GPU-based strong real-time agile radar data matrix transposition method

By adopting a strong real-time agile radar data matrix transposition method based on GPU in the radar system, using the coordinated work of CPU and GPU, the problems of slow data re-arrangement speed and low efficiency in the radar system are solved, and high-speed real-time data re-reloading is achieved Improved order and flexibility.

CN119001622APending Publication Date: 2024-11-22CNGC INST NO 206 OF CHINA ARMS IND GRP
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411101285.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The data re-arrangement speed in existing radar systems is slow, low efficiency and poor flexibility, and cannot meet the real-time and rapid R&D iteration requirements of modern radars.

Method used

Using a strong real-time variable radar data matrix transposition method based on GPU, the original beam data is sent to the GPU through the CPU for matrix transposition operation, the GPU memory access efficiency is improved by loading rows and columns, and the GPU is completed through the CPU. Logical control to achieve rapid data reordering.

Benefits of technology

It improves the speed and efficiency of data re-arrangement processing, realizes high-speed real-time re-arrangement processing of beam data, and meets the real-time and flexibility requirements of modern radar systems.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119001622A_ABST
    Figure CN119001622A_ABST
Patent Text Reader

Abstract

The invention discloses a strong real-time agile radar data matrix transposition method based on a GPU. The method comprises the steps that original beam data of a CPI read by a CPU from a DAM cache are acquired, one CPI comprises 128 PRIs, each PRI comprises 1088 distance points, and each distance point comprises 64 original beam data; performing zero padding operation on the original beam data to obtain original beam data after the zero padding operation, and storing the original beam data after the zero padding operation into a CPU cache; sending the original beam data after the zero padding operation to a GPU video memory; and receiving rearranged data corresponding to the original beam data after the zero padding operation sent by the GPU. The technical problems of low data rearrangement speed, low efficiency and poor flexibility in the prior art are solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of data rearrangement before pulse compression processing in a radar back-end, and in particular to a GPU-based strong real-time agile radar data matrix transposition method. Background Art

[0002] Data rearrangement is a process in which the radar rearranges the data according to beams after digital beamforming processing, and separates each beam to facilitate subsequent processing in sequence according to the beams. Data rearrangement is an essential processing step in connecting the front-end and back-end of radar signal processing. The speed of data rearrangement directly affects the real-time performance of the radar.

[0003] Many modern radar signal processing systems use the FPGA+DSP working architecture. Since data rearrangement involves a large amount of data and high real-time requirements, most radar signal processing will choose to perform data rearrangement and pulse compression on the Field Programmable Gate Array (FPGA) side.

[0004] With the development of radar signal processing technology, the central processing unit (CPU) has shown strong appeal in radar signal processing due to its flexibility, long-term mature applications and its easy development characteristics. Rearranging data to the back-end processing reduces the development difficulty and improves the development speed. However, the characteristics of the CPU determine that it is more suitable for processing complex logical operations and lacks real-time processing of large amounts of data. Using the CPU+GPU architecture for data rearrangement processing provides a better solution to solve the bottleneck caused by the CPU processing speed and bandwidth in high-speed system design and improve and enhance system performance.

[0005] Therefore, the use of CPU+GPU architecture for data rearrangement has great application prospects for large phased array radars. The current industry implementation method for data rearrangement is as follows: FPGA has fast processing speed and large data throughput, so the FPGA+DDR3 architecture is used for data rearrangement. However, due to the low efficiency of address jump reading and writing of the third-generation double-data-rate synchronous dynamic random access memory (Double-Data-Rate 3 Synchronous Dynamic Random Access Memory, referred to as DDR3), it can no longer meet the real-time requirements of modern radars. In addition, FPGA has the characteristics of low development efficiency, slow development speed, and poor flexibility, which makes it difficult to meet the current demand for rapid R&D and iteration of equipment. Therefore, the industry expects flexible and fast data rearrangement processing. Summary of the invention

[0006] The embodiment of the present invention provides a strong real-time agile radar data matrix transposition method based on GPU, so as to at least solve the technical problems of slow data rearrangement speed, low efficiency and poor flexibility in the prior art.

