interchanged or performed simultaneously, where permissible.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of the present application only and is not intended to be limiting of the application.

The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

In order to accurately locate the position of the probe target, in addition to the distance r of the target to the radar array, the angle θ of the target in the radar coordinate system needs to be known. Let r be the distance of the target from the origin of the radar coordinate system. As shown in fig. 2a, for a system in which a linear millimeter wave radar is arranged, the origin of the radar coordinate system is the center position of the linear array. The angle theta is the included angle between the line between the target and any antenna and the linear array normal. As shown in fig. 2b, the position of the target in the radar coordinate system can be determined by the distance r and the angle θ.

In one scheme (referred to as scheme 1), the radar may estimate an angle θ of the target in a radar coordinate system according to the received echo signal. If the receiving antenna of the radar is located in the far field of the signal source, the received echo signal can be assumed to be a plane wave, and the detection target is regarded as a point target. In particular, assume that each antenna element is spaced apart by a distance daThen, the path difference Δ d of the echo signal satisfying the far-field condition reaching the two adjacent antennas is:

Δd=dasin(θ) (1)

phase difference caused by the path differenceComprises the following steps:

wherein λ is the wavelength of radar millimeter waves.

For a linear array containing M antennas, the array response y, which consists of the received echo signals, can be expressed as:

if K targets exist in the space, the angle between each target and the radar is theta1,θ2,…,θkWherein thetakIs the value of the angle of the kth target of the K targets to the antenna. Then the array response y of the received signals of the M receive antennas is:

wherein the content of the first and second substances,the complex gain on each antenna for the echo of the kth target. The array response y is the array response of echo signals of K targets compounded on M receiving antennas.

According to the array response y, a machine learning method is adopted for target angle estimation, and the method which can be adopted comprises a method of scanning in the direction of incoming waves (DOA), a method of multiple signal classification (MUSIC), a rotation invariant parameter estimation technology (ESPRIT) and the like. Taking the direction of arrival (DOA) method as an example, the radar needs to construct a plurality of steering vectors according to the detected angle range. For example, the detected angle range is-40 degrees to 40 degrees, and the guide vector is constructed according to the angle interval of 1 degree, so that the guide vectorComprises the following steps:

wherein, thetaiThe ith angle in the range of angles is detected.

Each detection angleΘiCorresponding guide vectorCorrelated with the array response y, the correlation value g (i) for the ith angle is:

the correlation value g (i) represents the echo signal at the detection angle thetaiGain in the direction. A steering vector corresponding to each detection angle obtained by the formula (5)And the y-point multiplication of the array response obtained by the formula (4) and summation, when the ith detection angle is the same as the angle of the echo signal of the kth target, i.e., i ═ k, according to the formula (6), in-phase combination is performed,the dot product is 1, at which time the correlation g (i) is at a maximum and the array response y is at the detection angle ΘiThe gain in the direction is maximum. Under the condition that the target number or the source number K is known, each detection angle theta corresponding to K maximum values in the correlation values g (i)iI.e. the angles of the K sources.

The principle of the incoming wave direction scanning (DOA) method is that a guide vector corresponding to each detection angle is correlated with a received signal, and the detection angle corresponding to the scanned maximum correlation value is the angle of a target in a radar coordinate system. However, in a target with the same speed at a distance, a maximum correlation value obtained by a direction of arrival scanning (DOA) method cannot accurately distinguish the orientation of the target for the targets with the two angles close to each other. For example, in the automatic driving technology, when two vehicles are running side by side in front of each other, if one vehicle has an angle of 5 ° in a radar coordinate system and the other vehicle has an angle of 5.5 ° in the radar coordinate system, the vehicle or the large vehicle may be mistakenly recognized as running in front of each other by using a common detection algorithm such as the above-mentioned direction of arrival (DOA) method, which may result in a reduction in safety of automatic driving.

In order to improve the safety performance of automatic driving, in radar detection, a signal processing system of the radar can adopt certain special algorithms, such as an angle super-resolution algorithm, to detect the positions of two vehicles on the basis of accurate information source quantity estimation. Accurate source number estimation is often the basis for target position detection.

