WO2020049717A1

WO2020049717A1 - Signal processing circuit, radar device, signal processing method, and signal processing program

Info

Publication number: WO2020049717A1
Application number: PCT/JP2018/033218
Authority: WO
Inventors: 廣愛浅見; 將白石
Original assignee: 三菱電機株式会社
Priority date: 2018-09-07
Filing date: 2018-09-07
Publication date: 2020-03-12
Also published as: JP6887571B2; JPWO2020049717A1

Abstract

A radar device (1), wherein a signal processing circuit (30) comprises: a region conversion unit (31) that respectively converts a plurality of reception signals inputted from a reception circuit (21) into a plurality of frequency region signals; a target detection unit (32) that senses latest target detection information on the basis of the plurality of frequency region signals; and an arrival direction estimation unit (35). The arrival direction estimation unit (35): calculates a latest correlation matrix on the basis of the plurality of frequency region signals; acquires, from an information storage unit (34), a previous eigenvector corresponding to previous target detection information that coincides with or resembles the latest target detection information; and executes an iterative operation based on an eigenvalue decomposition algorithm using the previous eigenvector and the latest correlation matrix, thereby estimating an eigenvalue and an eigenvector of the latest correlation matrix. The arrival direction estimation unit (35) estimates the arrival direction of one or a plurality of arriving waves using the estimated eigenvalue and eigenvector.

Description

Signal processing circuit, radar device, signal processing method, and signal processing program

The present invention relates to radar technology, and more particularly to receiving an incoming wave reflected by a target object using an antenna array including a plurality of antenna elements, and estimating a direction of arrival (Direction-Of-Arrival, DOA) of the incoming wave. Radar technology.

In the radar technology, there is a technology for estimating the arrival direction of an incoming wave reflected by a target object by using an antenna array including a plurality of spatially arranged antenna elements. As techniques capable of estimating the direction of arrival with high resolution, super resolution algorithms such as ESPRIT (Estimation of Signal, Parameters, Via, Rotation, Innovation, Technologies) and MUSIC (Multiple Signal, Classification) are widely known.

Both the ESPRIT method and the MUSIC method form a correlation matrix based on outputs of a plurality of antenna elements forming an antenna array, estimate eigenvalues and eigenvectors of the correlation matrix, and use the estimated eigenvalues and eigenvectors. This is a technique capable of estimating the number and direction of arrival of one or a plurality of incoming waves with high resolution. For example, Non-Patent Document 1 below discloses a TLS-ESPRIT (Total-Least-Squares @ ESPRIT) method which is a kind of the ESPRIT method.

However, in the conventional super-resolution algorithm using the eigenvalues and eigenvectors of the correlation matrix, such as the ESPRIT method and the MUSIC method, when the number of antenna elements is increased to improve the estimation accuracy of the number of incoming waves and the direction of arrival, the correlation matrix The size also increases, and the amount of calculation required for eigenvalue estimation becomes enormous. In such a case, there is a problem that the calculation time increases and it becomes difficult to calculate an estimated value of the arrival direction within a desired limited time. Alternatively, in order to realize a signal processing circuit that calculates an estimated value in a short operation time, it is necessary to increase the circuit scale of the signal processing circuit, which causes a problem that the manufacturing cost is increased.

In view of the above, an object of the present invention is to provide a signal processing circuit, a radar device, a signal processing method, and a signal processing program that can estimate an arrival direction with a low calculation load using an antenna array.

A signal processing device according to one aspect of the present invention is an antenna array including a plurality of receiving antenna elements that receive an incoming wave reflected by a target object, and a plurality of received signals based on outputs of the plurality of receiving antenna elements. A signal processing circuit in a radar apparatus comprising: a receiving circuit that generates a plurality of frequency domain signals; and a domain conversion unit configured to convert the plurality of received signals into a plurality of frequency domain signals; and target detection information based on the plurality of frequency domain signals. And a direction estimating unit that calculates a correlation matrix based on the plurality of frequency domain signals, and estimates a direction of arrival of one or more incoming waves using a set of eigenvalues and eigenvectors of the correlation matrix. And at least one set of previously detected target detection information detected in the past and a plurality of previous eigenvectors estimated in the past. Comprising a storage unit, when the target detection unit detects the latest target detection information based on the plurality of frequency domain signals, the arrival direction estimation unit, the latest correlation matrix based on the plurality of frequency domain signals Is obtained from the information storage unit, a plurality of previous eigenvectors corresponding to the target detection information that matches or is similar to the latest target detection information, and the obtained plurality of previous eigenvectors and the latest An eigenvalue and a set of eigenvectors of the latest correlation matrix are estimated by performing an iterative operation based on a predetermined eigenvalue decomposition algorithm using the correlation matrix.

According to the present invention, the arrival direction estimating unit obtains a plurality of previous eigenvectors corresponding to the target detection information that matches or is similar to the latest target detection information from the information storage unit, and obtains the eigenvectors of the destination and the latest eigenvectors. By performing an iterative operation based on the eigenvalue decomposition algorithm using the correlation matrix, the latest set of the eigenvalues and eigenvectors of the correlation matrix is estimated, so that the eigenvalues can be estimated with high precision with a small number of iterations. Therefore, it is possible to suppress the calculation load required for eigenvalue estimation. Therefore, even if the number of receiving antenna elements forming the antenna array increases, the direction of arrival can be estimated with high accuracy in a short calculation time without increasing the circuit scale of the signal processing circuit.

FIG. 1 is a block diagram illustrating a schematic configuration of a radar device according to a first embodiment of the present invention. FIG. 2A is a graph showing an example of the time change of each of the frequency of the transmission wave and the frequency of the reception wave, and FIG. 2B is a graph showing an example of the frequency spectrum of the first frequency domain signal. FIG. 3 is a diagram schematically illustrating an example of an arrangement of reception antenna elements forming an antenna array according to the first embodiment. FIG. 3 is a block diagram schematically illustrating a hardware configuration example of a signal processing circuit according to the first embodiment. FIG. 3 is a block diagram schematically illustrating a configuration example of an area conversion unit according to the first embodiment. FIG. 5 is a diagram illustrating an example of target information stored in an information storage unit according to the first embodiment. 5 is a flowchart illustrating an example of radar processing in the radar device according to the first embodiment. 5 is a flowchart illustrating an example of an arrival direction estimation process according to Embodiment 1. 5 is a flowchart illustrating an example of an arrival direction estimation process according to Embodiment 1. 5 is a flowchart illustrating an example of an arrival direction estimation process according to Embodiment 1. FIG. 9 is a block diagram illustrating a schematic configuration of a radar device according to a second embodiment of the present invention. 15 is a flowchart illustrating an example of an arrival direction estimation process according to Embodiment 2.

Hereinafter, various embodiments according to the present invention will be described in detail with reference to the drawings. Note that components denoted by the same reference numerals throughout the drawings have the same configuration and the same function.

Embodiment 1 FIG.
FIG. 1 is a block diagram illustrating a schematic configuration of a radar device 1 according to a first embodiment of the present invention. As shown in FIG. 1, the radar device 1 includes a transmission circuit 11 that periodically generates a frequency modulation wave in a high frequency band such as a millimeter wave band or a microwave band, and a frequency modulation wave input from the transmission circuit 11. a transmitting antenna 10 for transmitting, the antenna array of the receiving antenna elements 20 _0-20 ₃ for receiving the frequency-modulated wave reflected by a target object existing in the external space (not shown.) as the incoming wave (signal wave) 20, a receiving circuit 21 for generating a receive antenna elements 20 0 _to the digital reception signal of the four reception channels based on the high frequency band of the signals outputted in parallel from 20 ₃ (digital beat signals), these digital reception signal To the target object, the relative speed of the target object, and the direction of arrival of the arriving wave reflected by the target object (Direction). -Of-Arrival, and a signal processing circuit 30 for calculating values such as DOA).

The transmission circuit 11 includes a signal generator 12, a distributor 13, and a transmission amplifier 14. The signal generator 12 generates a frequency-modulated signal by repeatedly generating a frequency-modulated wave modulated by a predetermined frequency modulation method according to the control signal supplied from the signal processing circuit 30. As a frequency modulation method, a frequency modulation continuous wave (Frequency Modulated Continuous Wave, FMCW) method can be used. The signal generator 12 can generate a frequency modulation signal (chirp signal) by repeatedly generating a frequency modulation wave (chirp wave) by the FMCW method. The frequency of the frequency modulation signal, that is, the transmission frequency, may be swept so as to repeatedly increase or decrease continuously with time within a certain frequency bandwidth. Alternatively, the transmission frequency may be swept so as to repeatedly increase and then continuously decrease with time within a certain frequency bandwidth.

Distributor 13 distributes the frequency modulation signal input from signal generator 12 into a transmission signal and local signal LO. The distributor 13 outputs a transmission signal to the transmission amplifier 14 and outputs a local signal LO to the reception circuit 21. The transmission amplifier 14 amplifies the transmission signal and outputs the amplified transmission signal to the transmission antenna 10. Then, the transmission antenna 10 radiates the amplified transmission signal to an external space.

送信 The transmitting circuit 11 continuously transmits M frequency modulated waves as one frame. More specifically, when the transmission circuit 11 continuously outputs the 0th to M−1th frequency modulated waves within a certain frame period, the transmission circuit 11 outputs the 0th to M−1th frequency modulation waves within a next frame period after a certain time interval. It is assumed that the (M-1) th frequency modulated wave is continuously output. Here, M is an integer of 2 or more.

