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

Abstract

To reduce a possibility of erroneously detecting noise as a target without increasing a possibility of overlooking detection of the target.SOLUTION: A signal processor is configured while including a first map generation part 1 for generating a first range Doppler map as a range Doppler map in an observation area from reception signals of reflection waves from the observation area, and a second map generation part 2 for generating a noise base matrix as a matrix for showing a base of noise included in the first range Doppler map from the first range Doppler map generated by the first map generation part 1, and for generating a second range Doppler map by using the noise base matrix. The signal processor includes a target detection part 5 for obtaining a difference between the first range Doppler map generated by the first map generation part 1 and the second range Doppler map generated by the second map generation part 2, and detecting a target from the differential range Doppler map.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a signal processing device, a signal processing method, and a signal processing program. [Background technology]

[0002] There is a signal processing device that includes a target detection unit that acquires a received signal of a reflected wave from an observation area and detects a target from the received signal (see, for example, Patent Document 1). The target detection unit generates a range-Doppler map of the observation area from the received signal, compares the signal levels of each of the multiple components included in the range-Doppler map with a threshold, and detects targets based on the comparison results of each signal level with the threshold. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6952939 Summary of the Invention [Problem to be solved by the invention]

[0004] The signal processing device disclosed in Patent Document 1 cannot detect targets whose signal level is smaller than the threshold. Therefore, if the threshold is set to a large value, the detection unit is more likely to miss the detection of the target. On the other hand, if the threshold is set to a small value, there is a problem in that the detection unit is more likely to erroneously detect noise as a target.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a signal processing device and a signal processing method that can reduce the possibility of falsely detecting noise as a target without increasing the possibility of missing the target. [Means for solving the problem]

[0006] A signal processing device according to the present disclosure includes a first map generation unit that generates a first range-Doppler map that is a range-Doppler map of an observation area from a received signal of a reflected wave from the observation area, and a second map generation unit that generates a noise basis matrix that is a matrix indicating a basis of noise included in the first range-Doppler map from the first range-Doppler map generated by the first map generation unit and generates a second range-Doppler map using the noise basis matrix. The signal processing device also includes a target detection unit that calculates a difference between the first range-Doppler map generated by the first map generation unit and the second range-Doppler map generated by the second map generation unit and detects targets from the differential range-Doppler map. [Effects of the Invention]

[0007] According to the present disclosure, it is possible to reduce the possibility of falsely detecting noise as a target without increasing the possibility of missing the target. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a configuration diagram showing a signal processing device according to a first embodiment. [Figure 2] 1 is a hardware configuration diagram showing hardware of a signal processing device according to a first embodiment. [Figure 3] FIG. 10 is a hardware configuration diagram of a computer in the case where the signal processing device is realized by software, firmware, or the like. [Figure 4] 1 is a flowchart showing a signal processing method that is a processing procedure of the signal processing device. [Figure 5] FIG. 3 is an explanatory diagram showing an example of a first range-Doppler map X. [Figure 6] 3 is an explanatory diagram showing the target signal and noise included in the first range-Doppler map X. FIG. [Figure 7] FIG. 2 is an explanatory diagram illustrating an example of a noise basis matrix H. [Figure 8]FIG. 2 is an explanatory diagram showing an example of a weighting matrix W. [Figure 9] FIG. 10 is an explanatory diagram showing an example of a second range-Doppler map X'. [Figure 10] FIG. 10 is an explanatory diagram showing an example of displaying a landmark. [Figure 11] FIG. 10 is a configuration diagram showing a signal processing device according to a second embodiment. [Figure 12] FIG. 10 is a hardware configuration diagram showing hardware of a signal processing device according to a second embodiment. [Figure 13] FIG. 11 is an explanatory diagram showing the processing content of a signal processing device according to a third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] In order to explain the present disclosure in more detail, embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0010] Embodiment 1 FIG. 1 is a configuration diagram showing a signal processing device according to the first embodiment. FIG. 2 is a hardware configuration diagram showing hardware of the signal processing device according to the first embodiment. The signal processing device shown in FIG. 1 includes a first map generating unit 1, a second map generating unit 2, a target detecting unit 5, and a display processing unit 6.

[0011] The first map generating unit 1 is realized by, for example, a first map generating circuit 21 shown in FIG. The first map generator 1 acquires a received signal of a reflected wave from an observation area, for example, from a radar device (not shown). The received signal is, for example, a complex signal. The first map generator 1 generates a first range-Doppler map, which is a range-Doppler map of the observation area, from the received signal of the reflected wave from the observation area. The first map generating unit 1 outputs the first range-Doppler map to the second map generating unit 2 and the target detecting unit 5, respectively.