In another scheme (denoted scheme 2), information theoretic criterion (AIC) or Minimum Description Length (MDL) algorithm is used for source number estimation. In these methods a correlation matrix of the array response y is required.

The correlation matrix is also called a correlation coefficient matrix and is formed by correlation coefficients among columns of the matrix, and elements of the ith row and the jth column of the correlation matrix are the correlation coefficients of the ith row and the jth column of the original matrix. The correlation coefficient is a statistical index reflecting the degree of closeness of correlation between variables.

For example, as in the implementation scenario of scheme 1, the receiving antenna of the radar is located in the far field of the signal source, the received echo signal is a plane wave, and the detection target is regarded as a point target and is denoted as the signal source. For a linear array containing M antennas, the array response, which consists of the received echo signals, is y. And taking N times of snapshots on the array response y, wherein the snapshots are obtained by sampling the array response at different moments. Obtaining the receiving signals of M receiving antennas according to the formula (4) to obtain the array response y, and then obtaining the array response y sampled at the jth snapshot timej

Wherein the content of the first and second substances,and the echo signal received by the ith antenna at the jth snapshot time is received. Array response yjAnd forming a vector for echo signals received by the M antennas at the jth snapshot time. The jth blockArray response at beat time yjWith its conjugate transpose (y)j)HMultiplying to obtain an M multiplied by M autocorrelation matrix at the jth snapshot time, adding the autocorrelation matrices of the N snapshots, and then averaging to obtain a correlation matrix R:

in the formula (8), R is a correlation matrix of M × M, N is the number of snapshots, (y)j)HArray response y for jth snapshot timejThe conjugate of (2) transposes the vector.

Then, performing eigenvalue decomposition on the correlation matrix R:

R=U∑UH (9)

wherein U is a decomposition matrix of the correlation matrix R, UHA conjugate transpose matrix that is a decomposition matrix U; the diagonal matrix Σ is:

the dimension of the diagonal matrix Σ is mxm, the diagonal element liIs the ith eigenvalue of the correlation matrix R; which reflects the eigenvalues of the number of echo signals that are ideally noise-free.

Extracting diagonal elements l of a diagonal matrix ΣiArranging the extracted diagonal elements in descending order to obtain a feature vector l ═ l { (l)1,…,li,…,lM'}: wherein l1≥l2≥…≥lM'

From the eigenvectors l of the correlation matrix R, the estimated number of sources Ns can be calculated with an information theoretic criterion (AIC) algorithm as:

wherein N iss0,1,2, …, M-1. The number of the sourceComprises the following steps:

number of sourcesTo make AIC (N)s) Variable N at minimumsThe value of (a).

In this scenario 2, the performance confidence of the source number estimation algorithm depends on the choice of the number of snapshots N. When the value of N is larger, the algorithm has better performance in a high signal-to-noise ratio interval and higher confidence coefficient; but the low signal-to-noise ratio interval has poor performance and low confidence. When the number N of times of snapshot is smaller, the algorithm has better performance in a low signal-to-noise ratio interval and higher confidence coefficient; but there is estimation bias in the high signal-to-noise ratio interval, and the confidence is lower. I.e. the estimation accuracy cannot approach 1.

Since the scheme 1 cannot accurately distinguish a plurality of targets with very close angles, and the positioning accuracy is not ideal because the performance of the low signal-to-noise ratio interval and the high signal-to-noise ratio interval has deviation when the scheme 2 estimates the number of the information sources, the method adopts a multi-classification Support Vector Machine (SVM) to estimate the number of the information sources, and introduces the concept of the method.

The embodiment of the application provides a method for estimating the number of information sources, which processes array response of an antenna array to obtain a correlation matrix, obtains a feature vector for detecting the number estimation of the information sources according to the correlation matrix, and inputs the feature vector into a Support Vector Machine (SVM), so that the number of a plurality of information sources on a distance and speed unit is reasonably estimated.