FIG. 2A is a graph showing an example of a time change of each of the frequency Tf of the transmission wave and the frequency Rf of the reception wave when a fast chirp modulation (FCM) system, which is a kind of the FMCW system, is adopted. . As shown in FIG. 2A, the frequency Tf of the transmitted wave, straight line as varied as sawtooth, from the lower limit frequency f ₁ that is specified, continuously change with to a specified upper limit frequency f ₂ Time Modulated. In the example of FIG. 2A, M frequency-modulated waves (chirp waves) are continuously transmitted within one frame period, and a received wave is received with a delay of Δt from the transmitted frequency-modulated wave. I have.

Referring to FIG. 1, when receiving antenna elements 20 ₀ to 203 ₃ receive an incoming wave reflected by a target object, the receiving antenna elements 20 ₀ to 203 output signals of high frequency bands of four reception channels to receivers 21 ₀ to 21 ₃ , respectively. I do. Figure 3 is a diagram schematically showing an arrangement example of the receiving antenna elements 20 _0-20 ₃ constituting the antenna array 20. Receive antenna elements 20 _0-20 ₃ shown in FIG. 3, are arranged at equal intervals d along the x-axis direction in a linear base line. The installation position of the receiving antenna element 20 ₃ as the origin (reference point), along the x-axis positive direction, the receiving antenna element ₂₀ _3, 20 _2, 20 1, 20 ₀ are arranged in this order. In the example of FIG. 3, a state is shown in which an incoming wave is incident at an incident angle θ with respect to a y-axis direction orthogonal to the x-axis direction. The y-axis direction is perpendicular to the antenna surface of the antenna array 20.

The receiving circuit 21 has four receivers 21 ₀ to 21 ₃ connected to the receiving antenna elements 20 ₀ to 20 ₃ , respectively. Each receiver 21 _ch includes a reception amplifier 22 _ch such as a low noise amplifier (LNA) that amplifies the output of the reception antenna element 20 _ch , and an amplified signal output from the reception amplifier 22 _ch as a local signal LO. by mixing it includes a mixer circuit 23 _ch for generating an analog beat signal in the intermediate frequency band, and the a / D converter for converting the analog beat signal into a digital beat signal (ADC) 24 _ch. Here, the subscript ch is a reception channel number and is an integer in the range of 0 to 3.

The ADC 24 _ch generates a digital beat signal B _k (ch, m, n) (n = 0 to N−1) by sampling the analog beat signal at a predetermined sampling interval for each chirp wave. Here, m is a chirp number (chirp index) indicating the number of a chirp wave, and N is the number of sampling points. The subscript k is an integer representing the current time, that is, the time of the current frame period. For example, a time one frame period before the time of the current frame period is represented by k-1. The ADC 24 _ch has a function of outputting the digital beat signal B _k (ch, m, n) as a complex signal including a quadrature component and an in-phase component.

The ADC 24 _ch outputs the digital beat signal B _k (ch, m, n) to the signal processing circuit 30 as a digital reception signal. Between the mixer circuit 23 _ch and ADC 24 _ch, filter circuit for removing unnecessary signal components of the analog beat signal may be disposed.

Next, referring to FIG. 1, the signal processing circuit 30 converts the digital reception signals B _k (0, m, n) to B _k (3, m, n) in the time domain into the frequency domain signal D in the two-dimensional frequency domain. a domain conversion unit 31 for converting _k (0, fv, fr) to D _k (3, fv, fr), respectively, and a frequency domain signal D _k (0, fv, fr) to D _k (3, fv, fr) And a correlation matrix Cxx based on the frequency domain signals D _k (0, fv, fr) to D _k (3, fv, fr), and a predetermined eigenvalue An arrival direction estimating unit 35 that estimates the eigenvalues and eigenvectors of the correlation matrix Cxx by executing a decomposition algorithm, and estimates the arrival directions of one or more arriving waves using the estimated eigenvalues and eigenvectors; Target detection information, estimation Information storage unit 34 in which the set of the obtained eigenvalues and eigenvectors and the combination of the estimated direction of arrival are stored; information output unit 38 for outputting target information Dc read from information storage unit 34 to the outside; 39 are provided. The control unit 39 controls the operation of the signal generator 12 in the transmission circuit 11 and controls the operations of the area conversion unit 31, the target detection unit 32, the information storage unit 34, the arrival direction estimation unit 35, and the information output unit 38. It has the function of controlling individually. The control unit 39 is connected to the transmission circuit 11, the area conversion unit 31, the target detection unit 32, the information storage unit 34, the arrival direction estimation unit 35, and the information output unit 38 via signal paths such as a system bus and control signal lines. Have been.

The hardware configuration of such a signal processing circuit 30 is, for example, a semiconductor integrated circuit using a processor such as a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), or a processor including a processor (Field-Programmable Gate Array) having an FPGA (Field-Programmable Gate Array). It should be done. Alternatively, the hardware configuration of the signal processing circuit 30 may be a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) that executes a program code (instruction group) of signal processing software or firmware read from the memory. May be realized by using a processor including the arithmetic unit. It is also possible to realize the hardware configuration of the signal processing circuit 30 using a processor having a combination of the semiconductor integrated circuit and the arithmetic device. Furthermore, each of the area conversion unit 31, the target detection unit 32, the arrival direction estimation unit 35, the information output unit 38, and the control unit 39 may be configured by dedicated hardware, or may be configured by one or a plurality of hardware. It may be configured.

FIG. 4 is a block diagram schematically showing an example of a hardware configuration of the signal processing circuit 30. The signal processing circuit 30 includes a processor 71, a memory 72, an input / output interface unit 73, and a signal path 74. The signal path 74 is a bus for interconnecting the processor 71, the memory 72, and the input / output interface unit 73. The input / output interface unit 73 has a function of transferring the digital reception signal input from the reception circuit 21 to the processor 71 via the signal path 74. Processor 71 performs digital signal processing on the transferred digital reception signal. The processor 71 can output the target information Dc obtained as a result of the digital signal processing to an external device (not shown) via the signal path 74 and the input / output interface unit 73.

The memory 72 is for realizing the functions of the non-volatile memory forming the storage area of the information storage unit 34 and the functions of the area conversion unit 31, the target detection unit 32, the arrival direction estimation unit 35, the information output unit 38, and the control unit 39. A non-volatile memory storing a signal processing program such as software or firmware for signal processing, a work memory used when the processor 71 executes digital signal processing, and data used in the digital signal processing are developed. And a temporary storage memory. For example, the memory 72 may be configured by a semiconductor memory such as a flash memory, a ROM (Read Only Memory), and an SDRAM (Synchronous Dynamic Random Access Memory).

In the example of FIG. 4, the number of the processors 71 is one, but the number is not limited to one. The hardware configuration of the signal processing circuit 30 may be realized using a plurality of processors operating in cooperation with each other.

Next, the configuration of the signal processing circuit 30 shown in FIG. 1 will be described in detail. The domain conversion unit 31 performs a two-dimensional orthogonal transform on the digital reception signal B _k (ch, m, n) (m = 0 to M−1, n = 0 to N−1) for each reception channel. A digital received signal B _k (ch, m, n) at M × N points in the time domain is converted into a frequency domain signal D _k (ch, fv, fr) at M × N points in the two-dimensional frequency domain (fv = 0 to M -1, fr = 0 to N-1). Here, fr is a frequency bin number (hereinafter, referred to as “distance frequency bin number”) assigned to a frequency corresponding to the distance to the target object (hereinafter, referred to as “distance frequency”), and fv is a target object. Is a frequency bin number (hereinafter, referred to as a “speed frequency bin number”) assigned to a frequency corresponding to the relative speed (hereinafter, referred to as a “speed frequency”). As the two-dimensional orthogonal transform, a two-dimensional discrete Fourier transform may be used, but the present invention is not limited to this. The frequency domain signal D _k (ch, fv, fr) is a three-dimensional data signal having an amplitude distribution related to a distance frequency and a speed frequency for each reception channel number ch.

FIG. 5 is a block diagram schematically illustrating a configuration example of the area conversion unit 31 according to the first embodiment. As shown in FIG. 5, the region converter 31, a window function processing unit ₄₁ _0, 41 _1, 41 2, 41 a first pre-processing unit 41 consisting of _three chirp in the

orthogonal transform unit

₄₂ 0, ₄₂ 1, 42 and _2, 42 ₃ the first orthogonal transformation unit 42 consisting of a window function processing unit ₄₃ _0, 43 _1, 43 2, 43 a second pre-processing unit 43 consisting of _three chirp between

orthogonal transform unit

₄₄ 0, 44 ₁ , and a second orthogonal transformation unit 44 consisting of ₄₄ 2, 44 _3.

Each window function processing unit 41 _ch in the first pre-processing unit 41 performs a window function on N digital reception signals B _k (ch, m, n) (n = 1 to N) input for each chirp wave. By executing the processing, signals at N points are output. In this window function process, for example, a known window function such as a hamming window function or a Blackman-Harris window function may be used.

Each of the intra-chirp orthogonal transform units 42 _ch in the first orthogonal transform unit 42 performs orthogonal transform on the N-point signal input from the window function processing unit 41 _ch, thereby forming the N-point first frequency-domain signal T _k. (Ch, m, fr) (fr = 1 to N) is generated. As the orthogonal transform, a discrete Fourier transform such as a fast Fourier transform (FFT) can be used. The window function processing in the above-described window function processing unit _41ch is processing for suppressing the distortion of the spectrum that occurs at the time of the orthogonal transformation to achieve both improvement in the spectral resolution and expansion of the dynamic range. FIG. 2B shows a frequency spectrum | T _k (ch, 0, fr) |, | T _k (ch, 1, fr) |,..., | T of the first frequency domain signal T _k (ch, m, fr). It is a graph which shows the example of _k (ch, M-1, fr) |.

Next, each window function processing unit 43 _ch in the second preprocessing unit 43 outputs the first frequency domain signal T _k (ch, m, fr) (m = m) at M points input for each distance frequency bin number fr. By executing the window function processing on 0 to M−1), a signal at M points is output. In this window function processing, for example, a known window function such as a Hamming window function or a Blackman-Harris window function may be used.

