WO2024060592A1 - 基于深度学习的射频通信设备空间辐射测试系统及方法 - Google Patents

基于深度学习的射频通信设备空间辐射测试系统及方法 Download PDF

Info

Publication number
WO2024060592A1
WO2024060592A1 PCT/CN2023/087808 CN2023087808W WO2024060592A1 WO 2024060592 A1 WO2024060592 A1 WO 2024060592A1 CN 2023087808 W CN2023087808 W CN 2023087808W WO 2024060592 A1 WO2024060592 A1 WO 2024060592A1
Authority
WO
WIPO (PCT)
Prior art keywords
neural network
test
fully connected
radiation
radiation performance
Prior art date
Application number
PCT/CN2023/087808
Other languages
English (en)
French (fr)
Inventor
全智
顾一帆
毕宿志
Original Assignee
深圳市中承科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市中承科技有限公司 filed Critical 深圳市中承科技有限公司
Publication of WO2024060592A1 publication Critical patent/WO2024060592A1/zh

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/0082—Monitoring; Testing using service channels; using auxiliary channels
    • H04B17/0087—Monitoring; Testing using service channels; using auxiliary channels using auxiliary channels or channel simulators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04B—TRANSMISSION
    • H04B17/00—Monitoring; Testing
    • H04B17/10—Monitoring; Testing of transmitters
    • H04B17/15—Performance testing

Definitions

  • the present invention relates to the field of wireless communication technologies, and in particular to a system and method for testing spatial radiation of radio frequency communication equipment based on deep learning.
  • the spatial radiation performance of radio frequency transmitters and receivers is a key indicator of wireless communication equipment, and is mainly determined by the overall performance of the radio frequency front-end system, including the antenna.
  • Standards such as the international standards organization CTIA and 3GPP require that the radiated power and receiving sensitivity of the radio frequency communication system be tested in all directions of three-dimensional space in a specific microwave anechoic chamber.
  • the traditional radio frequency radiation (OTA, Over-the-Air) test system is limited by physical conditions such as the size of the microwave anechoic chamber, the number of antenna probes, the rotation accuracy of the turntable, and the test time, so its spatial resolution is always limited.
  • test time cost required for each test point of radiation power and receiving sensitivity is high.
  • traditional test methods can only obtain the radiation performance of the device under test at the test point, and lack effective and accurate models to infer the radiation performance of the device under test at untested points, which usually introduces large inference errors. Therefore, traditional test methods are highly restricted in terms of system complexity, test accuracy, and test time cost, and it is urgent to explore new test methods to solve the above bottleneck problems.
  • the technical problem to be solved by the present invention is to provide a deep learning-based radio frequency communication equipment space radiation testing system and method, which can make full use of the real data obtained by a limited number of test points in the radio frequency radiation testing system, and accurately infer that the radiation is not detected at any time.
  • the spatial radiation performance in the direction can thus significantly reduce the number of test points required for the system while ensuring the accuracy of the RF space radiation test system.
  • the first aspect of the present invention discloses a deep learning-based radio frequency communication equipment space radiation testing method, which is characterized in that the method includes: obtaining the actual radiation performance parameters of the tested equipment at a limited number of test points ; Minimizing the mean square error training based on the actual radiation performance parameters to generate a fully connected deep neural network model; deducing the radiation performance parameters of the device under test in each direction of space based on the fully connected deep neural network model.
  • obtaining the actual radiation performance parameters of the device under test at a limited number of test points includes: placing the device under test on a high-precision turntable and a controller, and using the high-precision turntable and controller to pair the placed device under test.
  • the measuring equipment performs all-round high-precision steering on the horizontal plane; the measured equipment is measured in all directions of the space through antenna probes evenly distributed in the microwave anechoic chamber. Upward actual radiation performance parameters.
  • the radiation performance parameters include radiation power parameters and reception sensitivity parameters.