[0012] The second map generating unit 2 is realized by, for example, a second map generating circuit 22 shown in FIG. The second map generating unit 2 includes a matrix decomposition unit 3 and a matrix multiplication unit 4. The second map generator 2 acquires the first range-Doppler map from the first map generator 1. The second map generating unit 2 generates, from the first range-Doppler map generated by the first map generating unit, a noise basis matrix which is a matrix indicating the basis of noise included in the first range-Doppler map. The second map generator 2 generates a second range-Doppler map using the noise basis matrix, and outputs the second range-Doppler map to the target detector 5.

[0013] The matrix decomposition unit 3 acquires the first range-Doppler map from the first map generation unit 1. The matrix decomposition unit 3 decomposes the first range-Doppler map into a noise basis matrix and a weight matrix which is a matrix indicating the weight of the basis indicated by the noise basis matrix. The matrix decomposition unit 3 outputs the noise basis matrix and the weight matrix to the matrix multiplication unit 4. The matrix multiplication unit 4 calculates the product of the noise basis matrix and the weighting matrix, and outputs the calculation result of the product to the target detection unit 5 as a second range-Doppler map.

[0014] The target detection unit 5 is realized by, for example, a target detection circuit 23 shown in FIG. The target detection unit 5 acquires the first range-Doppler map from the first map generation unit 1 and acquires the second range-Doppler map from the second map generation unit 2. The target detection unit 5 obtains a range-Doppler map that is the difference between the first range-Doppler map and the second range-Doppler map. The target detection unit 5 detects a target from the range-Doppler map of the difference. Furthermore, the target detection unit 5 determines the distance from the antenna of the radar device that received the reflected wave to the target and the Doppler velocity of the target based on the range-Doppler map of the difference. The target detection unit 5 outputs to the display processing unit 6 detection information indicating the distance to the target and the Doppler velocity of the target.

[0015] The display processing unit 6 is realized by, for example, the display processing circuit 24 shown in FIG. The display processing unit 6 acquires the detection information from the target detection unit 5. The display processing unit 6 generates display data for displaying the target at a position that indicates the distance to the target and the Doppler velocity. The display processing unit 6 displays the target on a display (not shown) in accordance with the display data.

[0016] In Fig. 1, it is assumed that each of the components of the signal processing device, that is, the first map generation unit 1, the second map generation unit 2, the target detection unit 5, and the display processing unit 6, is realized by dedicated hardware as shown in Fig. 2. In other words, it is assumed that the signal processing device is realized by a first map generation circuit 21, a second map generation circuit 22, a target detection circuit 23, and a display processing circuit 24. Each of the first map generation circuit 21, the second map generation circuit 22, the target detection circuit 23 and the display processing circuit 24 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination thereof.

[0017] The components of the signal processing device are not limited to those realized by dedicated hardware, and the signal processing device may be realized by software, firmware, or a combination of software and firmware. Software or firmware is stored as a program in the memory of a computer. A computer refers to hardware that executes the program, such as a CPU (Central Processing Unit), central processing unit, processing unit, arithmetic unit, microprocessor, microcomputer, processor, or DSP (Digital Signal Processor).

[0018] FIG. 3 is a hardware configuration diagram of a computer in the case where the signal processing device is realized by software, firmware, or the like. When the signal processing device is realized by software, firmware, or the like, a signal processing program is stored in the memory 31 to cause a computer to execute a first map generation processing procedure, a second map generation processing procedure, a target detection processing procedure, and a display processing procedure as processing procedures in the first map generation unit 1, the second map generation unit 2, the target detection unit 5, and the display processing unit 6. Then, a processor 32 of the computer executes the signal processing program stored in the memory 31.

[0019] 2 shows an example in which each of the components of the signal processing device is realized by dedicated hardware, and Fig. 3 shows an example in which the signal processing device is realized by software, firmware, etc. However, this is merely an example, and some of the components in the signal processing device may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0020] Next, the operation of the signal processing device shown in FIG. 1 will be described. FIG. 4 is a flowchart showing a signal processing method, which is a processing procedure of the signal processing device. A transmitting antenna of the radar device (not shown) transmits radio waves toward an observation area where a target may exist. The reflected waves, which are radio waves reflected by the observation area, are received by a receiving antenna of the radar device, and the radar device outputs a received signal of the reflected waves to a signal processing device. Here, the radar device is assumed to transmit and receive radio waves, but this is merely an example, and it may be configured such that, for example, radio waves are transmitted from an antenna at a facility toward an observation area, and then an antenna at a facility other than the one at that facility receives the reflected waves from the observation area.