The inter-chirp orthogonal transform unit 44 _ch in the second orthogonal transform unit 44 performs orthogonal transform on the signal at M points input from the window function processing unit 43 _ch for each distance frequency bin number fr, thereby obtaining the second frequency. A frequency domain signal D _k (ch, fv, fr) (fv = 0 to M−1) at M points is generated as a domain signal. As the orthogonal transform, a discrete Fourier transform such as a fast Fourier transform can be used. The window function processing in the above-described window function processing unit _43ch is processing for suppressing the distortion of the spectrum that occurs at the time of the orthogonal transformation to achieve both improvement in the spectral resolution and expansion of the dynamic range.

Referring to FIG. 1, the target detection unit 32 receives frequency domain signals D _k (ch, fv, fr) (ch = 0 to 3) for four reception channels as inputs, and receives these frequency domain signals D _k (ch , Fv, fr) with respect to the reception channel number ch, to generate an integrated signal I _k (fv, fr) related to a two-dimensional frequency (distance frequency and velocity frequency). For example, the target detection unit 32 calculates the square | D _k (ch, fv, fr) | ² of the absolute value of the frequency domain signal D _k (ch, fv, fr) as shown in the following equation (1). The integration signal I _k (fv, fr) may be generated by adding (incoherent integration) for the channel number ch.

Next, the target detection unit 32 detects a peak value from the distribution of the integrated signal I _k (fv, fr), and a distance frequency bin number indicating the position of the detected peak value (hereinafter, referred to as “peak position”). And a set of velocity frequency bin numbers (fv _p , fr _p ). For example, using the threshold th0, the integrated signal I _k (fv, fr) that satisfies all of the following conditional expressions (2) to (6) can be detected as a peak value. Here, the threshold value th0 is a threshold value capable of excluding a signal of power corresponding to the noise level.

I _k (fv, fr−1) <I _k (fv, fr) (2)
I _k (fv, fr)> I _k (fv, fr + 1) (3)
I _k (fv−1, fr) <I _k (fv, fr) (4)
I _k (fv, fr)> I _k (fv + 1, fr) (5)
I _k (fv, fr)> th0 (6)

Target detection unit 32, the distance frequency bin number fr _p from it is possible to calculate the distance between the target object, it is possible to calculate the relative velocity of the target object from the speed frequency bin number fv _p. Note that the target detection unit 32 may identify the target object and generate an identification result. Then, the target detection unit 32 sends the arrival direction estimation unit 35 the target detection information such as the time information, the peak position (a set of the distance frequency bin number and the speed frequency bin number), the distance to the target object, and the relative speed of the target object. The information is supplied and stored in the information storage unit 34.

Note that, as described above, the target detection unit 32 performs incoherent integration to calculate the integration signal I _k (fv, fr), but is not limited thereto. Instead of incoherent integration, other processing including coherent integration may be employed.

Also, as for the threshold th0, the threshold th0 may be determined using a method such as CFAR (Constant False False Alarm) used in general radar technology, instead of the above method.

The arrival direction estimating unit 35 shown in FIG. 1 includes a correlation calculating unit 52 that calculates a correlation matrix Cxx based on the frequency domain signals D _k (0, fv, fr) to D _k (3, fv, fr), An eigenvalue calculation unit 53 that estimates an eigenvalue of the correlation matrix Cxx by executing an iterative operation based on an eigenvalue decomposition algorithm of Eq., An eigenvector calculation unit 54 that estimates an eigenvector of the correlation matrix Cxx, and Direction-of-arrival calculation unit 55 for estimating the direction of arrival of one or more incoming waves. The arrival direction estimation unit 35 stores the estimated arrival direction together with the estimated eigenvalue and eigenvector in the information storage unit 34 in association with the target detection information detected by the target detection unit 32.

FIG. 6 is a diagram illustrating an example of the target information DD _k (p) stored in the information storage unit 34. The target information DD _k (p) is information on the p-th incoming wave (p is an integer of 1 or more) detected at the current time k. As shown in FIG. 6, the target information DD _k (p) includes an identifier, the number of consecutive detections N _k indicating the frequency of detection of the target object, time information indicating the time k, and the peak position (distance frequency bin number and speed frequency bin number). ), The distance to the target object, the relative speed of the target object, the eigenvalue, the eigenvector, and the direction of arrival. The target detection information in the target information DD _k (p) includes time information, a peak position (a set of a distance frequency bin number and a speed frequency bin number), a distance, and a relative speed. As shown in FIG. 1, the information storage unit 34 also stores target information DD _k-1 (1), DD _k-1 (2),... Regarding time k-1 before current time k. I have.

Furthermore, the direction-of-arrival estimating unit 35 of the present embodiment includes a comparison search unit 51. The comparison search unit 51 searches for the target information stored in the information storage unit 34, and searches for the target detection information that matches or is similar to the latest target detection information detected at the current time k, and The destination eigenvector corresponding to the detection information can be acquired from the information storage unit 34. Here, the previous target detection information is the target detection information detected at time ki (i is an integer of 1 or more) before the current time k.

The comparison search unit 51 supplies the eigenvector obtained from the information storage unit 34 to the eigenvalue calculation unit 53. As described below, the eigenvalue calculation unit 53 can execute an iterative operation based on the eigenvalue decomposition algorithm using the preceding eigenvector and the correlation matrix Cxx. The eigenvalue of the correlation matrix Cxx can be estimated in a shorter operation time than in a case where an iterative operation based on a decomposition algorithm is executed. Accordingly, the arrival direction estimation unit 35 can estimate the arrival direction of the arrival wave reflected by the target object in a short calculation time.

Hereinafter, the operation of the radar device 1 according to the first embodiment will be described with reference to FIGS. 7 to 10, and the operation and configuration of the arrival direction estimating unit 35 will be described in detail.

FIG. 7 is a flowchart illustrating an example of radar processing in the radar device 1 according to the first embodiment. 8 to 10 are flowcharts illustrating an example of the arrival direction estimation processing executed by the arrival direction estimation unit 35. The flowchart of FIG. 8 is connected to the flowchart of FIG. 9 via the connector C0, and is connected to the flowchart of FIG. 10 via the connector C1. The flowchart of FIG. 10 is combined with the flowchart of FIG. 9 via the connector C2.

Referring to FIG. 7, as described above, when receiving a control signal indicating a transmission start command from control unit 39, transmission circuit 11 receives a control signal indicating a transmission start command, and in accordance with a predetermined frequency modulation scheme such as the FMCW scheme in accordance with the transmission start command, transmits the signal. A modulated wave (chirp wave) is transmitted (step ST10). Thereafter, the receiving antenna elements ₂₀ 0-20 ₃ When receiving the incoming waves reflected by the target object (step ST11), the reception circuit 21 based on the output of the receiving antenna elements ₂₀ 0-20 _3, the digital reception signal B _k (0,0, n), B _k (1,0, n), B _k (2,0, n), B _k (3,0, n) (n = 0 to N−1) (Step ST12). Thereafter, steps ST10, ST11, and ST12 are repeatedly executed until M frequency modulated waves for one frame are transmitted (NO in step ST13). As a result, M frequency-modulated waves are continuously transmitted within one frame period, and the receiving circuit 21 receives the digital reception signals B _k (0, m, n) and B _k (1, m) at M × N points. , N), B _k (2, m, n) and B _k (3, m, n) (m = 0 to M−1, n = 0 to N−1). When it is determined that the transmission is to be ended (YES in step ST13), the next step ST14 is executed.

In step ST14, as described above, the domain conversion unit 31 sets the frequency domain signals D _k (ch, fv, fr) at M × N points (fv = 0 to M−1, fr = 0 to N-1).

Next, the target detection unit 32 generates the integrated signal I _k (fv, fr) by integrating the frequency domain signal D _k (ch, fv, fr) input from the domain conversion unit 31 with respect to the reception channel number ch. (Step ST15). Subsequently, the target detection unit 32 attempts to detect target detection information (peak position, distance to the target object, and relative speed of the target object) based on the integration signal I _k (fv, fr) (step ST16). When the target detection information is not detected (NO in step ST17), the control unit 39 determines whether or not to continue the radar processing (step ST22). When it is determined that the radar processing is to be continued (YES in step ST22), the control unit 39 returns the procedure of the radar processing to step ST10. On the other hand, when it is determined that the radar processing is not continued (NO in step ST22), the control unit 39 ends the radar processing.

When the target detection information is detected in step ST17 (YES in step ST17), the target detection unit 32 stores the target detection information in the information storage unit 34 (step ST18). At this time, the target detection unit 32 sets the value of the number of continuous detections _Nk shown in FIG. 6 to zero.

Thereafter, the direction-of-arrival estimation unit 35 performs a direction-of-arrival estimation process (step ST20). Referring to FIG. 8, first, the correlation calculation unit 52 converts the frequency domain signals D _k (0, fv, fr) to D _k (3, fv, fr) based on the Spatial Smoothing Preprocessing (SSP). Then, a correlation matrix (latest correlation matrix) Cxx is calculated (step ST30). Specifically, the correlation calculation unit 52 uses a correlation matrix Rxx expressed by the following equation (7) for calculating the correlation matrix Cxx.

Rxx = X _k · X _k ^H (7)

In equation (7), the dot symbol “•” represents a matrix product, and the superscript H represents Hermite conjugate (conjugate transpose). The correlation matrix Rxx is a Hermitian matrix having the same number of rows as the number of reception channels and the same number of columns as the number of reception channels. X _k in equation (7) is a four-row, one-column element having frequency domain signals D _k (0, fv, fr) to D _k (3, fv, fr) as shown in the following equation (8). Is a vector.