  • the fully connected deep neural network model is implemented by: substituting actual radiation power parameters and actual reception sensitivity parameters of a limited number of test points as training samples into The stochastic gradient descent algorithm formula of the following formula:
  • FCDNN ⁇ (. ⁇ ⁇ ) represents the fully connected deep neural network model
  • ⁇ ⁇ is its parameter
  • the subscript ⁇ is used to distinguish the model for radiation power parameters or the model for reception sensitivity parameters
  • E[.] represents the model for The mathematical expectation of the random variable x
  • X represents the set of all test points
  • represents the number of all test points
  • y ⁇ (x) represents the radiation performance parameter of the device under test at any point on the three-dimensional sphere
  • p represents the polarization mode of the antenna probe during testing
  • ⁇ r represents the elevation angle of the spherical coordinate system
  • the method further includes: dynamically evaluating based on a preset accuracy threshold value to determine whether the fully connected deep neural network model needs to collect training samples again; if the average value in the verification set of the fully connected deep neural network model If the square error is less than the preset accuracy threshold, the fully connected deep neural network model needs to collect training samples again.
  • the method further includes: normalizing the actual radiation performance parameters of the device under test at a limited number of test points; minimizing the full connection depth based on the normalized actual radiation performance parameters.
  • Fully connected deep neural network model generated by neural mean square error training.
  • a deep learning-based radio frequency communication equipment space radiation testing system which is characterized in that the system includes: a multi-probe microwave anechoic chamber, a wireless communication testing instrument, a channel simulator and other equipment
  • the radio frequency radiation test system is used to obtain the actual radiation performance parameters of the device under test at a limited number of test points;
  • a fully connected deep neural network model is generated by minimizing the mean square error training based on the actual radiation performance parameters; a prediction module is used to deduce the radiation performance parameters of the device under test in each direction of space based on the fully connected deep neural network model.
  • the multi-probe microwave anechoic chamber includes: a high-precision turntable and a controller for performing all-round high-precision steering of the placed device under test on the horizontal plane; a microwave anechoic chamber; annular and evenly distributed in the microwave
  • the antenna probe in the darkroom is used to measure the actual radiation performance parameters of the device under test in all directions in space; the radio frequency switch box is used to switch the antenna probe; and the wireless communication is used to measure the radio frequency radiation performance of the device under test in the specified direction. test instrument.
  • the radiation performance parameters include radiation power parameters and receiving sensitivity parameters
  • the fully connected deep neural network model is implemented by substituting the actual radiation power parameters and actual receiving sensitivity parameters of a finite number of test points as training samples into the following stochastic gradient descent algorithm formula:
  • FCDNN ⁇ (. ⁇ ⁇ ) represents a fully connected deep neural network model
  • ⁇ ⁇ is its parameter
  • the subscript ⁇ is used to distinguish the model for the radiation power parameter or the model for the receiving sensitivity parameter
  • E[.] represents the mathematical expectation for the random variable x
  • X represents the set of all test points
  • represents the number of all test points
  • y ⁇ (x) represents the radiation performance parameter of the device under test at any point on the three-dimensional sphere, represents the radiation performance parameters of the device under test at point x calculated by the fully connected deep neural network
  • p represents the polarization mode of the antenna probe during the test
  • ⁇ r represents the elevation angle of the spherical coordinate system
  • the system further includes: a dynamic inspection model that dynamically evaluates and determines whether the fully connected deep neural network model needs to collect training samples again based on a preset accuracy threshold; if the fully connected deep neural network model If the mean square error in the verification set is less than the preset accuracy threshold, the fully connected deep neural network model needs to collect training samples again.
  • the system further includes: a normalization module for normalizing the actual radiation performance parameters of the device under test at a limited number of test points; based on the normalized actual radiation performance parameters
  • the fully connected deep neural network model is generated by minimizing the mean square error of the fully connected deep neural network.
  • the implementation of the present invention can assist in obtaining accurate radiation characteristic parameter data of the equipment under test through the independently developed multi-probe microwave anechoic chamber, and introduces deep learning methods to train a fully connected deep neural network (FCDNN) model using limited measurement data in three-dimensional space, thereby Estimate the radiation performance of the RF communication system under test in all directions in three-dimensional space.
  • FCDNN fully connected deep neural network
  • a dynamic test model accuracy is further proposed, and the number of training test points is gradually increased through the preset accuracy threshold until the model accuracy reaches the preset requirements. solution.
  • the deep learning-based test system proposed by the present invention only needs about 60% of the test points to accurately reconstruct the spatial radiation performance of the radio frequency communication equipment under test, verifying that this method Accuracy and efficiency of invention.
  • the system proposed by the present invention can also infer radiation performance parameters at any point in space, rather than being limited to test points.