[0007] According to one aspect of an embodiment of the present invention, a GPU-based strong real-time agile radar data matrix transposition method is provided, which is applied to a CPU side, wherein the method may include: obtaining the original beam data of a CPI read by the CPU from a DAM cache, wherein a CPI contains 128 PRIs, each PRI contains 1088 distance points, and each distance point contains 64 original beam data; performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, and storing the original beam data after the zero-padding operation in a CPU cache; sending the original beam data after the zero-padding operation to a GPU video memory; and receiving rearranged data corresponding to the original beam data after the zero-padding operation sent from the GPU.

[0008] Optionally, before performing zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, the method also includes: taking the number of distance points of each PRI in a CPI; and determining the number of zero-padding for each PRI based on the number of distance points of each PRI.

[0009] Optionally, performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation includes: performing a zero-padding operation on the original beam data based on the number of zero-paddings of each PRI to obtain the original beam data after the zero-padding operation.

[0010] According to one aspect of an embodiment of the present invention, another GPU-based strong real-time agile radar data matrix transposition method is provided, which is applied to the GPU side, wherein the method may include: receiving original beam data after zero-padding operation sent from the CPU; performing a matrix transposition operation on the original beam data after the zero-padding operation to obtain rearranged data corresponding to the original beam data after the zero-padding operation, and storing the rearranged data in the GPU cache; sending the rearranged data in the GPU cache to the CPU cache.

[0011] Beneficial effects of the present invention:

[0012] The present invention proposes a strong real-time agile radar data matrix transposition method based on GPU. The CPU sends beam data and control parameters to be processed to the GPU. The GPU performs matrix transposition operation by loading row storage columns. The rearranged beam data is sent back to the CPU or is processed in the GPU for the next step. The technical problems of slow data rearrangement speed, low efficiency and poor flexibility in the prior art are solved. The method achieves the goal of using GPU to realize high-speed real-time rearrangement processing of beam data. The method of loading row storage columns is used to perform matrix transposition operation, which improves the access efficiency of GPU memory and the speed of rearrangement processing. The CPU completes the logical control of the GPU and achieves the technical effect of fast data rearrangement. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0014] Figure 1 is a flowchart of a GPU-based strong real-time agile radar data matrix transposition method according to an embodiment of the present invention;

[0015] Figure 2 is a schematic diagram of original beam data arrangement of a CPI according to an embodiment of the present invention;

[0016] Figure 3 is a schematic diagram of original beam data arrangement after a zero-padding operation according to an embodiment of the present invention;

[0017] Figure 4 is a flowchart of another GPU-based strong real-time agile radar data matrix transposition method according to an embodiment of the present invention;

[0018] Figure 5 is a schematic diagram of rearranged data arrangement corresponding to original beam data after a zero-padding operation according to an embodiment of the present invention;

[0019] Figure 6 is a schematic diagram of a GPU-based strong real-time agile radar data matrix transposition system according to an embodiment of the present invention;

[0020] Figure 7 4 is a schematic diagram of another GPU-based strong real-time agile radar data matrix transposition system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only embodiments of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] Example 1

[0024] According to an embodiment of the present invention, a GPU-based strong real-time agile radar data matrix transposition method is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system comprising at least one set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0025] Figure 1 is a flowchart of a GPU-based strong real-time agile radar data matrix transposition method according to an embodiment of the present invention, such as Figure 1 As shown, the method may include the following steps:

[0026] Step S101, obtaining the original beam data of a CPI read by the CPU from the DAM cache, wherein a CPI contains 128 PRIs, each PRI contains 1088 range points, and each range point contains 64 original beam data.

[0027] In the technical solution provided in the above step S101 of the present invention, when the central processing unit (CPU) reads the raw beam data of a radar coherent processing index (CPI) from the direct memory access (DMA) cache, according to an embodiment of the present invention, a schematic diagram of the arrangement of the raw beam data of the CPI is provided. Figure 2 FIG. 4 is a schematic diagram of an arrangement of original beam data of a CPI according to an embodiment of the present invention. Figure 2 As shown, a CPI contains 128 PRIs, each PRI contains 1088 distance points, and each distance point contains 64 original beam data. It should be noted that the present invention is implemented by using the CPU and GPU in a radar digital signal processing chassis. The main boards in the digital signal processing chassis are an Intel CPU and an NVIDIA GPU. The CPU is Xeon E5-2658, using x86 architecture, 64 bits, with 14 cores and 28 threads. The GPU model is Tesla P6 version, and the memory type is GDDR5.