The method for estimating the number of the information sources is particularly applied to an angle spectrum estimation module in radar system signal processing, the number of the information sources is obtained according to a Support Vector Machine (SVM), and a specific target direction can be obtained by adopting some special algorithms, such as an angle super-resolution algorithm, in subsequent angle spectrum estimation. In the subsequent angle spectrum estimation, how to calculate the angle and the distance according to the number of the information sources is not explained in the present application.

The principle of Support Vector Machines (SVM) is described below.

Support Vector Machines (SVMs) are a class of classifiers that classify data in a supervised learning manner. Let the input vector of the classifier be X ═ X1,…,Xi,…,XNThe corresponding classification is z ═ z }1,…,zi,…,zNWherein the ith input vector is Xi=[x1,x2,…,xn]. In the binary problem, zi∈{-1,1},ziA value of-1 indicates a negative class, ziA value of 1 indicates a positive class.

In the classification problem of SVM, the parameters ω and b need to be found such that:

ziTXi+b)≥1 (12)

where ω is the normal vector of the hyperplane, ωTA transposed vector which is a hyperplane normal vector omega; b is the intercept.

In some circumstances, XiIs linearly inseparable, and at this time X is often required to be introducediMapping into a high dimensional space, when:

ziTφ(Xi)+b)≥1 (13)

wherein, ω isTPhi (-) is the kernel function.

The SVM algorithm is originally designed for a binary classification problem, and when processing a plurality of classes of problems, a suitable plurality of classes of classifiers need to be trained, and common methods include one-to-many (OVR SVMs for short) and one-to-one (OVO SVMs or pairwise for short).

In the one-to-many method, samples of a certain class are classified into one class and other remaining samples are classified into another class in sequence during training, so that m SVM are constructed by the samples of m classes. The classification classifies the unknown sample as the class having the largest classification function value.

For example, assume that there are four categories to be classified (i.e., 4 labels), A, B, C and D respectively. When the training sets are extracted, the following training sets are respectively extracted:

(1) the vector corresponding to A is used as a positive set, and the vectors corresponding to B, C and D are used as a negative set;

(2) the vector corresponding to B is used as a positive set, and the vectors corresponding to A, C and D are used as a negative set;

(3) the vector corresponding to C is used as a positive set, and the vectors corresponding to A, B and D are used as a negative set;

(4) the vector corresponding to D is used as a positive set, and the vectors corresponding to A, B and C are used as a negative set;

and respectively training by using the four training sets, and then obtaining four training result files as four trained support vector machines.

During testing, the test vectors are respectively tested by the four trained support vector machines, and each support vector machine outputs an evaluation value: f1(x), f2(x), f3(x) and f4 (x). The evaluation values are used to evaluate the similarity of the input test vectors to the corresponding training set, and the final classification result is the category corresponding to the largest one of the four evaluation values. Namely Max (f1(x), f2(x), f3(x) and f4 (x)).

A method and an apparatus for source number estimation provided by the embodiments of the present application are specifically described below with reference to fig. 3 to fig. 11.

Fig. 3 is a schematic view of an application scenario of the millimeter wave radar receiving system. As shown in fig. 3, in the millimeter wave radar receiving system, the radio frequency front end module receives the echo signal Y, the echo signal Y enters the signal processing unit to be processed to obtain the estimated number of the signal sources, and then the signal processing unit processes the signal to obtain the angle and the distance, and finally the specific signal source position is obtained. The signal processing unit comprises distance spectrum estimation, velocity spectrum estimation, angle spectrum estimation and other modules, and the method and the device for estimating the number of the information sources are mainly applied to the angle spectrum estimation module.

To reasonably estimate the number of the information sources, firstly, the array response Y of an echo signal Y on a resolution target with the same distance and the same speed on a radar antenna array needs to be obtained; then after a correlation matrix R of the array response y is calculated, carrying out eigenvalue decomposition on the correlation matrix R so as to extract the eigenvector of the correlation matrix R; and finally, estimating the number of the information sources by a Support Vector Machine (SVM) by using the extracted feature vectors.