  • the proposed system can not only significantly reduce the number of test points and system hardware complexity, but also accurately construct a radiation performance prediction model at any direction and angle in three-dimensional space, providing the industry with an efficient, accurate but low-cost space radiation test.
  • Technical solutions can not only significantly reduce the number of test points and system hardware complexity, but also accurately construct a radiation performance prediction model at any direction and angle in three-dimensional space, providing the industry with an efficient, accurate but low-cost space radiation test.
  • Figure 1 is a schematic diagram of a system for space radiation testing of radio frequency communication equipment based on deep learning disclosed in an embodiment of the present invention
  • FIG2 is a schematic diagram of a multi-probe microwave darkroom system of a deep learning-based radio frequency communication equipment space radiation test system disclosed in an embodiment of the present invention
  • Figure 3 is a schematic diagram of a spherical coordinate system of three-dimensional spatial direction and unit radius disclosed in an embodiment of the present invention
  • Figure 4 is a schematic diagram of the neural network model training process of a radio frequency communication equipment space radiation testing system based on deep learning disclosed in an embodiment of the present invention
  • Figure 5 is a schematic diagram of EIRP and EIS performance at a 30-degree sampling interval measured by the device under test of a deep learning-based radio frequency communication equipment space radiation testing system disclosed in an embodiment of the present invention
  • Figure 6 is a schematic diagram of EIRP and EIS performance at a 15-degree sampling interval measured by the device under test of a deep learning-based radio frequency communication equipment space radiation testing system disclosed in an embodiment of the present invention
  • Figure 7 is a schematic diagram of the space radiation performance of the radio frequency communication equipment under test inferred by an FCDNN model under the training set scale of a radio frequency communication equipment space radiation testing system based on deep learning disclosed in the embodiment of the present invention
  • Figure 8 is a schematic diagram of the space radiation performance of the radio frequency communication equipment under test inferred by another FCDNN model under the training set scale of the radio frequency communication equipment space radiation testing system based on deep learning disclosed in the embodiment of the present invention
  • Figure 9 is a schematic diagram of the space radiation performance of the radio frequency communication equipment under test inferred by the FCDNN model under another training set scale of the radio frequency communication equipment space radiation testing system based on deep learning disclosed in the embodiment of the present invention.
  • Figure 10 is a schematic flowchart of a deep learning-based radio frequency communication equipment space radiation testing method disclosed in an embodiment of the present invention.
  • Embodiments of the present invention disclose a deep learning-based radio frequency communication equipment space radiation testing system and method.
  • the self-developed multi-probe microwave anechoic chamber is used to assist in obtaining accurate radiation characteristic parameter data of the equipment under test, and the deep learning method is introduced to utilize
  • the limited measurement data in three-dimensional space trains a fully connected deep neural network (FCDNN) model to estimate the radiation performance of the tested radio frequency communication system in all directions in three-dimensional space.
  • FCDNN fully connected deep neural network
  • a dynamic test model accuracy is further proposed, and the number of training test points is gradually increased through the preset accuracy threshold until the model accuracy reaches the preset requirements. solution.
  • the deep learning-based test system proposed by the present invention only needs about 60% of the test points to accurately reconstruct the spatial radiation performance of the radio frequency communication equipment under test, verifying that this method Accuracy and efficiency of invention.
  • the system proposed by the present invention can also infer radiation performance parameters at any point in space, rather than being limited to test points.
  • the proposed system can not only significantly reduce the number of test points and the complexity of system hardware, but also accurately construct a radiation performance prediction model for any direction and angle in three-dimensional space, providing the industry with an efficient, accurate and low-cost space radiation testing technology solution.
  • FIG. 1 is a schematic diagram of a deep learning-based space radiation test system for radio frequency communication equipment disclosed in an embodiment of the present invention.
  • the deep learning-based space radiation test system for radio frequency communication equipment can be applied to a wireless communication system, and the application of the system is not limited in the embodiment of the present invention.
  • the deep learning-based space radiation test system for radio frequency communication equipment may include:
  • a larger microwave anechoic chamber and deploy more antenna probes to test RF radiation at more points in three-dimensional space. data to obtain better spatial resolution of radiofrequency radiation.
  • these methods will occupy larger site space and increase the complexity of the test system, thereby significantly increasing test time and test costs.