[0028] Step S102, performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, and storing the original beam data after the zero-padding operation in a CPU cache.

[0029] In the technical solution provided in the above step S102 of the present invention, according to an embodiment of the present invention, a schematic diagram of the arrangement of original beam data after a zero-padding operation is provided. Figure 3 FIG. 4 is a schematic diagram of an arrangement of original beam data after a zero-padding operation according to an embodiment of the present invention. Figure 3 As shown, each of the 120 PRIs is supplemented from distance point 1089 to distance point 2048.

[0030] Step S103, sending the original beam data after the zero-padding operation to the GPU memory.

[0031] In the technical solution provided in the above step S103 of the present invention, the CPU sends the original beam data after the zero-padding operation to the GPU display memory.

[0032] Step S104, receiving rearranged data corresponding to the original beam data after the zero-padding operation sent by the GPU.

[0033] In the technical solution provided in the above step S104 of the present invention, the rearranged data corresponding to the original beam data after the zero-padding operation sent by the GPU is received.

[0034] The above method of this embodiment is further introduced below.

[0035] As an optional implementation method, in step S102, before performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, the method also includes: obtaining the number of distance points for each PRI in a CPI; and determining the number of zero-padding for each PRI based on the number of distance points for each PRI.

[0036] In this embodiment, each PRI contains 1088 distance points, and the FFT requires the number of points to be processed to be a power of 2. When 2 to the power of 10 is 1024, and 1024 is less than 1088, then 2 to the power of 11 should be taken, and 2 to the power of 11 is 2048. Therefore, zeros need to be padded after each PRI to reach 2048 points.

[0037] As an optional implementation method, step S102, performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, includes: performing a zero-padding operation on the original beam data based on the number of zero-paddings of each PRI to obtain the original beam data after the zero-padding operation.

[0038] In this embodiment, each PRI needs to be padded with zeros to 2048 points, so the zero padding operation starts from distance point 1089 and continues to distance point 2048.

[0039] In an embodiment of the present invention, the original beam data of a CPI is obtained by obtaining the original beam data read from the DAM cache by the CPU, wherein one CPI contains 128 PRIs, each PRI contains 1088 distance points, and each distance point contains 64 original beam data; a zero-padding operation is performed on the original beam data to obtain the original beam data after the zero-padding operation, and the original beam data after the zero-padding operation is stored in the CPU cache; the original beam data after the zero-padding operation is sent to the GPU video memory; and the rearranged data corresponding to the original beam data after the zero-padding operation sent by the GPU is received, so as to solve the technical problems of slow data rearrangement speed, low efficiency and poor flexibility in the prior art, and achieve the goal of using the GPU to realize high-speed and real-time rearrangement processing of the beam data, adopt the method of loading rows and storing columns to perform matrix transposition operations, improve the access efficiency of the GPU memory, improve the speed of the rearrangement processing, complete the logical control of the GPU by the CPU, and achieve the technical effect of fast data rearrangement.

[0040] According to an embodiment of the present invention, another GPU-based strong real-time agile radar data matrix transposition method is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system comprising at least one set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0041] Figure 4 FIG. 1 is a flowchart of another GPU-based strong real-time agile radar data matrix transposition method according to an embodiment of the present invention. Figure 4 As shown, applied to the GPU side, the method may include the following steps:

[0042] Step S401, receiving original beam data after zero padding operation sent by the CPU.

[0043] In the technical solution provided in the above step S201 of the present invention, a graphics processing unit (GPU) receives the original beam data after the zero-padding operation sent by the CPU.

[0044] Step S402 , performing a matrix transposition operation on the original beam data after the zero-padding operation to obtain rearranged data corresponding to the original beam data after the zero-padding operation, and storing the rearranged data in a GPU cache.