Fig. 4 is a flowchart of a method for source number estimation according to an embodiment of the present application. The implementation subject of the method may be equipment, a server or an electronic device with computing processing capability, as in the implementation scenario of scheme 1, a receiving antenna of a radar array is located in a far field of a signal source, a received echo signal is a plane wave, and a detection target is regarded as a point target and is marked as a signal source. The present application is specifically set forth below for each step.

As shown in fig. 4, step S401 is first executed to obtain an array response of the radar antenna array.

Specifically, the radar array is a linear array formed by M equally spaced antenna units, the antenna index is 1, …, M, and the normal direction of the antenna units of the radar is 0 degree. A schematic diagram of the radar array receiving the echo signal can be seen with reference to fig. 5. Suppose that there are N point targets with the same speed in the same distance area in front of the radar array, and the point targets are marked as source 1, source 2, …, source N, and the azimuth angle of source 1 is theta1Azimuth angle of source 2 is θ2…, azimuth angle θ of source Nn. The array response y received by a radar array with M antenna elements is:

wherein, y1,y2,…,yMReceived echo signals, y, of M antenna elements, respectivelyMThe received signal of the Mth antenna is M ≥ 1; the array response y is a row/column vector consisting of echo signals received by the M antenna units; snThe complex signal response coefficient of the information source n is related to the material of the information source; n is a radical ofnThe noise vector is determined by the thermal noise caused by the device;A(θn) For steering vectors, the azimuth angle is theta over M antenna elementsnThe steering vector A (θ) of the echo signal of (2)n) Comprises the following steps:

when the radar array is an area array, if the area array is arranged as W rows and M columns with equal spacing, W, M is any natural number, the radar area array comprises W × M antenna elements, and the indexes of the antenna elements are 1,2, … and W × M. Equivalent to the splicing of W line vectors, each line vector consists of echo signals of M antenna units, the array response y of each linear array unit is obtained by a formula (14), and the radar area array response is obtained by splicing the W array responses yWherein y isW×MIs the received signal of the W × M antenna element.

When the radar array is arranged as a circular array, if the circular array comprises M antenna units uniformly arranged along the circumference, where M is any natural number, the M antenna units uniformly arranged along the circumference may be stretched into a linear arrangement with equal spacing, and the array response y received by the radar circular array of the M antenna units is obtained from formula (14)

After the array response y is obtained, step S402 is executed to calculate the array response y, and obtain the correlation matrix R.

In practice, considering that the received signal is of finite length, the correlation matrix R can be calculated from the N snapshots of the array response y using a time-space estimation method. Specifically, J times of snapshot sampling are performed on the array response y obtained by the formula (14), and the array response y at the jth snapshot timejWith its conjugate transpose (y)j)HMultiplying to obtain M × M autocorrelation matrix at the jth snapshot time, adding the autocorrelation matrices of the J snapshots, averaging, and referring toEquation (8) yields a correlation matrix R as:

wherein, yjFor the array response at the jth snapshot time, for the echo signal received by the ith antenna at the jth snapshot time,for the response of the Mth antenna in the jth snapshot, (y)j)HArray response y for jth snapshot timejThe conjugate of (2) transposes the vector.

In the step S402, the correlation matrix R calculated by taking N photographs has a large dimension, and although the accuracy of the extracted eigenvector is high, the array response y needs to be sampled N times, which increases the workload and data amount of the sampling operation, and the calculation is complicated.

As an alternative embodiment, a single snapshot may be performed on the array response y acquired in step S401, and the spatial correlation matrix R is acquired in a smooth grouping manner. Specifically, in step S402, one snapshot sampling may be performed on the array response y; the specific method is that the received signals of the array response y are divided into groups smoothly, the array response of every M 'antennas is a group, then the M array responses are divided into M-M' +1 groups of vectors, and the acquired array responses areConversion to vector y':

for example, in a linear array of 16 antennas with a radar front, 8 antennas are usedThe antennas are grouped, and the array response to be collected isThe smoothing is divided into y' consisting of 9 row vectors:

the correlation matrix R of the array response y is then:

where s is the index value, M 'is the number of antennas per group, M' < M.