  • This type of method speeds up testing by sacrificing the spatial resolution and measurement accuracy of the radiation performance of the radio frequency communication system, and is suitable for scenarios that require rapid design verification.
  • it is proposed to use compact field technology to achieve far-field measurement of the spatial radiation performance of radio frequency communication equipment at close range.
  • this type of measurement method still cannot effectively infer the radiation performance of radio frequency communication equipment at untested points of the equipment under test. , will introduce higher errors.
  • the present invention utilizes a self-designed multi-probe microwave anechoic chamber to obtain limited measurement data, namely actual radiation performance parameters, in three-dimensional space, and trains a fully connected deep neural network FCDNN to obtain the radiation performance of the RF communication system under test in all directions in the entire three-dimensional space.
  • the multi-probe microwave anechoic chamber 1 is used to obtain the actual radiation performance parameters of the device under test at a limited number of test points.
  • the multi-probe microwave anechoic chamber 1 includes: a high-precision turntable and a controller for performing all-round high-precision steering of the placed equipment under test on the horizontal plane; a microwave anechoic chamber; annular and evenly distributed antenna probes in the microwave anechoic chamber. It is used to measure the actual radiation performance parameters of the device under test in all directions in space.
  • Figure 2 it is a schematic diagram of a multi-probe microwave anechoic chamber system in a specific application.
  • the system mainly consists of a broadband wireless communication test instrument (RF communication test instrument), a microwave anechoic chamber (Anechoic chamber), and an antenna probe. It consists of (Probe antennas), high-precision turntable and controller (Turn table and controller), and a personal computer with automatic test software (Automatic testequipment, ATE). Among them, in order to avoid the reflection and echo of radio frequency signals in space radiation testing, absorbing materials can be used inside the microwave anechoic chamber to create an ideal testing environment.
  • the device under test DUT
  • the turntable controller is used to make the device under test perform all-round high-precision steering on the horizontal plane.
  • multiple antenna probes are usually deployed in a darkroom in a uniformly distributed annular structure to measure the radiation performance of the radio frequency communication equipment under test in all directions in space, that is, the actual radiation performance parameters of the equipment under test.
  • These antenna probes are connected to the RF switch and space channel simulator, and are controlled by a broadband wireless communication tester to accurately measure the radiation performance of the device under test in different directions in space and in different wireless environments.
  • the broadband wireless communication tester and turntable controller are globally controlled through a personal computer, and the self-developed automatic test software provides a user interface to further realize efficient test methods, as well as visual data output and storage.
  • the radiation power of the radio frequency communication system in all directions of the three-dimensional space in the microwave anechoic chamber is a key indicator to measure its uplink transmission performance.
  • the reception sensitivity parameter can be used as a key indicator for its downlink transmission. Therefore, the actual radiation performance parameters include radiation power parameters and receiving sensitivity parameters.
  • the radiation power parameter represents the amount of radiation power of the antenna in each direction, which is defined as the product of the power of the antenna in a specific polarization mode and the absolute gain of the antenna in a given direction.
  • ( ⁇ r , ) can be expressed as a point in the unit radius spherical coordinate system, that is, a test point in a specific direction in three-dimensional space.
  • 0 ⁇ r ⁇ is the elevation angle of the spherical coordinate system, which can be adjusted by selecting different test antenna probes.
  • the device under test can be placed in the maximum transmit power state, the direction of the control turntable can be adjusted, and the antenna probes can adopt horizontal and vertical polarization methods respectively. Multiple deployed probes can be used to measure the device under test in various directions. The value of the radiation power parameter.
  • the equivalent isotropic radiated power parameter can be used as the actual radiated power parameter.
  • the test of the equivalent isotropic radiated power parameter EIRP gives the radiated power performance of the device under test in all directions in space, and The total radiated power reflects its overall radiated power situation and is defined as the integral of the radiated power parameter value of each point in space on the three-dimensional sphere. Since the number of test points that an actual system can carry is often limited by its hardware complexity, such as turntable accuracy, number of antenna probes, etc., the total radiated power is usually calculated using numerical integration.
  • the total radiation power TRP of the device under test can be expressed as:
  • the test of receiving sensitivity parameters can reflect the sensitivity of the receiving device under test in a specific direction in space when the transmitting antenna probe is in a certain polarization mode.
  • First place the device under test on the turntable and adjust the angle of the turntable.