Here, the superscript T indicates transposition.

For example, as shown in the following equation (9), the correlation calculating unit 52 calculates the Q partial correlation matrices Rxx (q) (q = 1,..., Q) extracted from the correlation matrix Rxx shown in the equation (7). ) Is weighted by a coefficient z _q (= 1 / Q), and the weighted partial correlation matrix z _q × Rxx (q) is added, whereby a correlation matrix Cxx based on the spatial average method can be calculated.

Equation (9) is an equation for executing the averaging process of the Q point. Generally, when the averaging process of the Q points is performed, if the correlation matrix Rxx is a square matrix of K rows and K columns (K is an integer of 3 or more), the correlation matrix Cxx is (K−Q + 1) rows (K− Q + 1) square matrix. In the present embodiment, it is possible to extract two 3-row, 3-column partial correlation matrices Rxx (1) and Rxx (2) from the 4-row, 4-column correlation matrix Rxx along the diagonal of the correlation matrix Rxx. It is. By using the correlation matrix Cxx of Expression (9) without using the correlation matrix Rxx of Expression (7) as it is, there is an advantage that the cross-correlation between arriving waves can be suppressed. The correlation matrix Cxx may be calculated using not only the above-described spatial averaging method but also another spatial averaging method.

After the execution of step ST30, the comparison search unit 51 determines in step ST16 from the target detection information detected at the time ki preceding the current time k and stored in the information storage unit 34. The target detection information that matches or is similar to the latest detected target detection information is searched (step ST31). Here, the time ki is a time preceding the current time k by an i-frame period (i is, for example, an integer in the range of 1 to 9). The smaller the difference between the current time k at which the latest target detection information was detected and the time ki at which the previous target detection information stored in the information storage unit 34 was detected, the smaller the latest target detection information. It can be expected that the correlation between the target detection information and the target detection information is high. Therefore, the comparison search unit 51 can limit the time ki to a time k−1 (i = 1) immediately before the current time k. When there is no previous target detection information detected at time k−1, the comparison search unit 51 selects the latest target detection information from the previous target detection information detected at a time earlier than time k−1. The target detection information that matches or is similar to the target detection information may be searched.

More specifically, the comparison search unit 51 determines, for each of the latest target detection information, the degree of similarity or difference between the previous target detection information stored in the information storage unit 34 and the latest target detection information. (E.g., a value inversely proportional to the similarity) is calculated and, based on the similarity or the difference, matches the latest target detection information from the previous target detection information stored in the information storage unit 34. Alternatively, one similar target detection information can be found.

As illustrated in FIG. 6, the target detection information includes a plurality of elements such as time information, a peak position (a set of a distance frequency bin number and a speed frequency bin number), a distance to a detected target object, and a relative speed. Therefore, a vector composed of all or some of the plurality of elements can be configured based on the target detection information. The comparison search unit 51 determines a vector-to-vector distance (norm) such as a Euclidean distance or a Manhattan distance between a vector configured based on the latest target detection information and a vector configured based on the previous target detection information. The square of the distance between the vectors can be calculated as the degree of difference. Alternatively, the comparison search unit 51 may calculate the reciprocal of the degree of difference as the degree of similarity.

For example, p _k-th detected at time k (p _k is an integer of 1 or more) the distance between the target object r _k of _(p _k), the relative velocity of the detected p _k-th target object at time k v _k _(p k), the distance indicating the _{p k-th} peak position detected at time k frequency bin number and speed frequency bin number, respectively _fr k _(p _k), and represents the fv k _{(p k)} . Further, _{p t} th detected at than time k before time t (t = k-i) the distance between the target object _{(p t} is an integer of 1 or more) _r t _(p t), at time t The relative speed of the detected _pt- th target object is v _t ( _pt ), and the distance frequency bin number and the speed frequency bin number indicating the _pt- th peak position detected at time t are fr _t ( _pt _), it shall be expressed as fv _{t (p} t). At this time, it constructed based on the latest target detection information p _k-th vector V _{k (p} _k) and configured based on the target detection information of the detected previously before time t vector V _t ( p _k ) can be expressed by, for example, a set of the following equations (10A) and (11A) or a set of the following equations (10B) and (11B).

In this case, the comparison search unit 51 uses one of the set of the formulas (10A) and (11A) or the set of the formulas (10B) and (11B) to express the following formula (12A) or (12B). Thus, the difference Δ (k, _pk , t, _pt ) can be calculated.

Alternatively, comparison search unit 51, the dissimilarity Δ using the weight coefficients _w 1, _{w 2} as shown in the following equation (13A) or _{(13B) (k, p k} , t, p t) be calculated Good.

The values of the weight coefficients w ₁ and w ₂ may be set or calculated in advance based on past measurement results. By adjusting the values of the weight coefficients w ₁ and w ₂ , the comparison search unit 51 can perform a search that emphasizes either the relative speed or the distance.

The comparison search unit 51, for each of the latest target detection information, from among the previous target detection information stored in the information storage unit 34, the degree of similarity with the latest target detection information is equal to or more than a certain value, In addition, it is sufficient to find only one having the maximum similarity. Alternatively, for each of the latest target detection information, the comparison search unit 51 determines, from among the previous target detection information stored in the information storage unit 34, that the degree of difference from the latest target detection information is equal to or less than a certain value. It is only necessary to find one that has the least difference.

Next, for each of the latest target detection information, if the previous target detection information that matches or is similar to the latest target detection information is not found (NO in step ST32), the comparison search unit 51 and the eigenvalue calculation The unit 53 and the eigenvector calculation unit 54 execute steps ST34, ST41 to ST49, and ST61 to ST66 (FIG. 9). On the other hand, when the target detection information that matches or is similar to the latest target detection information is found for each of the latest target detection information (YES in step ST32), the comparison search unit 51 and the eigenvalue calculation unit 53. In addition, the eigenvector calculation unit 54 can execute steps ST35 and ST36 (FIG. 8), steps ST37 to ST39, ST51 to ST59 (FIG. 10), and steps ST49 and ST61 to ST66 (FIG. 9). For example, when two latest target detection information are detected, the target detection information that matches or is similar to the first latest target detection information among the detected latest target detection information is not found. At this time, steps ST34, ST41 to ST49, and ST61 to ST66 (FIG. 9) are executed, and the target detection information of the destination that matches or is similar to the second latest target detection information among the detected latest target detection information. Are found, steps ST35 and ST36 (FIG. 8), steps ST37 to ST39, ST51 to ST59 (FIG. 10), and steps ST49 and ST61 to ST66 (FIG. 9) are executed.

If the target detection information that matches or is similar to the latest target detection information has not been found (NO in step ST32), the eigenvalue calculation unit 53 and the eigenvector calculation unit 54 calculate the correlation matrix Cxx calculated in step ST30. The eigenvalue and the eigenvector of the correlation matrix Cxx are calculated by executing an iterative operation based on an eigenvalue decomposition (Eigenvalue @ Decomposition) algorithm using as an initial matrix (steps ST34 and ST41 to ST48 in FIG. 9).

Specifically, eigenvalue calculation section 53 converts correlation matrix Cxx to Hessenberg matrix A ⁽⁰⁾ using conversion matrix H ₀ based on, for example, the known Householder method (step ST34), and iteratively. The value of the number j is set to zero (step ST41). Then, eigenvalue calculation section 53, decomposed into a product of a unitary matrix matrix ^{A (j)} based on known QR decomposition (QR decomposition) method _{Q j} and an upper triangular matrix _{R j} (step ST42). Then, eigenvalue calculation section 53, the matrix ^{A (j)} as shown in the following equation (14) calculating a similarity transformation matrix T by similarity transformation with a unitary matrix _{Q j} (step ST43).

T = Q _j ^H · A ^(j) · Q _j (14)

Next, the eigenvalue calculation unit 53 determines whether or not the similarity transformation matrix T has converged, that is, whether or not the similarity transformation matrix T satisfies a predetermined convergence condition (step ST44). When it is determined that the similarity transformation matrix T has not converged (NO in step ST44), the eigenvalue calculation unit 53 increments the number of iterations j by 1 (step ST45), and replaces the elements of the similarity transformation matrix T with the matrix A ^{( j)} (step ST46). Subsequently, the eigenvalue calculation unit 53 executes steps ST42, ST43, and ST44. Finally, when it is determined that the similarity transformation matrix T has converged (YES in step ST44), the eigenvalue calculation unit 53 converts the diagonal elements of the similarity transformation matrix T into the eigenvalues Λ (0), ..., of the correlation matrix Cxx. Λ (n−1) (n is a positive integer) can be estimated (step ST47).

After the execution of step ST47, the eigenvector calculation unit 54 performs an inverse transformation using the unitary matrices Q ₀ , Q ₁ ,..., Q _Nq−1 used in step ST43 and the transformation matrix H ₀ used in step ST34. By executing, the eigenvectors v (0),..., V (n-1) of the correlation matrix Cxx are calculated (step ST48). The eigenvectors v (0), ..., v (n-1) correspond to the eigenvalues Λ (0), ..., Λ (n-1), respectively. Specifically, the eigenvector calculation unit 54 can calculate the eigenvectors v (0), ..., v (n-1) based on the following equation (15).

Here, Nq is the number of iterations required until the similarity transformation matrix T converges.

After the execution of step ST48, the arrival direction calculation unit 55 rearranges the eigenvalues Λ (0),..., Λ (n−1) and the eigenvectors v (0),..., V (n−1) of the correlation matrix Cxx ( Step ST49). Specifically, the arrival direction calculation unit 55 rearranges the eigenvalues Λ (0),..., Λ (n−1) in descending order (in descending order), so that the eigenvalue λ ( 0),..., Λ (n-1).