[0045] In the technical solution provided in the above step S402 of the present invention, according to an embodiment of the present invention, a schematic diagram of rearranged data arrangement corresponding to the original beam data after the zero-padding operation is provided. Figure 5 FIG. 4 is a schematic diagram of rearranged data arrangement corresponding to original beam data after a zero-padding operation according to an embodiment of the present invention. Figure 5 As shown, according to Figure 5 It can be seen that the arrangement of 128 PRIs corresponding to each beam is obtained. For example, beam 1 arranges data1 of different distance points together.

[0046] Step S403, sending the rearranged data in the GPU cache to the CPU cache.

[0047] In the technical solution provided in the above step S403 of the present invention, the GPU sends the rearranged data in the GPU cache to the CPU cache.

[0048] In an embodiment of the present invention, original beam data after zero padding operation is received from a CPU; a matrix transposition operation is performed on the original beam data after zero padding operation to obtain rearranged data corresponding to the original beam data after zero padding operation, and the rearranged data is stored in a GPU cache; and the rearranged data in the GPU cache is sent to the CPU cache, thereby solving the technical problems of slow data rearrangement speed, low efficiency and poor flexibility in the prior art, achieving high-speed and real-time rearrangement processing of beam data by using a GPU, adopting a method of loading rows and storing columns to perform matrix transposition operation, thereby improving the access efficiency of GPU memory, improving the speed of rearrangement processing, completing logical control of the GPU by the CPU, and achieving the technical effect of fast data rearrangement.

[0049] Example 2

[0050] According to an embodiment of the present invention, a GPU-based strong real-time agile radar data matrix transposition system is also provided. It should be noted that the GPU-based strong real-time agile radar data matrix transposition system can be used to execute the GPU-based strong real-time agile radar data matrix transposition method in Example 1, and is applied to the CPU side.

[0051] Figure 6 Schematic diagram of a GPU-based strong real-time agile radar data matrix transposition system according to an embodiment of the present invention. Figure 6 As shown, a GPU-based strong real-time agile radar data matrix transposition system may include: a first acquisition unit, a first processing unit, a first sending unit and a first receiving unit.

[0052] The first acquisition unit is used to acquire the original beam data of a CPI read by the CPU from the DAM cache, wherein a CPI contains 128 PRIs, each PRI contains 1088 distance points, and each distance point contains 64 original beam data.

[0053] The first processing unit is used to perform a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, and store the original beam data after the zero-padding operation in a CPU cache.

[0054] The first sending unit is used to send the original beam data after the zero-padding operation to the GPU display memory.

[0055] The first receiving unit is used to receive rearranged data corresponding to the original beam data after the zero-padding operation sent by the GPU.

[0056] Optionally, before performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, the system also includes: a second acquisition unit, used to obtain the number of distance points of each PRI in a CPI; and a determination unit, used to determine the number of zero-padding for each PRI based on the number of distance points of each PRI.

[0057] Optionally, the first processing unit may further include: an obtaining module, configured to perform a zero-padding operation on the original beam data based on the number of zero-padding of each PRI, to obtain the original beam data after the zero-padding operation.

[0058] In this embodiment, a first acquisition unit is used to acquire the original beam data of a CPI read by the CPU from the DAM cache, wherein a CPI contains 128 PRIs, each PRI contains 1088 distance points, and each distance point contains 64 original beam data; a first processing unit is used to perform a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, and store the original beam data after the zero-padding operation in the CPU cache; a first sending unit is used to send the original beam data after the zero-padding operation to the GPU video memory; a first receiving unit is used to receive the rearranged data corresponding to the original beam data after the zero-padding operation sent by the GPU, so as to solve the technical problems of slow data rearrangement speed, low efficiency and poor flexibility in the prior art, and achieve the high-speed and real-time rearrangement processing of beam data by using GPU, adopt the method of loading rows and storing columns to perform matrix transposition operation, improve the access efficiency of GPU memory, improve the speed of rearrangement processing, complete the logical control of GPU by CPU, and achieve the technical effect of fast data rearrangement.

[0059] According to an embodiment of the present invention, another GPU-based strong real-time agile radar data matrix transposition system is also provided. It should be noted that the other GPU-based strong real-time agile radar data matrix transposition system can be used to execute a GPU-based strong real-time agile radar data matrix transposition method in Example 1, and is applied to the GPU end.