  • BER downlink reception bit error rate
  • the received signal strength of the device under test that is, the minimum received signal strength that can be tolerated under a given bit error rate, represents its receiving sensitivity.
  • the receiving sensitivity of the device under test in each direction of the three-dimensional space is established.
  • the total omnidirectional sensitivity parameter TIS can be used as the receiving sensitivity parameter, which can be calculated by integrating the value of the receiving sensitivity EIS at each point on the three-dimensional sphere.
  • the total omnidirectional sensitivity parameter TIS of the device under test can be expressed as :
  • TIS the reciprocal of EIS is adopted for integration. This is because the sensitivity of the device under test in each direction of space is often very different. If the reciprocal is not adopted, the test points with poor sensitivity, that is, the term with the higher receiving sensitivity EIS value, will dominate the integral result in this equation, and will As a result, the integration ignores the impact of test points with better sensitivity on the overall sensitivity of the device under test.
  • FCDNN ⁇ (. ⁇ ⁇ ) represents the fully connected deep neural network model
  • ⁇ ⁇ is its parameter
  • the subscript ⁇ is used to distinguish the model for radiation power parameters or the model for reception sensitivity parameters
  • E[.] represents the model for The mathematical expectation of the random variable x
  • X represents the set of all test points
  • represents the number of all test points
  • y ⁇ (x) represents the radiation performance parameter of the device under test at any point on the three-dimensional sphere
  • p represents the polarization mode of the antenna probe during testing
  • ⁇ r represents the elevation angle of the spherical coordinate system
  • the stochastic gradient descent algorithm can optimize the neural network parameters of the fully connected deep neural network model in a supervised learning manner to help build a fully connected neural network model.
  • the prediction module 3 can directly calculate the radiation performance parameters of the device under test in all directions of space based on the fully connected deep neural network model. Therefore, based on a limited number of test points, it can Accurately reconstruct the spatial radiation performance of the radio frequency communication equipment under test and accurately infer the test data at untested points, thus significantly reducing test time and test costs.
  • the system also includes: a dynamic inspection model 4, which is used to dynamically evaluate and determine whether the fully connected deep neural network model needs to collect training samples again based on the preset accuracy threshold value.
  • the designed radio frequency communication equipment radiation test system will first randomly select N test test points in the three-dimensional space for measurement and store the data ⁇ N test ,X,y E ⁇ Used in the verification pool for all subsequent evaluations and inspections.
  • additional N train test points will be randomly selected to collect data ⁇ N train ,X,y E ⁇
  • the FCDNN model will be optimized using supervised learning around the above stochastic gradient descent algorithm formula. neural network parameters.
  • the accuracy of the trained FCDNN model will be evaluated against the samples in the validation pool. Only when its mean square error performance in the validation set is better than the given preset accuracy threshold value, that is, the tolerance threshold, will the model be finally established, otherwise Training data will be further collected and the above training and testing steps will be repeated.
  • the system in order to obtain a better training model, also includes a normalization module (not shown in the figure), which is used to normalize the actual radiation performance parameters of the device under test, and generate a fully connected deep neural network model by minimizing the mean square error of the fully connected deep neural network based on the normalized actual radiation performance parameters.
  • a normalization module (not shown in the figure), which is used to normalize the actual radiation performance parameters of the device under test, and generate a fully connected deep neural network model by minimizing the mean square error of the fully connected deep neural network based on the normalized actual radiation performance parameters.
  • the mobile phone will be used as the device under test to measure the space radiation of its 802.11n Wi-Fi radio frequency communication equipment. performance.
  • the device under test adopts the following settings: the access channel is channel 1 in the 2.4GHz frequency band, and its carrier center frequency is 2412MHz; the system bandwidth is 20MHz; the modulation and coding strategy used is MCS3, which is the physical The layer rate is 26Mbps.
  • the EIRP and EIS of the equipment under test are measured based on the traditional space radiation performance testing method of radio frequency communication equipment, which serves as the benchmark for subsequent experimental performance comparisons.
  • test point density consider two methods with intervals of 30 degrees ( Figure 5) and 15 degrees ( Figure 6).
  • Test points are established for the elevation and azimuth angle sampling of the spherical coordinate system in Figure 3.
  • the total number of measurement points under the two different test point densities is 120 and 528 respectively.