[0021] The first map generating unit 1 acquires, for example, from a radar device, a received signal of a reflected wave from an observation area. The first map generating unit 1 generates a first range-Doppler map X as shown in FIG. 5 as a range-Doppler map of the observation area from the received signal of the reflected wave from the observation area (step ST1 in FIG. 4). The first map generating unit 1 outputs the first range-Doppler map to the second map generating unit 2 and the target detecting unit 5, respectively.

[0022] FIG. 5 is an explanatory diagram showing an example of the first range-Doppler map X. As shown in FIG. 5 is expressed as a matrix with m rows and n columns, where m and n are integers of 2 or greater. The elements of the matrix representing the first range-Doppler map X are x 11 ,x 12 ,···,x 1n ,x 21 ,x 22 ,···,x 2n ,···,x m1 ,x m2 ,···,x mn It is expressed as follows. The row direction of the matrix representing the first range-Doppler map X corresponds to the distance from the receiving antenna of the radar device to the target, and the column direction of the matrix corresponds to the Doppler velocity of the target. As shown in Fig. 6, the first range-Doppler map X contains noise in addition to the target signal. In the example of Fig. 6, the noise is represented as sea surface clutter and ionospheric clutter. The "noise" shown in Fig. 5 is noise other than these clutters. FIG. 6 is an explanatory diagram showing the target signal and noise included in the first range-Doppler map X. As shown in FIG.

[0023] The second map generator 2 acquires the first range-Doppler map X from the first map generator 1. The second map generator 2 generates, from the first range-Doppler map X, a noise basis matrix H, which is a matrix indicating the basis of the noise included in the first range-Doppler map. The second map generator 2 generates a second range-Doppler map X' using the noise basis matrix H (step ST2 in FIG. 4). The second map generating unit 2 outputs the second range-Doppler map X′ to the target detecting unit 5. The process of generating the second range-Doppler map X' by the second map generating unit 2 will now be described in detail.

[0024] The matrix decomposition unit 3 decomposes the first range-Doppler map X into a noise basis matrix H and a weighting matrix W by a matrix decomposition method. The matrix decomposition unit 3 can use NMF (Nonnegative Matrix Factorization) as a matrix decomposition method. However, the matrix decomposition method is not limited to NMF, and the matrix decomposition unit 3 can use, for example, SVD (Singular Value Decomposition), LU decomposition, Cholesky decomposition, QR decomposition, Eigendecomposition, or Polar decomposition. Here, the first range-Doppler map X is assumed to be a two-dimensional matrix, and the matrix decomposition unit 3 decomposes the two-dimensional matrix into a noise basis matrix H and a weighting matrix W. However, this is merely an example, and the first range-Doppler map X may be a tensor of a higher order than the two-dimensional matrix, and the matrix decomposition unit 3 may decompose the higher-order tensor into the noise basis matrix H and the weighting matrix W. The matrix decomposition unit 3 may use, for example, Tucker decomposition, Canonical Polyadic decomposition (CP decomposition), or tensor train decomposition as a higher-order tensor decomposition method. The matrix decomposition unit 3 outputs the noise basis matrix H and the weighting matrix W to the matrix multiplication unit 4 .