{Λ (0) ≧... ≧ λ (n−1)}

Equation (16) means that λ (α) ≧ λ (β) always holds for arbitrary integers α and β that satisfy α <β. Further, the arrival direction calculation unit 55 rearranges the eigenvectors v (0),..., V (n−1), and thereby the eigenvectors vc (0) respectively corresponding to the eigenvalues λ (0),. ),..., Vc (n-1).

Next, the direction-of-arrival calculating section 55 estimates the number Ni of incoming waves based on the magnitudes of the eigenvalues λ (0) to λ (n-1) obtained in step ST49 (step ST61 in FIG. 9). Specifically, the direction-of-arrival calculation unit 55 uses the assumed value σ ² of the noise level as a threshold, and among the eigen values λ (0) to λ (n−1), the eigen value λ (0) larger than the threshold σ ² ,..., Λ (Ni−1) (Ni is a positive integer) can be estimated as the number of incoming waves Ni. Alternatively, DOA computation unit 55, the eigenvalue λ (0), ..., λ a value obtained by multiplying a preset factor to a large eigenvalue k _z th of (n-1) is used as a threshold th1 , Λ (n−1), the number of eigenvalues λ (0),..., Λ (Ni−1) larger than the threshold th1 may be estimated as the number of arriving waves Ni. Here, _kz is an integer less than or equal to n, and is a number of an eigenvalue that is larger than the assumed number of incoming waves and can be estimated to be equivalent to a noise level.

Next, the direction-of-arrival calculating unit 55 executes an algorithm based on the ESPRIT method to estimate the direction of arrival of one or more arriving waves reflected by the target object (steps ST62 to ST65).

Specifically, the direction-of-arrival calculation unit 55 firstly selects, from among the eigenvectors vc (0),..., Vc (n-1), the eigenvectors vc (0),. 1) is extracted, and partial matrices Ex and Ey are generated from a partial space matrix Es composed of these eigenvectors vc (0),..., Vc (Ni-1) (step ST62). The subspace matrix Es is given by the following equation (17).

{Es = [vc (0), ..., vc (Ni-1)]} (17)

The 空間 subspace matrix Es is a matrix having the eigenvectors vc (0),..., Vc (Ni-1) as column elements (column vectors), as shown in Expression (17). The arrival direction calculation unit 55 can generate the submatrix Ex with eigenvectors from the first row to the L-1 row (L is the number of receiving antenna elements) of the subspace matrix Es, for example. , The sub-matrix Ex can be generated using the eigenvectors from the second row to the L-th row.

{After performing step ST62, the arrival direction calculating unit 55 calculates a matrix を満たす that satisfies the following equation (18) (step ST63), and calculates an eigenvalue of the matrix ψ (step ST64).

Ey = Ex · ψ (18)

As a method of calculating the matrix ψ, an LS (Least Squares) -ESPRIT method using a pseudo inverse matrix corresponding to an inverse matrix of the sub-matrix Ex, and a TLS (Minimizing the influence of errors included in the sub-matrices Ex and Ey) Total-Least-Squares) -ESPRIT method is known. The arrival direction calculation unit 55 may calculate the matrix ψ based on a known ESPRIT method such as the LS-ESPRIT method or the TLS-ESPRIT method. Since the size of the columns of the sub-matrix Ey changes according to the number Ni of incoming waves, the sub-matrix Ey is not always a square matrix. In the LS-ESPRIT method, a matrix ψ can be calculated by multiplying a partial matrix Ey by a pseudo inverse matrix of the partial matrix Ex. Also, the matrix ψ is not necessarily a Hermitian matrix. As a method of calculating the eigenvalue of the matrix ψ, for example, the same method as in steps ST34 and ST41 to ST47 can be used, but is not particularly limited.

Next, the direction-of-arrival calculating unit 55 calculates the direction of arrival of each arriving wave using the complex argument (phase) φ of each eigenvalue of the matrix 得 obtained in step ST64 (step ST65). When the receiving antenna elements 20 _0-20 ₃ as shown in FIG. 3 are arranged at equal intervals, when the eigenvalues λ of the matrix ψ is given, the incident angle θ indicating the direction of arrival of the incoming wave, For example, it can be calculated based on the following equation (19).

θ = Arcsin (λ · φ / (2πd)) (19)
Here, Arcsin () is a function for obtaining an inverse sine, and λ is a signal wavelength.

Then, the direction-of-arrival calculation unit 55 calculates the latest eigenvectors vc (0) to vc (n-1), the eigenvalues λ (0) to λ (n-1) and the directions of arrival θ ₁ to θ _Ni of the correlation matrix Cxx. The information is stored in the information storage unit 34 in association with the target detection information (step ST66).

After the execution of step ST66, the information output unit 38 reads out the target information Dc such as the distance to the target object, the relative speed and the arrival direction of the target object from the information storage unit 34, and outputs the target information Dc to the outside (step ST21 in FIG. 7). ). The target information Dc output to the outside is used, for example, for tracking processing in a subsequent processing device, or is used as information displayed on a display device. Thereafter, when it is determined that the radar process is to be continued (YES in step ST22), the control unit 39 returns the procedure of the radar process to step ST10. On the other hand, when it is determined that the radar processing is not continued (NO in step ST22), the control unit 39 ends the radar processing.

On the other hand, referring to FIG. 8, if the previous target detection information that matches or is similar to the latest target detection information has been found in step ST31 (YES in step ST32), the comparison search unit 51 returns to the latest search target. the continuous detection number N _k of target detection information is set to the found previous continuous detection number N _k-i was obtained by incrementing by 1 the value of the target detection information _{(= N k-i +1)} ( step ST35). Then, comparing the search unit 51, the continuous detection number _{N k-i} determines whether a predetermined threshold _{N th} or more (step ST36). When the continuous detection number _{N k-i} is determined to be the threshold value _{N th} or more (YES in step ST36), eigenvalue calculation section 53 and the eigenvector computing section 54, ahead of eigenvectors vb corresponding to the target detection information of the destination By using (0),..., Vb (n−1) (n is a positive integer) and the correlation matrix Cxx, an eigenvalue and an eigenvector of the correlation matrix Cxx are calculated by executing an iterative operation based on an eigenvalue decomposition algorithm ( Steps ST37 to ST39 and ST51 to ST59 in FIG. 10). When the continuous detection number _{N k-i} is determined to be less than the threshold value _{N th} (NO in step ST36), steps ST34 (FIG. 9) is executed. As described above, when the number of consecutive detections _Nk indicating the detection frequency of the target object is _{equal to} or greater than the predetermined threshold Nth, the eigenvalue calculation unit 53 and the eigenvector calculation unit 54 perform the repetition using the target detection information with high reliability. Operations can be performed. For the threshold value N _th, for example, it is possible to set two.

Next, referring to FIG. 10, in step ST37, the eigenvalue calculation unit 53 determines, based on the following equation (20), the previous eigenvectors vb (0),..., Vb (n− A transformation matrix B is calculated based on 1).

{B = [vb (0), ..., vb (n-1)]} (20)

The correlation matrix Cxx is a square matrix with n rows and n columns. The transformation matrix B is a matrix having the above eigenvectors vb (0) to vb (n-1) as column elements (column vectors) as shown in Expression (20). Since the correlation matrix Cxx is a Hermitian matrix, the column vectors vb (0) to vb (n-1) are generally orthogonal to each other, and the transformation matrix B is a unitary matrix having n rows and n columns.

Next, the eigenvalue calculation unit 53 executes similarity conversion using the conversion matrix B (step ST38). That is, the intrinsic value calculation unit 53, as shown in the following equation (21), multiplied from the right side of the transformation matrix B in the correlation matrix Cxx, and multiplication from the left to the correlation matrix Cxx the adjoint matrix B ^H of the transformation matrix B Then, a similarity transformation matrix Γ is calculated.

Γ = B ^H・ Cxx ・ B (21)

After step ST38, the eigenvalue calculation unit 53, for example, converts, based on a known Householder method, the similarity transformation matrix Γ using the transformation matrix H ₁ in Hessenberg matrix A ⁽⁰⁾ (step ST39), the number of iterations The value of j is set to zero (step ST51).

Then, eigenvalue calculation section 53, as in step ST42, the matrix ^{A (j)} based on the QR decomposition method decomposes the product of the unitary matrix _{Q j} and an upper triangular matrix _{R j} (step ST52), and step ST43 Similarly, the matrix ^a and ^(j) calculating a similarity transformation matrix T by similarity transformation with a unitary matrix _{Q j} (step ST53). Next, as in step ST44, the eigenvalue calculation unit 53 determines whether the similarity transformation matrix T has converged, that is, whether the similarity transformation matrix T satisfies a predetermined convergence condition (step ST54). When it is determined that the similarity transformation matrix T has not converged (NO in step ST54), the eigenvalue calculation unit 53 increments the number of iterations j by 1 (step ST55), and replaces the elements of the similarity transformation matrix T with the matrix A ^{( j)} (step ST56). Subsequently, the eigenvalue calculation unit 53 executes steps ST52, ST53, and ST54.

Finally, when it is determined that the similarity transformation matrix T has converged (YES in step ST54), the eigenvalue calculation unit 53 converts the diagonal elements of the similarity transformation matrix T into the eigenvalues Λ (0),. Λ (n−1) (n is a positive integer) can be estimated (step ST57).

After execution of step ST57, the eigenvector computing section 54, similarly to the step ST48, by performing the inverse transformation using the transformation matrix _{H 1} used in the unitary matrix _{Q j} and the step ST39, which is used in step ST53, The eigenvectors x (0),..., X (n-1) of the transformation matrix Γ are calculated (step ST58).