  • the measured EIRP and EIS performance of the device under test are shown in Figure 5 and Figure 6. It can be seen that there are obvious differences in the radiation performance of the device under test in different directions in space.
  • denser test points can characterize the spatial radiation performance more accurately, but at the same time, the number of test points required is also higher, which will lead to higher test time and system complexity.
  • test data of Ntest 60 measurement points will be randomly sampled to form the verification set of the experiment.
  • additional measurement data at 140, 280, and 420 test points will be randomly sampled to construct training sets of three different sizes, and the prediction accuracy of their trained models will be compared.
  • hyperparameters of the model a four-layer FCDNN architecture is adopted for EIRP and EIS respectively.
  • the number of neurons in each layer is 3, 64, 64, and 1 respectively.
  • the activation function of each layer of neural network is a modified linear unit ( ReLu).
  • ReLu modified linear unit
  • a represents the original data
  • ⁇ a represents the standard deviation of the data.
  • SGD algorithm is used, and each FCDNN model is trained with 0.02 as the learning rate, and the number of training iterations (epoch) is 15,000. Training of each model is completed within ten seconds and will not become a bottleneck for the test system.
  • the following table shows the MSE error performance comparison of the FCDNN models trained with different numbers of training samples in the test set.
  • the FCDNN model can more accurately predict the spatial radiation performance of the device under test at unmeasured points.
  • the accuracy of model prediction is greatly improved, while when the number of training samples is increased from 280 to 420, the accuracy only achieves a small increase. Therefore, in order to reduce the test samples required to train the model, the proposed method of dynamically testing the accuracy of the model and gradually increasing the number of training samples can be adopted.
  • Figures 7, 8 and 9 show the spatial radiation performance of the radio frequency communication equipment under test inferred by the three FCDNN models under different training set sizes.
  • the proposed deep learning-based testing method can accurately depict the characteristics of the device under test. Space radiation performance.
  • the measurement results obtained by the proposed method are close to the results when the traditional testing method adopts 528 test points. Therefore, this observational result verifies the feasibility, efficiency, and accuracy of the proposed deep learning-based testing method.
  • test set is constructed and used as the number of training samples collected by the system each time.
  • 6 is used as the MSE threshold when the system evaluates the model.
  • the designed system only needs to collect two rounds of training data, that is, data on 280 test points, to meet the required test set MSE threshold requirements.
  • the proposed testing method based on deep learning needs to collect data on a total of 340 test points for training and testing.
  • the trained model will be used to infer the EIRP and EIS values of the device under test in all directions in space, and the TRP and TIS of the device will be calculated based on the inferred results.
  • Experimental results show that compared with traditional testing methods, the proposed deep learning-based testing method can obtain very accurate TRP and TIS measurement results, but only requires about 60% of test points, thus greatly improving testing efficiency and reducing system hardware complexity.
  • FIG10 is a flow chart of a method for testing spatial radiation of radio frequency communication equipment based on deep learning, and the method includes:
  • the self-designed multi-probe microwave anechoic chamber is used to obtain limited measurement data, that is, actual radiation performance parameters, in three-dimensional space, and is trained into a fully connected deep neural network FCDNN to obtain the radiation of the tested radio frequency communication system in all directions in the entire three-dimensional space. performance.
  • the probe microwave anechoic chamber is used to obtain the actual radiation performance parameters of the device under test.
  • the multi-probe microwave anechoic chamber includes: a high-precision turntable and a controller for performing all-round high-precision steering of the placed equipment under test on the horizontal plane; a microwave anechoic chamber; annular and evenly distributed antenna probes in the microwave anechoic chamber for Measure the actual radiation performance parameters of the device under test in all directions in space.
  • the radiation power of the radio frequency communication system in all directions of the three-dimensional space in the microwave anechoic chamber is a key indicator to measure its uplink transmission performance.
  • the reception sensitivity parameter can be used as a key indicator for its downlink transmission. Therefore, the actual radiation performance parameters include radiation power parameters and receiving sensitivity parameters.
  • the radiation power parameter represents the amount of radiation power of the antenna in each direction, which is defined as the product of the power of the antenna in a specific polarization mode and the absolute gain of the antenna in a given direction.
  • ( ⁇ r , ) can be expressed as a point in the unit radius spherical coordinate system, that is, a test point in a specific direction in three-dimensional space.