[0025] The noise basis matrix H is the noise basis h k where k is an integer equal to or greater than 1. The value of k is set to a value equal to or less than the number of types of noise that may be included in the first range-Doppler map X. The noise appearing in the first range-Doppler map X may include, for example, sea surface clutter, ionospheric clutter, and noise N other than these clutters. A and noise other than these clutters N B If it is known that there are four types of noise, for example, k is set to 4. However, if the noise N B is set to, for example, k=3 if it is known that the signal level is low noise, which has very little effect on target detection. Here, an example is shown in which k is set to 4 or k to 3. As a method for setting the value of k, for example, the following two methods can be used. ·Method(1) A curve is drawn with the value of k on the horizontal axis and the reconstruction error (value of |X-X'|) on the vertical axis. In the region where the value of k is small, this curve is drawn so that the reconstruction error (value of |X-X'|) rapidly attenuates as the value of k increases. Then, when the value of k becomes larger than a certain boundary value, the curve is drawn so that the decline in the reconstruction error (value of |X-X'|) becomes more gradual. In method (1), the boundary value is set to the value of k. When the value of k is small, the expressive power of the noise basis matrix H is poor, so normal data cannot be correctly restored, resulting in a large restoration error. As the value of k increases, the expressive power of the noise basis matrix H becomes increasingly rich, resulting in a small restoration error. However, when the value of k exceeds the boundary value, the expressive power of the noise basis matrix H becomes too high, resulting in a situation where even abnormal data is restored. This is known as overfitting. By setting the boundary value to the value of k, the restoration error can be reduced within a range that does not result in overfitting. Any method may be used to detect the boundary value, but examples of methods that can be used to detect the boundary value include the elbow method and the silhouette analysis method. ·Method(2) A binary classification index that quantifies the quality of target detection is determined in advance. Examples of binary classification indexes include the F1 score or ROC-AUC (Receiver Operating Characteristic-Area Under the Curve). Method (2) calculates the binary classification index using several pieces of data that include the detection target, and finds the value of k that best improves the binary classification index when the value of k is changed. The found value is then set as the value of k.

[0026] FIG. 7 is an explanatory diagram illustrating an example of the noise basis matrix H. As shown in FIG. The noise basis matrix H shown in Fig. 7 is expressed as a matrix with k rows and n columns, where k=4 in the example of Fig. 7. The elements of the noise basis matrix H are h 11 ,h 12 ,···,h 1n ,h 21 ,h 22 ,···,h 2n ,h31 ,h 32 ,···,h 3n ,h 41 ,h 42 ,···,h 4n It is expressed as follows. The dimension of the noise basis matrix H obtained by matrix decomposing the first range-Doppler map X is reduced compared to the dimension of the matrix representing the first range-Doppler map X.

[0027] The weighting matrix W is a matrix that indicates the weight of the basis indicated by the noise basis matrix H. FIG. 8 is an explanatory diagram showing an example of the weighting matrix W. The weighting matrix W shown in Fig. 8 is expressed as a matrix with m rows and k columns, where k=4 in the example of Fig. 8. The elements of the weight matrix W are 11 ,w 12 ,w 13 ,w 14 ,w 21 ,w 22 ,w 23 ,w 24 ,w 31 ,w 32 ,w 33 ,w 34 ,···,w m1 ,w m2 ,w m3 ,w m4 It is expressed as follows.

[0028] The relationship between the elements of the matrix representing the first range-Doppler map X, the elements of the noise basis matrix H, and the elements of the weighting matrix W is expressed by the following equation (1).

[0029]

number

[0030] In equation (2), the h1 vector represents, for example, the base of sea surface clutter, and the h2 vector represents, for example, the base of ionospheric clutter. The h3 vector represents a noise basis different from, for example, sea surface clutter and ionospheric clutter, and the h4 vector represents a noise basis different from, for example, sea surface clutter and ionospheric clutter, and different from the noise associated with the h3 vector. In the text of the specification, due to electronic filing, it is not possible to add the symbol "→" above the letters "h1" etc., so they are written as "h1 vector" etc. Usually, the ratio of the number of target signals included in the first range-Doppler map X to the number of noise signals included in the first range-Doppler map X is extremely low. Therefore, the possibility that the target basis will appear in the noise basis matrix H is extremely low. Equation (1) can be expressed as the following equation (3).

[0031] TIFF0007784938000002.tif21166

[0032] The matrix multiplication unit 4 acquires the noise basis matrix H and the weighting matrix W from the matrix decomposition unit 3 . The matrix multiplication unit 4 calculates the product HW of the noise basis matrix H and the weighting matrix W. The calculation result of the product HW is the second range-Doppler map X', where X'=HW. FIG. 9 is an explanatory diagram showing an example of the second range-Doppler map X'. The second range-Doppler map X' shown in FIG. 9 is expressed as a matrix with m rows and n columns. The elements of the matrix representing the second range-Doppler map X' are x 11 ',x 12 ',···,x 1n ',x 21 ',x 22 ',···,x 2n ',···,x m1 ',x m2 ',···,x mn It is represented as: The dimension of the noise basis matrix H is reduced compared to the dimension of the matrix representing the first range-Doppler map X, and the noise basis matrix H is a matrix that does not include a target basis. The second range-Doppler map X' is the calculation result of the product HW of the noise basis matrix H and the weighting matrix W, and therefore does not include a target signal. The matrix multiplication unit 4 outputs the second range-Doppler map X′ to the target detection unit 5 .