The correlation matrix R of the array response y is calculated by adopting the method, only one sampling is needed, the obtained correlation matrix R has low dimensionality, the calculated amount is reduced, and the spatial correlation matrix R is evaluated by a smooth grouping mode, so that the orientation can be accurately realized. This approach is applicable to correlation matrices R based on the array response y of the radar linear and area arrays, but not to radar circular arrays.

After obtaining the correlation matrix R, step S403 is performed to perform eigenvalue decomposition on the correlation matrix R, and extract the eigenvector of the correlation matrix R. The specific flow is as shown in fig. 6, and can be realized by executing the following steps S4031-4035.

S4031, the characteristic value decomposition is carried out on the correlation matrix R, and the formula (9) is shown:

R=U∑UH

where U is the decomposition matrix of the correlation R, UHA conjugate transpose matrix for U; Σ is a square matrix of eigenvalues of the ideal noiseless array response, as in equation (10):

diagonal element l in eigenvalue square matrix sigma1 … li … lM'To a characteristic value reflecting the number of ideal noise-free received signals.

S4032, diagonal elements of the feature matrix Σ are extracted to obtain a feature vector l ═ l of the correlation matrix R1 … li… lM']TWherein M' ═ M.

To reduce the amount of computation of a Support Vector Machine (SVM), the feature vector l may be given as l ═ l1 … li … lM']TThe value domain mapping is performed by the function processing, and steps S4033 to S4035 are performed.

S4033, using the feature vector l ═ l1 … li … lM']TFor the first feature vector, each element of the first feature vector is mapped through function processing, and a feature function x ═ f (l) is obtained, where f () represents a feature function, and the feature function may be a logarithmic function.

S4034, calculating the characteristic function x ═ f (l), and obtaining each characteristic value liCorresponding characteristic function value xi

In one possible example, each element of the first feature vector may be mapped by a logarithmic function process with each feature value/iCalculating the value x of a logarithmic function for its argumentiThereby obtaining a second eigenvector x ═ x1 … xn]T. Specifically, the logarithmic feature function is xi=C*log(li) Or xi=C*log10(li) Wherein C is a constant.

The logarithmic function can be used for converting the characteristic parameters from a linear domain to a log domain, reducing the difference between the maximum value and the minimum value, leading the value domain of the characteristic function to be concentrated, having small discreteness and reducing the calculation amount of a subsequent Support Vector Machine (SVM).

In another possible example, each feature value in the feature vector may be compared to obtain a maximum feature value, then a ratio of each feature value to the maximum feature value is calculated, the ratio is taken as an argument of a logarithmic function, a value of the logarithmic function is calculated, and a feature parameter value x corresponding to each feature value is obtainedi

In particular, a feature letterNumber xi=C*log(li(l) max or xi=C*log10(li(l) where C is a constant.

Through the normalization processing and the logarithmic function mapping of the feature vectors, the distribution domains of the feature parameters are concentrated, the discreteness is small, and the calculation amount of a subsequent Support Vector Machine (SVM) is reduced.

In a third possible example, each eigenvalue in the eigenvector may be sorted, each eigenvalue in the sorted eigenvector is taken as an argument of a logarithmic function, a value of the logarithmic function is calculated, and an eigenvalue value x corresponding to each eigenvalue is obtainedi

Specifically, for the feature vector liSorting from large to small to obtain a feature vector l' ═ sort (l)1 … ln) Then the ith characteristic parameter can be xi=C*log(li') to a host; or, xi=C*log10(li') or xi=C*log(l'i(l) max or (x)i=C*log10(l'i/max(l))。

After sorting, the characteristic function is adopted to convert the characteristic parameters from a linear domain to a log domain, and the difference between the maximum value and the minimum value can be further reduced through normalization processing of the characteristic vector, so that the distribution domain of the characteristic function is concentrated, the discreteness is small, and the calculation is facilitated.