After that, as shown in the following equation (22), the eigenvector calculation unit 54 uses the transformation matrix B used in step ST38 to convert the eigenvector x (κ) of the transformation matrix Γ to the eigenvector v (κ) of the correlation matrix Cxx. Is calculated (step ST59).

{V (κ) = B · x (κ)} (22)

Κ In the formula (22), κ is an arbitrary integer in the range of 0 to n−1. The eigenvectors v (0), ..., v (n-1) correspond to the eigenvalues Λ (0), ..., Λ (n-1), respectively.

Since the correlation matrix Cxx is a Hermitian matrix, the eigenvectors v (0) to v (n-1) corresponding to the eigenvalues Λ (0) to Λ (n-1) are generally n linearly independent vectors. is there. Further, even when the eigenvalues Λ (0) to Λ (n−1) are degenerated, it is possible to select linearly independent eigenvectors v (0) to v (n−1). Similarly, the above eigenvectors vb (0) to vb (n-1) are also n linearly independent eigenvectors, and the transformation matrix B generated from the above eigenvectors vb (0) to vb (n-1) Is a unitary matrix. In general, it is known that the eigenvalue of the correlation matrix Cxx matches the eigenvalue of the similarity transformation matrix ても even when the similarity transformation as shown in Expression (21) is performed. However, since the eigenvectors x (0) to x (n-1) of the similarity transformation matrix Γ are different from the eigenvectors v (0) to v (n-1) of the correlation matrix Cxx, the eigenvectors x ( By multiplying the unitary matrix B by (0) to x (n-1), eigenvectors v (0) to v (n-1) can be obtained.

Next, referring to FIG. 9, the direction-of-arrival calculating unit 55 calculates the eigenvalues Λ (0),..., Λ (n−1) and the eigenvectors v (0),. By rearranging 1), eigenvalues λ (0),..., Λ (n−1) and eigenvectors vc (0),..., Vc (n−1) are obtained (step ST49). Next, the direction-of-arrival calculating unit 55 estimates the number Ni of incoming waves based on the magnitude of the eigenvalues λ (0) to λ (n-1) (step ST61). Next, the direction-of-arrival calculation unit 55 executes an algorithm based on the ESPRIT method to estimate the direction of arrival of one or more arriving waves reflected by the target object (steps ST62 to ST65).

After the execution of step ST66, the information output unit 38 reads out the target information Dc such as the distance to the target object, the relative speed and the arrival direction of the target object from the information storage unit 34, and outputs the target information Dc to the outside (step ST21 in FIG. 7). ). Thereafter, when it is determined that the radar process is to be continued (YES in step ST22), the control unit 39 returns the procedure of the radar process to step ST10. On the other hand, when it is determined that the radar processing is not continued (NO in step ST22), the control unit 39 ends the radar processing.

As described above, when target detection information that matches or is similar to the latest target detection information is found (YES in step ST32 of FIG. 8), the eigenvalue calculation unit 53 sets the eigenvalue calculation unit 53 before the current time k. Is calculated from the previous eigenvectors vb (0) to vb (n-1) estimated at the time (step ST37 in FIG. 10), and the correlation matrix Cxx is subjected to similarity transformation using the transformation matrix B. Generates a similarity transformation matrix Γ (step ST38). The eigenvalue calculation unit 53 and the eigenvector calculation unit 54 calculate an eigenvalue and an eigenvector of the correlation matrix Cxx by executing an iterative operation based on an eigenvalue decomposition algorithm using the similarity transformation matrix として as an initial matrix (steps ST39, ST51 to ST59). ). For this reason, the iterative operation shown in FIG. 10 requires a shorter operation time than the iterative operation using the correlation matrix Cxx as it is as the initial matrix (steps ST34 and ST41 to ST48 in FIG. 9). T can be made to converge, whereby the eigenvalue of the correlation matrix Cxx can be calculated in a short time.

By the way, in the processing of the comparison search unit 51, the target detection information based on the previous eigenvectors vb (0) to vb (n-1) and the latest target detection information based on the correlation matrix Cxx are mutually different. When it is determined that they match or are similar (YES in step ST32), the latest target detection information and the target detection information ahead are detected based on an incoming wave (signal wave) reflected by the same target object. Information can be expected. In particular, if the previous target detection information is information detected at a time when there is no large difference in time from the current time k, the physical position and relative position of the target object detected as the latest target detection information The velocity is a value that is substantially close to the physical position and relative velocity of the target object detected as the target detection information, and the eigenvectors vc (0) to vc (n−1) are the eigenvectors vb ( 0) to vb (n-1).

If the eigenvectors vb (0) to vb (n-1) and the previous eigenvectors vc (0) to vc (n-1) are vectors having completely the same elements, the similarity transformation matrix 、 It is a diagonal matrix having diagonal elements. On the other hand, if the elements of the eigenvectors vb (0) to vb (n-1) and the elements of the previous eigenvectors vc (0) to vc (n-1) have values close to each other, the similarity transformation matrix Γ It can be assumed that the matrix is close to a diagonal matrix. This is the same as the state where the similarity transformation matrix T has converged to some extent when the eigenvalue is calculated by an iterative operation based on the eigenvalue decomposition algorithm. For this reason, when eigenvalues are calculated from the similarity transformation matrix 反復 by iterative calculation, it can be expected that the similarity transformation matrix T can be converged with a smaller number of iterations than in the case where eigenvalues are calculated from the correlation matrix Cxx.

Further, even if the determination result of step ST32 is incorrect in the processing of the comparison search unit 51 and the conversion matrix B is generated from the previous target detection information on a physically different target object, the similarity Since the eigenvalue of the transformation matrix Γ and the eigenvalue of the correlation matrix Cxx theoretically match, the output result finally obtained does not substantially change.

By the way, the smaller the time difference between the time k at which the latest target detection information is detected and the time at which the target detection information used for calculating the transformation matrix B is detected, the more the latest target detection information and the time It can be expected that the correlation with the previous target detection information is high. However, even when the time difference is large, when the similarity between the pair of the distance and relative speed of the latest target detection information and the pair of the distance and relative speed of the previous target detection information is high, The comparison search unit 51 may determine that the latest target detection information and the preceding target detection information are similar to each other (YES in steps ST31 and ST32 in FIG. 8). As a result, when the intensity of the arriving wave (signal wave) from the target object is weak, or the comparison search unit 51 cannot find the previous target detection information detected at the previous time k-1. Even in this case, there is an advantage that the transformation matrix B can be generated.

The information storage unit 34 may store not only the target information generated by the target detection unit 32 and the arrival direction estimation unit 35 but also target information created by the user in advance. For example, when the arrangement state of a specific target object is likely to occur from the terrain information, the target storage information, eigenvalues and eigenvectors of the correlation matrix, which are calculated in advance from the arrangement of the specific target object, are converted into a database and stored in an information storage unit. 34. In this case, the arrival direction estimating unit 35 can generate the transformation matrix B from such database-based target information.

As described above, in the first embodiment, the arrival direction estimating unit 35 refers to the information storage unit 34 and refers to the information storage unit 34, and the eigenvector vb () corresponding to the target detection information that matches or is similar to the latest target detection information. 0) to vb (n-1) are obtained from the information storage unit 34, and an iterative operation based on the eigenvalue decomposition algorithm is performed using the eigenvectors vb (0) to vb (n-1) and the latest correlation matrix Cxx. Since it is executed (steps ST37 to ST39 and ST51 to ST59 in FIG. 10), when an iterative operation is executed without using the above eigenvectors vb (0) to vb (n-1) (steps ST34 and ST41 in FIG. 9). -ST48), the eigenvalue can be estimated with a smaller number of repetitions. Therefore, it is possible to suppress the calculation load required for eigenvalue estimation. Therefore, even if the number of receiving antenna elements constituting the antenna array 20 is large, the arrival direction can be estimated in a short calculation time without increasing the circuit scale of the signal processing circuit 30. In addition, it is possible to achieve both the reduction in size and weight of the signal processing circuit 30 and the reduction in cost.

Embodiment 2 FIG.
Next, a second embodiment according to the present invention will be described. FIG. 11 is a block diagram illustrating a schematic configuration of the radar device 1A according to the first embodiment of the present invention. As shown in FIG. 11, the configuration of the radar device 1A of the present embodiment is different from that of the first embodiment except that the signal processing circuit 30A is provided instead of the signal processing circuit 30A of the first embodiment. This is the same as the configuration of FIG. The configuration of the signal processing circuit 30A is the same as the arrival direction estimating unit 35 of the first embodiment, except that the signal processing circuit 30A has an arrival direction estimating unit 35A instead of the arrival direction estimating unit 35 of the first embodiment. As shown in FIG. 11, the configuration of the arrival direction estimating unit 35A is different from the eigenvalue calculating unit 53, the eigenvector calculating unit 54, and the arriving direction calculating unit 55 in the first embodiment in that the eigenvalue calculating unit 53A, the eigenvector calculating unit 54A The configuration is the same as that of the arrival direction estimating unit 35 of the first embodiment, except that it has an arrival direction calculating unit 55A.

As described above, in the first embodiment, the direction-of-arrival estimation unit 35 calculates the destination eigenvectors vb (0) to vb (n-1) acquired from the information storage unit 34 as shown in the above equation (20). Is used to calculate the transformation matrix B. On the other hand, as described later, the arrival direction estimating unit 35A according to the present embodiment includes h number of eigenvectors vb (0) to vb (n−1) obtained from the information storage unit 34, which are described later. The transformation matrix E is calculated using the vectors vb (0) to vb (h-1). Here, h is a positive integer smaller than n.