[0033] The target detection unit 5 acquires the first range-Doppler map X from the first map generation unit 1 and acquires the second range-Doppler map X′ from the second map generation unit 2. The target detection unit 5 obtains a range-Doppler map ΔX, which is the difference between the first range-Doppler map X and the second range-Doppler map X′, as shown in the following equation (4). ΔX=X-X' (4) The first range-Doppler map X contains the target signal and noise, and the second range-Doppler map X' contains the noise but not the target signal. Therefore, the differential range-Doppler map ΔX contains only the target signal. The target detection unit 5 detects a target from the differential range-Doppler map ΔX (step ST3 in FIG. 4).

[0034] As a method for detecting a target from the range-Doppler map ΔX, for example, there is a method for detecting a portion of the differential range-Doppler map ΔX where the signal level of the component contained in the range-Doppler map ΔX is equal to or higher than a first threshold value as the target. The target detection unit 5 may be configured to detect an area where the signal level is equal to or greater than a first threshold, but determine that the area is a target if the area of ​​the area is equal to or greater than a second threshold, and determine that the area is not a target if the area of ​​the area is less than the second threshold. The second threshold value is, for example, the minimum area of ​​the target. Specifically, the area where the number of matrix elements in the row direction is p and the number of matrix elements in the column direction is q is used as the second threshold value. p is an integer between 2 and m, and q is an integer between 2 and n. When an area having a size where the number of matrix elements in the row direction is p and the number of matrix elements in the column direction is q is used as the second threshold, the target detection unit 5 determines that the area is a target if the number of matrix elements in the row direction of the area whose signal level is equal to or greater than the first threshold is p or more and the number of matrix elements in the column direction of the area is q or more. The target detection unit 5 determines that a part is not a target if the number of matrix elements in the row direction of the part whose signal level is equal to or greater than the first threshold is less than p, or if the number of matrix elements in the column direction of the part is less than q. Here, if the number of matrix elements in the row direction of a part is less than p or the number of matrix elements in the column direction of a part is less than q, the target detection unit 5 determines that the part is not a target. Even if the number of matrix elements in the row direction of a part is less than p or the number of matrix elements in the column direction of a part is less than q, the target detection unit 5 may determine that the part is a target if the number of matrix elements in the column direction × the number of matrix elements in the column direction is greater than p × q.

[0035] The target detection unit 5 also determines the distance from the receiving antenna of the radar device to the target and the Doppler velocity of the target based on the differential range-Doppler map ΔX. The vertical axis of the range-Doppler map ΔX represents the distance to the target, and the horizontal axis of the range-Doppler map ΔX represents the Doppler velocity of the target. The target detection unit 5 outputs to the display processing unit 6 detection information indicating the distance to the target and the Doppler velocity of the target.

[0036] Here, the differential range-Doppler map ΔX is assumed to contain only the target signal. However, in addition to the target signal, a small amount of noise may remain in the differential range-Doppler map ΔX. Even if slight noise remains in addition to the target signal in the differential range-Doppler map ΔX, the target detection unit 5 determines that the portion where the signal level is equal to or greater than the first threshold is a target only if the area of ​​the portion is equal to or greater than the second threshold. Therefore, the possibility that the target detection unit 5 will erroneously detect the slight remaining noise as a target is extremely small.

[0037] The display processing unit 6 acquires the detection information from the target detection unit 5. The display processing unit 6 generates display data for displaying a target at a position that indicates the distance to the target and the Doppler velocity. The process of generating the display data itself is a known technique, so a detailed description will be omitted. As shown in FIG. 10, the display processing unit 6 displays the target on a display (not shown) in accordance with the display data. FIG. 10 is an explanatory diagram showing an example of displaying a target object. The vertical axis of FIG. 10 represents the distance to the target, and the horizontal axis of FIG. 10 represents the Doppler velocity of the target. In the example of FIG. 10, the landmark is highlighted by drawing a dashed line around it. Methods for highlighting a target include surrounding it with a dashed line, surrounding it with a square, coloring it, drawing an arrow pointing to the target, or making the rest of the area white.