S4035, the value x of each characteristic function is usediComposing a feature vector x ═ x for the element1 … xi … xn]T. Wherein xiIs the ith characteristic value liAnd corresponding characteristic parameter values.

After the feature vectors are obtained, step S404 is executed, the feature vectors are input into the trained support vector machine corresponding to each classification item, the evaluation value of each classification item corresponding to the number of the information sources is output, and the number of the information sources corresponding to the classification item with the largest evaluation value is used as the estimated number of the information sources.

The feature vector may be the feature vector l ═ l of the correlation matrix R extracted in step S40321… li…lM']TPreference is given toThe feature vector x ═ x obtained after calculation by the logarithmic function mapping in step S40351 … xi … xn]T

Finally, the feature vectors are input into a Support Vector Machine (SVM) to determine the number of the information sources. Specifically, as shown in FIG. 7, the method includes the following steps S4041-S4043.

S4041, pre-estimating the number of the information sources, and classifying the estimated number of the information sources to obtain classification items with the number corresponding to the number of the information sources.

Specifically, the number of pre-estimated sources is NsWhen estimating the number of sources N in advancesWhen the number is 4, the possible situations of the number of the information sources are classified within 4, which may be 0,1,2,3 or 4, and the obtained classification items s are respectively 0,1,2,3 and 4. Each classification item corresponds to a support vector machine.

S4042, inputting the feature vectors into the trained support vector machine corresponding to each classification item, and outputting the estimated value phi of each classification item01,…,φn-1

The trained support vector machine corresponding to each classification item can evaluate the value phi01,…,φmA corresponding algorithm. For example, the evaluation value phi of the output of the support vector machine corresponding to the classification item with the number m of the information sourcesmThe algorithm is as follows:

φm=Hm×K(Dm,x)+bm (18)

wherein HmThe dimension of the parameter vector trained by the support vector machine is 1 xL. K (D)mX) is a kernel function in the support vector machine, the dimension is L multiplied by 1, and L is the number of support vectors in the support vector machine; dmThe support vector is a support vector trained by a support vector machine; x is the input feature vector. bmThe dimensionality of the numerical variable trained by the support vector machine is 1 multiplied by 1.

Then the support vector machine corresponding to the classification item with the source number m can be expressed as: hm×K(Dm,x)+bm

FIG. 8 is a diagram illustrating source number estimation using a trained support vector machine. As shown in fig. 8, the feature vectors are respectively input into the trained support vector machine corresponding to each classification item, and the evaluation values phi of the corresponding information source number are respectively output01,…,φn-1

The kernel function of the support vector machine may be one of a linear kernel function, a polynomial kernel function, a gaussian kernel function, or a sigmiod kernel function.

Wherein the linear kernel function is:the function value of the linear kernel function takes a characteristic vector x as an independent variable and supports a vector DmProduct with an argument.

The polynomial kernel function is:the function value of the linear kernel function takes a second characteristic vector x as an independent variable and supports a vector DmThe product of the argument is summed to the x power of 1.

The Gaussian kernel function is:where γ is the configuration parameter and diag (.) is the operation to take the diagonal.

The sigmiod kernel function is: where β and θ are configuration parameters and s is a variable.

It is to be noted in particular that: estimation of source values for linear kernel functionsmCan be further expressed asPmIs a linear support vector. In the concrete examplesIn the application process, Dm、HmAnd bmAre parameters trained by a support vector machine. The subscript m is 0,1,2, …, n-1.

Finally, step S4043 is performed to determine the value of the source number according to the estimated value of each classification item.

Specifically, the estimated value φ of each classification term is compared01,…,φn-1Obtaining the maximum estimated value, and the value N of the information source numbersIs phi01,…,φn-1Subscript corresponding to the medium maximum, i.e.