The entire operation of the radar apparatus 1A according to the second embodiment is the same as the entire operation of the radar apparatus 1 according to the first embodiment, except for an arrival direction estimation process. Hereinafter, the operation and configuration of the arrival direction estimation unit 35A according to the second embodiment will be described in detail with reference to FIGS. 8, 9 and 12. FIG. 12 is a flowchart illustrating an example of an arrival direction estimation process performed by the arrival direction estimation unit 35A. The flowchart of FIG. 12 is combined with the flowchart of FIG. 8 via the connector C1, and is combined with the flowchart of FIG. 9 via the connector C0. Steps ST51 to ST57 shown in FIG. 12 are the same as steps ST51 to ST57 shown in FIG.

Referring to FIG. 8, when the continuous detection number _{N k-i} is determined to be the threshold value _{N th} or more in step ST36 (YES in step ST36), eigenvalue calculation section 53A and the eigenvector calculator 54A is of the destination target By using the eigenvectors vb (0),..., Vb (n-1) (n is a positive integer) corresponding to the detection information and the correlation matrix Cxx, an iterative operation based on an eigenvalue decomposition algorithm is performed to obtain a correlation matrix. The eigenvalues and eigenvectors of Cxx are calculated (steps ST37A to ST39A, ST51 to ST57, ST58A, ST59A in FIG. 12).

Specifically, the eigenvalue calculation unit 53A forms a part of the previous eigenvectors vb (0),..., Vb (n−1) corresponding to the target detection information based on the following equation (23). The transformation matrix E is calculated based on the vectors vb (0),..., Vb (h−1) (step ST37A in FIG. 12).

{E = [vb (0), ..., vb (h-1)]} (23)

The correlation matrix Cxx is a square matrix having n rows and n columns, and the transformation matrix E is a matrix having the above eigenvectors vb (0) to vb (h-1) as column elements (column vectors). The number h of vectors may be the same value as the assumed number of incoming waves. Alternatively, when the above-mentioned manner k _z-th incoming waves Ni with large eigenvalues are estimated, may be set to the number of the vector h is k _z.

After performing step ST37A, eigenvalue calculating section 53A performs similarity conversion using conversion matrix E (step ST38A). That is, the intrinsic value calculation unit 53A, as shown in the following equation (24), multiplied from the right transformation matrix E into the correlation matrix Cxx, and multiplying the adjoint matrix E ^H of the transformation matrix E from the left to the correlation matrix Cxx Then, a similarity transformation matrix Ω is calculated.

Ω = E ^H · Cxx · E (24)

After execution of step ST38A, eigenvalue calculation section 53A converts similarity transformation matrix Ω to Hessenberg matrix A ⁽⁰⁾ using a transformation matrix, for example, based on the well-known Householder method (step ST36A). The processing contents of the following steps ST51 to ST57 are the same as the processing contents of steps ST51 to ST57 shown in FIG. In step ST57, the eigenvalue calculation unit 53A estimates the diagonal elements of the similarity transformation matrix T determined to have converged as the eigenvalues Λ (0),..., Λ (h−1) of the correlation matrix Cxx.

In step ST58A, the eigenvector calculator 54A uses the conversion was used matrix with the unitary matrix _{Q j} steps ST39A used in step ST53, eigenvector y (0) of the similarity transformation matrix Omega, ..., y ( h-1) is calculated.

After execution of step ST58A, the eigenvector calculation unit 54A uses the transformation matrix E used in step ST35A to convert the eigenvector y (κ) of the similarity transformation matrix Ω into the correlation matrix Cxx as shown in the following equation (25). An eigenvector v (κ) is calculated (step ST59A).

v (κ) = E · y (κ) (25)
Here, κ is an integer in the range of 0 to h−1.

Then, the direction-of-arrival calculation unit 55A calculates the correlation matrix Cxx, the eigenvalues Λ (0) to Λ (h−1) calculated in step ST57, and the eigenvectors v (0) to v (h−1) calculated in step ST59A. ) Is used to verify whether or not the preceding target detection information can be trusted based on a predetermined verification formula (step ST60). If the preceding target detection information is sufficiently reliable, the product of the correlation matrix Cxx and its eigenvector v (κ) should match the product of the eigenvector v (κ) and the eigenvalue Λ (κ). . Therefore, the arrival direction calculation unit 55A can perform verification using the following verification formula (26).

{S (κ) = {(κ) · v (κ) −Cxx · v (κ)} (26)

The arrival direction calculation unit 55A calculates the verification vector s (κ) on the left side of Expression (26), and determines whether the target detection information is reliable based on the norm of the verification vector s (κ). Can be performed (step ST60A). For example, the direction-of-arrival calculating unit 55A can trust the previous target detection information if the norm of the verification vector s (κ) is less than a predetermined threshold value for all of the eigenvalues Λ (0) to Λ (h−1). (YES in step ST60A), and if the norm of the verification vector s (κ) is equal to or greater than a predetermined threshold, it can be determined that the target detection information is not reliable (NO in step ST60A).

If it is determined that the preceding target detection information is reliable (YES in step ST60A), the arrival direction calculation section 55A executes steps ST49 and ST61 to ST66 shown in FIG. On the other hand, when it is determined that the previous target detection information is not reliable (NO in step ST60A), the arrival direction calculation unit 55A sets the value of the number _Nk of consecutive detections of the latest target detection information to zero. (Step ST60B). After that, the eigenvalue calculation unit 53A and the eigenvector calculation unit 54A execute steps ST34 and ST41 to ST48 shown in FIG. Thereafter, the direction-of-arrival calculation unit 55A executes steps ST49, ST61 to ST66 shown in FIG.

After the execution of step ST66, the information output unit 38 reads the target information Dc such as the distance to the target object, the relative speed of the target object, and the arrival direction from the information storage unit 34, similarly to step ST21 of FIG. Output to the outside. The target information Dc output to the outside is used, for example, for tracking processing in a subsequent processing device, or is used as information displayed on a display device. Thereafter, the control unit 39 returns the procedure of the radar process to the first step when determining to continue the radar process, and terminates the radar process when determining not to continue the radar process.

As described above, the eigenvalue calculation unit 53A of the present embodiment calculates the h vectors vb forming a part of the previous eigenvectors vb (0) to vb (n-1) acquired from the information storage unit. The transformation matrix E is calculated using (0) to vb (h-1) (step ST37A in FIG. 12). The technical meaning of the number h of vectors will be described below.

移動 At the same time, it is rare that two target objects that move at the same distance from each other at the same distance and that exist in different directions (angles) at the same time are generated at the same time. Assuming that the number of arriving waves Ni is usually at most about 2 or 3, when the platform on which the radar device 1A is mounted is moving forward, a fixed object on the ground as viewed from the platform is considered. Is different depending on the line of sight. Further, even if a plurality of target objects are present at the same distance from each other, the relative speed at which the target objects are observed differs depending on the direction in which the target objects exist. Therefore, it is considered that there is a low possibility that three or more target objects exist at positions separated by the same distance and move at the same relative speed.

のうち In addition, the eigenvalue having a value equal to or higher than the noise level among the eigenvalues λ (0) to λ (n−1) of the correlation matrix Cxx corresponds to a value proportional to the power of the reflected wave from the specific target object. For this reason, among the n eigenvalues λ (0) to λ (n−1) of the correlation matrix Cxx, the eigenvalues λ (Ni + 1) to λ (n−1) whose magnitude is equal to or smaller than the (Ni + 1) th are noise level. It can be expected that the value will be a considerable value or a value of almost 0 caused by a calculation error.

In addition, in step ST30, the rank of the correlation matrix Rxx before the spatial averaging process is performed is 1 or less, and it is generally considered that the correlation matrix Rxx has at most one non-zero eigenvalue. The rank of the correlation matrix Cxx is also about the spatial average score Q. Therefore, even when the spatial average score Q is smaller than the size n of the correlation matrix Cxx, the correlation matrix Cxx is degenerate, and the eigenvalues λ (Q + 1) to λ (n−1) whose size is the (Q + 1) th or less are used. ) Can be expected to be a value of almost 0 caused by an arithmetic error. Therefore, it can be inferred that the correlation matrix Cxx is in a degenerate state or a state having an almost zero eigenvalue close to the degenerated state.

On the other hand, applying the transformation matrices E and E ^H to the correlation matrix Cxx as shown in Expression (24) corresponds to the projection from the n-dimensional vector space to the h-dimensional vector space. Since the similarity transformation matrix Ω is a matrix having a size of h, the calculation load of the similarity transformation matrix Ω can be significantly reduced as compared with the case of calculating the eigenvalue of the similarity transformation matrix の having a size of n. it can. In the iterative operation of the eigenvalue decomposition algorithm, matrix multiplication is frequently used. Generally, in matrix multiplication, the amount of calculation increases in the order of the cube of the size, so that the amount of calculation can be reduced even if the size of the matrix is reduced by one.

As described above, the DOA estimating method according to the present embodiment approximately calculates the eigenvalues and eigenvectors of the correlation matrix Cxx with a smaller amount of calculation than the DOA estimating method according to the first embodiment. be able to. Therefore, even if the number of receiving antenna elements constituting the antenna array 20 is large, the arrival direction can be estimated in a short calculation time without increasing the circuit scale of the signal processing circuit 30. Therefore, it is possible to achieve both the reduction in size and weight of the signal processing circuit 30 and the reduction in cost.

When the correlation matrix Cxx is calculated based on the vectors vb (0) to vb (h-1) forming a part of the eigenvectors vb (0) to vb (n-1), the arrival direction calculation unit 55A is based on a predetermined verification equation using the correlation matrix Cxx, the estimated eigenvalues λ (0) to λ (n-1) and the eigenvectors vc (0) to vc (n-1). Then, it verifies whether or not the target detection information can be trusted (steps ST60 and ST60A). When the arrival direction calculation unit 55A determines that the previous target detection information is unreliable, it does not use the estimated eigenvalues and eigenvectors, so that it is possible to avoid an unreliable arrival direction estimation.