[0038] In the above-described first embodiment, the signal processing device is configured to include a first map generation unit 1 that generates a first range-Doppler map, which is a range-Doppler map of the observation area, from a received signal of a reflected wave from the observation area, and a second map generation unit 2 that generates a noise basis matrix, which is a matrix indicating the basis of noise included in the first range-Doppler map, from the first range-Doppler map generated by the first map generation unit 1, and generates a second range-Doppler map using the noise basis matrix. The signal processing device also includes a target detection unit 5 that calculates the difference between the first range-Doppler map generated by the first map generation unit 1 and the second range-Doppler map generated by the second map generation unit 2, and detects targets from the differential range-Doppler map. Therefore, the signal processing device can reduce the possibility of erroneously detecting noise as a target without increasing the possibility of missing the target.

[0039] Embodiment 2 In the second embodiment, a signal processing device in which a second map generating unit 7 includes a learning model 8 will be described.

[0040] Fig. 11 is a configuration diagram showing a signal processing device according to embodiment 2. In Fig. 11, the same reference numerals as in Fig. 1 indicate the same or corresponding parts, and therefore description thereof will be omitted. Fig. 12 is a hardware configuration diagram showing hardware of a signal processing device according to embodiment 2. In Fig. 12, the same reference numerals as in Fig. 2 indicate the same or corresponding parts, and therefore description thereof will be omitted. The signal processing device shown in FIG. 11 includes a first map generating unit 1, a second map generating unit 7, a target detecting unit 5, and a display processing unit 6.

[0041] The second map generating unit 7 is realized by, for example, a second map generating circuit 25 shown in FIG. The second map generating unit 7 includes a learning model 8 . The second map generator 7 acquires the first range-Doppler map X from the first map generator 1. The second map generator 7 provides the first range-Doppler map X to the learning model 8 and obtains the second range-Doppler map X′ from the learning model 8. The second map generating unit 7 outputs the second range-Doppler map X′ to the target detecting unit 5.

[0042] The learning model 8 is an unsupervised learning model, and is realized by, for example, a neural network. As a neural network technique, for example, a non-convolutional autoencoder, a convolutional autoencoder, or a variational autoencoder can be used. When a first range-Doppler map X is given, the learning model 8 outputs a second range-Doppler map X'. The input data provided to the learning model 8 is a first range-Doppler map X. However, since the first range-Doppler map X is a two-dimensional matrix, it is converted into a one-dimensional matrix when provided to the learning model 8. Then, each of the multiple matrix elements constituting the one-dimensional matrix is ​​provided to each of the multiple input nodes of the neural network that constitutes the learning model 8. The neural network constituting the learning model 8 has multiple output nodes that output multiple matrix elements constituting a one-dimensional matrix corresponding to the second range-Doppler map X'. The one-dimensional matrix having the multiple matrix elements output from the learning model 8 is converted into a two-dimensional matrix. The two-dimensional matrix is ​​the second range-Doppler map X'. Here, each of the multiple matrix elements constituting the one-dimensional matrix is ​​provided to each of the multiple input nodes of the neural network constituting the learning model 8. However, this is merely an example, and the first range-Doppler map X may be regarded as an image of a two-dimensional matrix and provided to the neural network. In this case, for example, UNet, ENet, or BoxENet may be used as the neural network.

[0043] 11, it is assumed that each of the components of the signal processing device, that is, the first map generation unit 1, the second map generation unit 7, the target detection unit 5, and the display processing unit 6, is realized by dedicated hardware as shown in Fig. 12. That is, it is assumed that the signal processing device is realized by the first map generation circuit 21, the second map generation circuit 25, the target detection circuit 23, and the display processing circuit 24. Each of the first map generation circuit 21, the second map generation circuit 25, the target detection circuit 23 and the display processing circuit 24 may be, for example, a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

[0044] The components of the signal processing device are not limited to those realized by dedicated hardware, and the signal processing device may be realized by software, firmware, or a combination of software and firmware. When the signal processing device is realized by software, firmware, or the like, a signal processing program for causing a computer to execute the respective processing procedures in the first map generation unit 1, the second map generation unit 7, the target detection unit 5, and the display processing unit 6 is stored in memory 31 shown in Fig. 3. Then, a processor 32 shown in Fig. 3 executes the signal processing program stored in memory 31.

[0045] 12 shows an example in which each of the components of the signal processing device is realized by dedicated hardware, while Fig. 3 shows an example in which the signal processing device is realized by software, firmware, etc. However, this is merely an example, and some of the components in the signal processing device may be realized by dedicated hardware, and the remaining components may be realized by software, firmware, etc.

[0046] Next, the operation of the signal processing device shown in Fig. 11 will be described. Since the components other than the second map generating unit 7 are the same as those of the signal processing device shown in Fig. 1, only the operation of the second map generating unit 7 will be described here.