Ns=argmaxm01,…,φn-1} (19)。

The method for estimating the number of the information sources provided by the embodiment of the application further comprises the step of training the support vector machine corresponding to each classification item. In the embodiment of the application, the estimation of the number of the information sources belongs to a multi-classification problem, so that the multi-classification problem is realized in a one-to-many mode in the training of the support vector machine.

Specifically, assuming that there are Q sets of training vectors, the training set is ═ d(1),d(2),…,d(Q)In which d is(q)A training vector of M' × 1. The number of the information sources corresponding to the Q groups of training vectors is y(1),y(2),…,y(Q)The source number set is Y ═ Y(1),y(2),…,y(Q)In which y is(q)∈{0,1,…,M'-1}。

In the embodiment of the present application, a parameter D with m number of information sources is to be trainedm、HmAnd bmThe number of the information sources can be m, y(q)Selecting the vector m and corresponding truth value r(q)Set 1, will y(q)True value r corresponding to training vector other than m(q)Set-1, i.e.:

setting the kernel function dimension of each support vector machineIs L multiplied by 1, L is the number of the support vectors in the support vector machine; training set ═ d(1),d(2),…,d(Q)R and the truth set R ═ R(1),r(2),…,r(Q)Sending into a support vector machine for training so that r(q)(Hq×K(Dq,d(q))+bq)≥1。

Obtaining L support vectors through trainingWhereinThe support vector matrix isThe parameter vector matrix isWhereinIs composed ofA corresponding true value is set for the value of,a0,a1,…aLas a variable of training, bqAre trained parameters.

As an alternative embodiment, the input vector used for training or estimation may beOr y, or a function f (y) d ═ y or d ═ f (y), in particular,the rest of the process is similar to the above embodiment.

In a simulation environment, the confidence degree comparison of the information source quantity estimation is carried out by respectively adopting the information source quantity estimation method and the improved MDL method provided by the application. Under the same condition, the radar array is a linear array with 16 antennas, the wavelength is taken as the antenna spacing, and the number of the information sources is respectively as follows: 1. 2 and 3, of which 2 sources are spaced at 3 angular intervals and 3 sources are spaced at 7 angular intervals.

FIG. 9a is a graph of confidence of source number estimation using the method of source number estimation provided herein. As shown in fig. 9a, the abscissa of the graph is signal-to-noise ratio (SNR), the ordinate is accuracy of the estimation result, and an estimation result confidence curve of one source, an estimation result confidence curve of two sources, and an estimation result confidence curve of three sources are respectively shown in the coordinate system. The reliability curve analysis can be carried out, under the environment that the signal-to-noise ratio is 10, the accuracy of the estimation result of one information source is 99%, the accuracy of the estimation result of two information sources is 95%, and the accuracy of the estimation result of three information sources is 45%. The accuracy of the estimation results of the number of the information sources is increased along with the increase of the signal-to-noise ratio, under the environment that the signal-to-noise ratio is 20, the accuracy of the estimation results of one information source and two information sources is 100%, and the accuracy of the estimation results of three information sources is close to 90%.

Fig. 9b is a confidence plot of source number estimation using the improved MDL. As shown in fig. 9b, in an environment with a signal-to-noise ratio of 10, the accuracy of the estimation results of two information sources is 80%, and the accuracy of the estimation results of three information sources is only 28%; under the environment with the signal-to-noise ratio of 20, the accuracy of the estimation results of one and two information sources is 100%, and the accuracy of the estimation results of three information sources is close to 80%.

As can be obtained by comparing and analyzing fig. 9a and fig. 9b, the result of estimating the number of signal sources by using the method for estimating the number of signal sources provided by the present application has higher confidence, and the problem of larger deviation between the performance of the low signal-to-noise ratio interval and the performance of the high signal-to-noise ratio interval in the conventional method for estimating the number of signal sources can be solved.

The method for estimating the number of the information sources provided by the embodiment of the application has good performance under high signal-to-noise ratio or low signal-to-noise ratio through large-scale data training. When the support vector machine adopts the linear kernel function, the calculation complexity of the method is very low.