Note that the hardware configuration of the signal processing circuit 30A of the second embodiment may be realized by a processor having a semiconductor integrated circuit such as a DSP, an ASIC, or an FPGA, as in the signal processing circuit 30 of the first embodiment. Alternatively, the hardware configuration of the signal processing circuit 30 </ b> A is realized by a processor including a processing unit such as a CPU or a GPU that executes program codes (instructions) of software or firmware for signal processing read from a memory. Is also good. The hardware configuration of the signal processing circuit 30A can be realized by a processor having a combination of the semiconductor integrated circuit and the arithmetic device.

As described above, various embodiments according to the present invention have been described with reference to the drawings. However, these embodiments are merely examples of the present invention, and various embodiments other than these embodiments can be adopted. For example, in the first and second embodiments, the receiving antenna elements 20 _0-20 ₃ number the number of the reception channel but are both 4, but is not limited thereto. The receiving arrangement of the antenna elements 20 _0-20 ₃ is also not limited to the arrangement shown in FIG.

Also, in the first and second embodiments, the iterative operation based on the eigenvalue decomposition algorithm using the Householder method and the QR decomposition method is executed, but the present invention is not limited to this. Iterative operations based on eigenvalue decomposition algorithms other than the eigenvalue decomposition algorithm described above can be used.

において Within the scope of the present invention, any combination of the first and second embodiments, modification of any component of each embodiment, or omission of any component of each embodiment is possible.

The signal processing circuit, the radar device, the signal processing method, and the signal processing program according to the present invention can estimate the direction of arrival of an incoming wave from a single or a plurality of target objects with high resolution using an antenna array. The present invention can be applied to a radar device mounted on a moving body such as a vehicle. Further, the signal processing circuit, the radar apparatus, the signal processing method, and the signal processing program according to the present invention can estimate the direction of arrival with high resolution with a low calculation load, and therefore can be applied to a small-sized radar device that operates with low power consumption. Applicable.

1,1A radar apparatus, 10 transmission antenna 11 transmitting circuit, 12 a signal generator, 13 a distributor, 14 a transmission amplifier, 20 an antenna _array, 20 0-20 ₃ receive antenna elements, 21 receiving _circuit, 21 0-21 ₃ received _vessel, 22 0-22 ₃ receiving _amplifier, 23 0 to 23 ₃ mixer _{_{circuit, 24 0 ~ 24 3 A /}} D converter (ADC), 30, 30A signal processing circuit, 31 domain transforming section, 32 target detection unit, 34 Information storage unit, 35, 35A arrival direction estimation unit, 38 information output unit, 39 control unit, 41 first preprocessing unit, 42 first orthogonal transformation unit, 43 second preprocessing unit, 44 second orthogonal transformation unit, 51 Comparison search section, 52 correlation calculation section, 53, 53A eigenvalue calculation section, 54, 54A eigenvector calculation section, 55, 55A arrival direction calculation section, 71 processor, 72 memory, 73 input / output Interface section, 74 signal path.

Claims

An antenna array including a plurality of receiving antenna elements for receiving an incoming wave reflected by a target object, and a radar device including a receiving circuit configured to generate a plurality of received signals based on outputs of the plurality of receiving antenna elements, respectively. A signal processing circuit,
A domain conversion unit configured to convert the plurality of reception signals into a plurality of frequency domain signals,
A target detection unit that detects target detection information based on the plurality of frequency domain signals,
Calculating a correlation matrix based on the plurality of frequency domain signals, an arrival direction estimating unit that estimates an arrival direction of one or more arriving waves using a set of eigenvalues and eigenvectors of the correlation matrix,
An information storage unit in which at least one pair of a previously detected target detection information detected in the past and a plurality of previously eigenvectors estimated in the past is stored,
When the target detection unit detects the latest target detection information based on the plurality of frequency domain signals, the arrival direction estimation unit calculates the latest correlation matrix based on the plurality of frequency domain signals, and calculates the latest correlation matrix. Obtain a plurality of previous eigenvectors corresponding to the target detection information that matches or is similar to the target detection information from the information storage unit, using the obtained plurality of previous eigenvectors and the latest correlation matrix. A signal processing circuit which estimates an eigenvalue and an eigenvector set of the latest correlation matrix by executing an iterative operation based on a predetermined eigenvalue decomposition algorithm.
2. The signal processing circuit according to claim 1, wherein the direction-of-arrival estimating unit calculates a transformation matrix using the plurality of eigenvectors obtained from the information storage unit, and uses the transformation matrix to calculate the transformation matrix. A signal processing circuit, wherein a similarity transformation matrix is calculated by performing similarity transformation on a latest correlation matrix, and the iterative operation is performed using the similarity transformation matrix.
3. The signal processing circuit according to claim 2, wherein the direction-of-arrival estimating unit calculates the transformation matrix based on a plurality of vectors forming a part of the plurality of eigenvectors obtained from the information storage unit. A signal processing circuit.
The signal processing circuit according to claim 3, wherein the arrival direction estimating unit uses the latest correlation matrix and a set of the estimated eigenvalue and eigenvector based on a predetermined verification equation, It is determined whether or not the previous target detection information that matches or is similar to the latest target detection information can be trusted. If it is determined that the previous target detection information cannot be trusted, the estimated A signal processing circuit that does not estimate a direction of arrival using a set of eigenvalues and eigenvectors.
5. The signal processing circuit according to claim 4, wherein the verification formula is a product of the product of the latest correlation matrix and the estimated eigenvector, and a product of the estimated eigenvalue and the estimated eigenvector. A signal processing circuit having an equation for calculating a difference between the two.
(6) The signal processing circuit according to any one of (1) to (5), wherein the predetermined eigenvalue decomposition algorithm includes an algorithm based on a QR decomposition method.
The signal processing circuit according to any one of claims 1 to 5, wherein
The arrival direction estimation unit includes a comparison search unit that calculates a similarity or a difference between the latest target detection information and the previous target detection information stored in the information storage unit.
The comparison search unit determines whether the previous target detection information stored in the information storage unit matches or is similar to the latest target detection information based on the similarity or the difference.
A signal processing circuit characterized by the above-mentioned.
8. The signal processing circuit according to claim 7, wherein the comparison and search unit is configured to include a set of a distance and a relative speed included in the latest target detection information and a target search information stored in the information storage unit. 9. A signal processing circuit for calculating the similarity or the difference between a pair of a distance and a relative speed included in the above.
The signal processing circuit according to any one of claims 1 to 5, wherein
The radar device includes:
A signal generator that generates a chirp signal having a transmission frequency that periodically changes with time as the frequency modulation signal,
A distributor that distributes the frequency-modulated signal into a transmission signal and a local signal;
A signal processing circuit that mixes signals output in parallel from the plurality of reception antenna elements with the local signal to generate a plurality of beat signals as the plurality of reception signals.
The signal processing circuit according to claim 9,
The signal generator generates the chirp signal by continuously generating M frequency-modulated waves (M is an integer of 2 or more) within each frame period,
The area conversion unit converts the plurality of reception signals into the plurality of frequency modulation signals for each frame period,
The direction-of-arrival estimating unit calculates, based on the previous target detection information detected at a time earlier than the time at which the latest target detection information was detected by a time within a range from 1 frame period to 9 frame periods, Finding out the previous target detection information that matches or is similar to the latest target detection information,
A signal processing circuit characterized by the above-mentioned.
6. The signal processing circuit according to claim 1, wherein the arrival direction estimating unit is configured to set a detection frequency of a target object specified by the latest target detection information to a predetermined threshold value. 7. A signal processing circuit which performs the iterative operation at the time described above.
A signal processing circuit according to any one of claims 1 to 11,
Said antenna array;
A radar device comprising the receiving circuit.
An antenna array including a plurality of receiving antenna elements for receiving an incoming wave reflected by a target object, and a receiving apparatus that generates a plurality of received signals based on outputs of the plurality of receiving antenna elements, respectively. A signal processing method to be performed,
Converting the plurality of received signals into a plurality of frequency domain signals,
Detecting the latest target detection information based on the plurality of frequency domain signals,
Calculating the latest correlation matrix using the plurality of frequency domain signals,
By referring to an information storage unit in which at least one set of previously detected previous target detection information and a plurality of previously estimated previous eigenvectors is stored, it matches or is similar to the latest target detection information. Acquiring a plurality of previous eigenvectors corresponding to the target detection information to be performed from the information storage unit,
Estimating a set of eigenvalues and eigenvectors of the latest correlation matrix by executing an iterative operation based on a predetermined eigenvalue decomposition algorithm using the obtained plurality of previous eigenvectors and the latest correlation matrix,
Estimating the direction of arrival of one or more arriving waves using the set of the estimated eigenvalues and eigenvectors.
An antenna array including a plurality of receiving antenna elements that receive an incoming wave reflected by a target object, a receiving circuit that generates a plurality of received signals based on outputs of the plurality of receiving antenna elements, and a signal processing program are stored. And a processor that executes the signal processing program read from the memory, the signal processing program includes:
Converting the plurality of received signals into a plurality of frequency domain signals,
Detecting the latest target detection information based on the plurality of frequency domain signals,
Calculating the latest correlation matrix using the plurality of frequency domain signals,
By referring to an information storage unit in which at least one set of previously detected previous target detection information and a plurality of previously estimated previous eigenvectors is stored, it matches or is similar to the latest target detection information. Acquiring a plurality of previous eigenvectors corresponding to the target detection information to be performed from the information storage unit,
Estimating a set of eigenvalues and eigenvectors of the latest correlation matrix by executing an iterative operation based on a predetermined eigenvalue decomposition algorithm using the obtained plurality of previous eigenvectors and the latest correlation matrix,
Estimating the direction of arrival of one or more incoming waves using the set of the estimated eigenvalues and eigenvectors.