[0047] The second map generator 7 acquires the first range-Doppler map X from the first map generator 1. The second map generator 7 provides the first range-Doppler map X to the learning model 8. When a first range-Doppler map X is given, the learning model 8 outputs a second range-Doppler map X'. The second map generation unit 7 acquires the second range-Doppler map X′ from the learning model 8 and outputs the second range-Doppler map X′ to the target detection unit 5.

[0048] In the above-described second embodiment, the signal processing device shown in FIG. 11 is configured so that the second map generation unit 7 includes an unsupervised learning model 8, the first range-Doppler map generated by the first map generation unit 1 is provided to the learning model 8, a second range-Doppler map is obtained from the learning model 8, and the second range-Doppler map is output to the target detection unit 5. Therefore, similar to the signal processing device shown in FIG. 1, the signal processing device shown in FIG. 11 can reduce the possibility of erroneously detecting noise as a target without increasing the possibility of missing the detection of the target.

[0049] 11, the second map generation unit 7 includes a learning model 8. However, this is merely an example, and the second map generation unit 7 may include, as the learning model 8, a first learning model that outputs a noise basis matrix H and a weighting matrix W when a first range-Doppler map X is given, and a second learning model that outputs a second range-Doppler map X′ when a noise basis matrix H and a weighting matrix W are given. In this case, the second map generator 7 provides the first range-Doppler map X to the first learning model, and obtains the noise basis matrix H and the weighting matrix W from the first learning model. Then, the second map generating unit 7 provides each of the noise basis matrix H and the weighting matrix W to a second learning model, and obtains a second range-Doppler map X' from the second learning model.

[0050] Embodiment 3 In the signal processing device according to the first and second embodiments, the first map generating unit 1 acquires a received signal of a reflected wave at a certain reception time t from, for example, a radar device. In the third embodiment, a signal processing device will be described in which the first map generating unit 1 acquires, for example, from a radar device, received signals r(t) of a plurality of reflected waves having different reception times t.

[0051] FIG. 13 is an explanatory diagram illustrating the processing content of the signal processing device according to the third embodiment. As shown in FIG. 13, the first map generating unit 1 generates a first range-Doppler map X(t) from the reception signals r(t) of the reflected waves at the respective reception times t. The first map generating unit 1 outputs the first range-Doppler map X(t) at each reception time t to the second map generating unit 2 and the target detecting unit 5, respectively.

[0052] The second map generating unit 2 acquires from the first map generating unit 1 the first range-Doppler map X(t) for each reception time t. The second map generator 2 generates a noise basis matrix H from the first range-Doppler map X(t) at each reception time t. As shown in FIG. 13, the second map generator 2 generates a second range-Doppler map X'(t) using the noise basis matrix H at each reception time t. The second map generating unit 2 outputs the second range-Doppler map X′(t) for each reception time t to the target detecting unit 5.

[0053] The target detection unit 5 acquires a first range-Doppler map X(t) for each reception time t from the first map generation unit 1, and acquires a second range-Doppler map X'(t) for each reception time t from the second map generation unit 2. The target detection unit 5 obtains a range-Doppler map ΔX(t) that is the difference between the first range-Doppler map X(t) at each reception time t and the second range-Doppler map X'(t) at each reception time t, as shown in the following equation (5). ΔX(t)=X(t)-X'(t) (5) The first range-Doppler map X(t) contains the target signal and noise, and the second range-Doppler map X'(t) contains noise but not the target signal. Therefore, the differential range-Doppler map ΔX(t) contains only the target signal. If the second range-Doppler map X'(t) contains slight noise, the differential range-Doppler map ΔX(t) may contain slight noise in addition to the target signal. The target detection unit 5 detects a target from the differential range-Doppler map ΔX(t). Furthermore, the target detection unit 5 determines the distance from the receiving antenna of the radar device to the target and the Doppler velocity of the target at each reception time t based on the differential range-Doppler map ΔX(t). The target detection unit 5 outputs to the display processing unit 6 detection information indicating the distance to the target and the Doppler velocity of the target at each reception time t.

[0054] The display processing unit 6 acquires the detection information from the target detection unit 5 at each reception time t. The display processing unit 6 generates display data for displaying the target at a position that indicates the distance to the target and the Doppler velocity at each reception time t. The display processing unit 6 displays the target on a display (not shown) in accordance with the display data at each reception time t.