The method for estimating the number of the information sources provided by the embodiment of the application can estimate the number of the information sources according to the response of the antenna array on the same distance and speed unit, and can also estimate the number of the information sources of other dimensions, such as the number of the information sources of the speed dimension in the same distance and azimuth direction, or the total number of the information sources in the same distance unit, the speed dimension and the angle dimension, and the like.

The embodiment of the present application provides an apparatus for estimating the number of signal sources, as shown in fig. 10, the apparatus includes: a data acquisition module 1001 configured to acquire an array response of a radar antenna array, where the radar antenna array includes at least one antenna unit; a correlation matrix calculation module 1002, configured to perform autocorrelation calculation on the array response to obtain a correlation matrix; a feature extraction module 1003, configured to perform eigenvalue decomposition on the correlation matrix, and extract a feature vector; an evaluation value calculation module 1004, configured to set a pre-estimated number of information sources, classify the pre-estimated number of information sources, and obtain at least one classification item; and a source number determining module 1005, configured to input the feature vector into a support vector machine corresponding to the at least one classification item, and output an evaluation value corresponding to the at least one classification item; comparing the evaluation values of the at least one classification item to obtain the largest evaluation value; and determining the number of the information sources according to the classification item corresponding to the maximum evaluation value.

The device for estimating the number of the signal sources provided by the embodiment of the application further comprises a training support vector machine module 1006, wherein the training support vector machine module groups training vectors in a training set according to different numbers of the signal sources; the training vector is a characteristic vector with information source quantity labels; selecting the training vector of the corresponding group of each classification item according to the grouping, and taking the training vector of the corresponding group as positive; forming a truth set by taking the training vectors outside the corresponding groups in the training set as negative values; inputting the training set and the truth set into the support vector machine corresponding to each classification item, training the support vector, the parameter vector and the variable of the support vector machine, and obtaining the trained support vector machine corresponding to each classification item.

An embodiment of the present application provides an electronic device 1100, including a processor 1101 and a memory 1102; the processor 1101 is configured to execute the computer executable instructions stored in the memory 1102, and the processor 1101 executes the computer executable instructions to perform the method for estimating the source number according to any of the above embodiments.

The embodiment of the present application provides a storage medium 1103, which includes a readable storage medium and a computer program stored in the readable storage medium, where the computer program is used to implement the method for estimating the number of signal sources according to any one of the above embodiments.

Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the embodiments of the present application.

Moreover, various aspects or features of embodiments of the application may be implemented as a method, apparatus, or article of manufacture using standard programming and/or engineering techniques. The term "article of manufacture" as used herein is intended to encompass a computer program accessible from any computer-readable device, carrier, or media. For example, computer-readable media can include but are not limited to magnetic storage devices (e.g., hard disk, floppy disk, magnetic strips, etc.), optical disks (e.g., Compact Disk (CD), Digital Versatile Disk (DVD), etc.), smart cards, and flash memory devices (e.g., erasable programmable read-only memory (EPROM), card, stick, or key drive, etc.). In addition, various storage media described herein can represent one or more devices and/or other machine-readable media for storing information. The term "machine-readable medium" can include, without being limited to, wireless channels and various other media capable of storing, containing, and/or carrying instruction(s) and/or data. It should be understood that, in various embodiments of the present application, the sequence numbers of the above-mentioned processes do not imply an order of execution, and the order of execution of the processes should be determined by their functions and inherent logic, and should not limit the implementation processes of the embodiments of the present application.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the embodiments of the present application, which essentially or partly contribute to the prior art, may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or an access network device) to execute all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.

The above description is only a specific implementation of the embodiments of the present application, but the scope of the embodiments of the present application is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the embodiments of the present application, and all the changes or substitutions should be covered by the scope of the embodiments of the present application.

26页详细技术资料下载
上一篇:一种医用注射器针头装配设备
下一篇:测角方法以及雷达设备

网友询问留言

已有0条留言

还没有人留言评论。精彩留言会获得点赞!

精彩留言,会给你点赞!