  • 0 ⁇ r ⁇ is the elevation angle of the spherical coordinate system, which can be adjusted by selecting different test antenna probes.
  • the device under test can be placed in the maximum transmit power state, the direction of the control turntable can be adjusted, and the antenna probes can adopt horizontal and vertical polarization methods respectively. Multiple deployed probes can be used to measure the device under test in various directions. The value of the radiation power parameter.
  • the equivalent isotropic radiated power parameter can be used as the actual radiated power parameter.
  • the test of the equivalent isotropic radiated power parameter EIRP gives the radiated power performance of the device under test in all directions in space, and The total radiated power reflects its overall radiated power situation and is defined as the integral of the radiated power parameter value of each point in space on the three-dimensional sphere. Since the number of test points that an actual system can carry is often limited by its hardware complexity, such as turntable accuracy, number of antenna probes, etc., the total radiated power is usually calculated using numerical integration.
  • the total radiated power TRP of the device under test can be expressed as:
  • the test of receiving sensitivity parameters can reflect the sensitivity of the receiving device under test in a specific direction in space when the transmitting antenna probe is in a certain polarization mode.
  • the device under test is first placed on the turntable. And adjust the turntable angle.
  • activate each antenna probe in the darkroom system in sequence adopt horizontal and vertical polarization methods respectively, and gradually reduce or increase the transmit power of the antenna probe, so that the downlink reception bit error rate (Bit ErrorRate, BER) of the device under test is approaches a given threshold, such as 10%.
  • BER downlink reception bit error rate
  • the received signal strength of the device under test that is, the minimum received signal strength that can be tolerated under a given bit error rate, represents its receiving sensitivity.
  • the test is carried out by considering the combination of different turntable directions, different antenna probes, and different antenna polarization methods to establish the receiving sensitivity of the device under test in all directions in three-dimensional space.
  • the total omnidirectional sensitivity parameter TIS can be used as the receiving sensitivity parameter, which can be calculated by integrating the value of the receiving sensitivity EIS at each point on the three-dimensional sphere.
  • the total omnidirectional sensitivity parameter TIS of the device under test can be expressed as :
  • TIS the reciprocal of EIS is adopted for integration. This is because the sensitivity of the device under test in each direction of space is often very different. If the reciprocal is not adopted, the test points with poor sensitivity, that is, the term with a higher receiving sensitivity EIS value, will dominate the integral result in this equation, and will As a result, the integration ignores the impact of test points with better sensitivity on the overall sensitivity of the device under test.
  • FCDNN ⁇ (. ⁇ ⁇ ) represents a fully connected deep neural network model
  • ⁇ ⁇ is its parameter
  • the subscript ⁇ is used to distinguish the model for the radiation power parameter or the model for the receiving sensitivity parameter
  • E[.] represents the mathematical expectation for the random variable x
  • X represents the set of all test points
  • represents the number of all test points
  • y ⁇ (x) represents the radiation performance parameter of the device under test at any point on the three-dimensional sphere, represents the radiation performance parameters of the device under test at point x calculated by the fully connected deep neural network
  • p represents the polarization mode of the antenna probe during the test
  • ⁇ r represents the elevation angle of the spherical coordinate system
  • the stochastic gradient descent algorithm can optimize the neural network parameters of the fully connected deep neural network model in a supervised learning manner to help build a fully connected neural network model.
  • the spatial radiation performance of the radio frequency communication equipment under test can be accurately reconstructed, and the test data at untested points can be accurately inferred, thus significantly reducing test time and test costs.
  • a solution will be further proposed to achieve accuracy through dynamic verification and gradually increase the number of training test points until the model accuracy reaches the preset requirements, and finally build a spatial radiation test system for RF communication equipment based on deep learning. Therefore, it also includes screening a fully connected deep neural network model that meets the preset accuracy threshold.
  • in order to obtain a better training model also includes normalizing the actual radiation performance parameters of the device under test, and minimizing the full connection depth based on the normalized actual radiation performance parameters.
  • Fully connected deep neural network model generated by neural mean square error training.
  • An embodiment of the present invention discloses a computer-readable storage medium storing a computer program for electronic data exchange, wherein the computer program enables a computer to execute the described deep learning-based spatial radiation testing method for radio frequency communication equipment.
  • An embodiment of the present invention discloses a computer program product.