[0055] In the third embodiment described above, the target at each reception time t can be displayed on the display, so that the movement of the target can be tracked.

[0056] In addition, the present disclosure allows for free combination of the respective embodiments, modification of any of the components of the respective embodiments, or omission of any of the components of the respective embodiments. [Explanation of symbols]

[0057] 1 First map generation unit, 2 Second map generation unit, 3 Matrix decomposition unit, 4 Matrix multiplication unit, 5 Target detection unit, 6 Display processing unit, 7 Second map generation unit, 8 Learning model, 21 First map generation circuit, 22 Second map generation circuit, 23 Target detection circuit, 24 Display processing circuit, 25 Second map generation circuit, 31 Memory, 32 Processor.

Claims

1. a first map generator that generates a first range-Doppler map, which is a range-Doppler map of an observation area, from a received signal of a reflected wave from the observation area; a second map generation unit that generates, from the first range-Doppler map generated by the first map generation unit, a noise basis matrix that is a matrix indicating a basis of noise included in the first range-Doppler map, and generates a second range-Doppler map using the noise basis matrix; a target detection unit that calculates a difference between a first range-Doppler map generated by the first map generation unit and a second range-Doppler map generated by the second map generation unit, and detects the target from the range-Doppler map of the difference; A signal processing device comprising:

2. The second map generation unit a matrix decomposition unit that decomposes the first range-Doppler map generated by the first map generation unit into the noise basis matrix and a weight matrix that is a matrix indicating weights of the basis indicated by the noise basis matrix; 2. The signal processing device according to claim 1, further comprising a matrix multiplication unit that calculates the product of the noise basis matrix and the weighting matrix and outputs the calculated product to the target detection unit as the second range-Doppler map.

3. the second map generator comprises an unsupervised learning model; 2. The signal processing device according to claim 1, wherein the first range-Doppler map generated by the first map generation unit is provided to the learning model, the second range-Doppler map is obtained from the learning model, and the second range-Doppler map is output to the target detection unit.

4. The first map generation unit acquiring a plurality of received signals having different reception times as the received signals of the reflected waves, and generating a first range-Doppler map from each of the received signals; The second map generation unit generating the noise basis matrices from the first range-Doppler maps generated by the first map generators, and generating second range-Doppler maps using the respective noise basis matrices; The target detection unit 2. The signal processing device according to claim 1, wherein a difference between each of the first range-Doppler maps generated by the first map generating unit and each of the second range-Doppler maps generated by the second map generating unit is calculated, and the target is detected from each of the range-Doppler maps of the difference.

5. The target detection unit 2. The signal processing device according to claim 1, wherein a portion of the range-Doppler map of the difference where the signal level of a component included in the range-Doppler map is equal to or greater than a first threshold is detected, and if the area of ​​the portion is equal to or greater than a second threshold, the portion is determined to be the target, and if the area of ​​the portion is less than the second threshold, the portion is determined to not be the target.

6. The target detection unit 2. The signal processing device according to claim 1, wherein the target is detected from the range-Doppler map of the difference, and the distance from the antenna that received the reflected wave to the target and the Doppler velocity of the target are determined based on the range-Doppler map of the difference.

7. 7. The signal processing device according to claim 6, further comprising a display processing unit that generates display data for displaying the target identified by the target detection unit at a position that indicates the distance to the target and the Doppler velocity of the target.

8. a first map generator generating a first range-Doppler map of an observation area from a received signal of a reflected wave from the observation area; a second map generation unit generates, from the first range-Doppler map generated by the first map generation unit, a noise basis matrix that is a matrix indicating a basis of noise included in the first range-Doppler map, and generates a second range-Doppler map using the noise basis matrix; A target detection unit calculates a difference between the first range-Doppler map generated by the first map generation unit and the second range-Doppler map generated by the second map generation unit, and detects the target from the range-Doppler map of the difference. Signal processing methods.

9. a first map generation process for generating a first range-Doppler map, which is a range-Doppler map of an observation area, from a received signal of a reflected wave from the observation area; a second map generation process step of generating a noise basis matrix, which is a matrix indicating a basis of noise included in the first range-Doppler map generated in the first map generation process step, from the first range-Doppler map generated in the first map generation process step, and generating a second range-Doppler map using the noise basis matrix; a target detection processing procedure for calculating a difference between a first range-Doppler map generated in the first map generation processing procedure and a second range-Doppler map generated in the second map generation processing procedure, and detecting the target from the range-Doppler map of the difference.

Images