[0060] Figure 7 FIG. 1 is a schematic diagram of another GPU-based strong real-time agile radar data matrix transposition system according to an embodiment of the present invention. Figure 7 As shown, another GPU-based strong real-time agile radar data matrix transposition system may include: a second receiving unit, a second processing unit and a second sending unit.

[0061] The second receiving unit is used to receive the original beam data after the zero-padding operation sent by the CPU.

[0062] The second processing unit is used to perform a matrix transposition operation on the original beam data after the zero-padding operation to obtain rearranged data corresponding to the original beam data after the zero-padding operation, and store the rearranged data in a GPU cache.

[0063] The second sending unit is used to send the rearranged data in the GPU cache to the CPU cache.

[0064] In this embodiment, a second receiving unit is used to receive the original beam data after the zero-padding operation sent from the CPU; a second processing unit is used to perform a matrix transposition operation on the original beam data after the zero-padding operation to obtain rearranged data corresponding to the original beam data after the zero-padding operation, and store the rearranged data in the GPU cache; a second sending unit is used to send the rearranged data in the GPU cache to the CPU cache, so as to solve the technical problems of slow data rearrangement speed, low efficiency and poor flexibility in the prior art, and achieve the goal of using GPU to realize high-speed and real-time rearrangement processing of beam data, adopt the method of loading rows and storing columns to perform matrix transposition operation, improve the access efficiency of GPU memory, improve the speed of rearrangement processing, complete the logical control of GPU by CPU, and achieve the technical effect of fast data rearrangement.

[0065] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0066] In the above embodiments of the present invention, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0067] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0068] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed over multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0069] In addition, each functional unit in each embodiment of the present invention may be integrated into a first processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0070] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A GPU-based strong real-time agile radar data matrix transposition method, applied to the CPU side, characterized in that: include: Get the original beam data of a CPI read by the CPU from the DAM cache, where a CPI contains 128 PRIs, each PRI contains 1088 range points, and each range point contains 64 original beam data; Performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, and storing the original beam data after the zero-padding operation in a CPU cache; Send the original beam data after zero padding to the GPU memory; The rearranged data corresponding to the original beam data after the zero-padding operation sent by the GPU is received.

2. The method according to claim 1, characterized in that Before performing a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, the method further includes: Get the number of distance points for each PRI in a CPI; Based on the number of distance points of each PRI, the number of zero padding for each PRI is determined.

3. The method according to claim 1, characterized in that Perform zero padding operation on the original beam data to obtain the original beam data after the zero padding operation, including: A zero-padding operation is performed on the original beam data based on the number of zero-padding of each PRI to obtain the original beam data after the zero-padding operation.

4. A GPU-based strong real-time agile radar data matrix transposition method, applied to the GPU side, characterized in that: include: Receive the original beam data after zero padding operation sent by the CPU; Perform a matrix transposition operation on the original beam data after the zero-padding operation to obtain rearranged data corresponding to the original beam data after the zero-padding operation, and store the rearranged data in the GPU cache; Send the rearranged data in the GPU cache to the CPU cache.

5. A GPU-based strong real-time agile radar data matrix transposition system, applied to the CPU side, characterized in that: include: A first acquisition unit is used to acquire the original beam data of a CPI read by the CPU from the DAM cache, wherein one CPI contains 128 PRIs, each PRI contains 1088 range points, and each range point contains 64 original beam data; A first processing unit is used to perform a zero-padding operation on the original beam data to obtain the original beam data after the zero-padding operation, and store the original beam data after the zero-padding operation in a CPU cache; A first sending unit, used for sending the original beam data after the zero-padding operation to the GPU memory; The first receiving unit is used to receive rearranged data corresponding to the original beam data after the zero-padding operation sent by the GPU.

6. A GPU-based strong real-time agile radar data matrix transposition system, applied to the GPU side, characterized in that: include: A second receiving unit is used to receive the original beam data after the zero-padding operation sent by the CPU; A second processing unit is used to perform a matrix transposition operation on the original beam data after the zero-padding operation to obtain rearranged data corresponding to the original beam data after the zero-padding operation, and store the rearranged data in a GPU cache; The second sending unit is used to send the rearranged data in the GPU cache to the CPU cache.

7. A processor, characterized in that: The processor is used to run a program, wherein the program executes the method according to any one of claims 1 to 4 when running.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 4.