  • the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause the computer to perform the described radio frequency communication based on deep learning. Equipment space radiation test methods.
  • modules described as separate components may or may not be physically separated.
  • the components shown as modules may or may not be physical modules, that is, they may be located in a place, or can be distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Persons of ordinary skill in the art can understand and implement the method without any creative effort.
  • each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware.
  • the above technical solutions can be embodied in the form of software products in essence or in part that contribute to the existing technology.
  • the computer software products can be stored in computer-readable storage media, and the storage media includes read-only memories.
  • Read-Only Memory ROM
  • RAM Random Access Memory
  • PROM Programmable Read-only Memory
  • EPROM Erasable Programmable Read Only Memory
  • OTPROM One-time Programmable Read-Only Memory
  • EEPROM Electronically Erasable Programmable Read-Only Memory
  • CD-ROM Compact Disc Read -Only Memory
  • the deep learning-based radio frequency communication equipment space radiation testing method and system disclosed in the embodiments of the present invention are only the preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention. It is not intended to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that they can still modify the technical solutions recorded in the foregoing embodiments, or modify some of the techniques. Equivalent substitutions are made for the features; however, these modifications or substitutions do not cause the essence of the corresponding technical solutions to depart from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Theoretical Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • Mobile Radio Communication Systems (AREA)
  • Monitoring And Testing Of Transmission In General (AREA)

Abstract

本发明公开了一种基于深度学习的射频通信设备空间辐射测试方法，该方法包括：获取被测设备在空间有限个测试点的实际辐射性能参数；基于实际辐射性能参数进行最小化的均方误差训练生成全连接深度神经网络模型；根据全连接深度神经网络模型推算被测设备在空间各方向的辐射性能参数。根据本发明提供的方法和系统能够充分利用射频通信设备测试系统中各个测试点所获取的真实数据，推断出未测试点的空间辐射性能，由此能够在保证射频辐射测试系统精度的条件下，大幅减少系统所需的测试点数目。

Description

基于深度学习的射频通信设备空间辐射测试系统及方法 技术领域
本发明涉及无线通信技术领域，尤其涉及一种基于深度学习的射频通信设备空间辐射测试系统及方法。
背景技术
射频发射机与接收机的空间辐射性能是无线通信设备的关键指标，主要由射频前端系统包括天线整体性能决定。国际标准组织CTIA与3GPP等规范要求在特定的微波暗室中测试射频通信系统在三维空间各个方向的发射的辐射功率与接收灵敏度。然而，传统的射频辐射(OTA，Over-the-Air)测试系统受到微波暗室尺寸大小、天线探头数量、转台旋转精度等物理条件及测试时间限制，其空间分辨力始终受限。
实际应用中，常常需要更高的空间分辨力来评估射频通信设备及天线系统在空间各个方向的辐射性能，用于预测无线通信设备在三维空间上的信号覆盖与接收能力。在对辐射功率和接收灵敏度的每个测试点所需的测试时间成本较高，尤其针对接收灵敏度测试，需在测试点上不断调整探头的发射功率以穷搜该点被测设备的接收灵敏度，导致实际测试系统很难支撑高密集度的测试点。此外，传统测试方法只能获取测试点上被测设备的辐射性能，缺乏有效准确的模型推断被测设备在未测试点上的辐射性能，通常会引入较大的推断误差。因此，传统的测试方法无论在系统复杂度、测试精度，还是在测试时间成本方面上均受到了较高的制约，亟需探寻新的测试方法以解决上述瓶颈问题。
发明内容
本发明所要解决的技术问题在于，提供一种基于深度学习的射频通信设备空间辐射测试系统及方法，能够充分利用射频辐射测试系统中有限个测试点所获取的真实数据，准确推断出未在任意方向上的空间辐射性能，由此能够在保证射频空间辐射测试系统精度的条件下，大幅减少系统所需的测试点数目。
为了解决上述技术问题，本发明第一方面公开了一种基于深度学习的射频通信设备空间辐射测试方法，其特征在于，所述方法包括：获取被测设备在有限个测试点的实际辐射性能参数；基于所述实际辐射性能参数进行最小化的均方误差训练生成全连接深度神经网络模型；根据所述全连接深度神经网络模型推算被测设备在空间各方向的辐射性能参数。
在一些实施方式中，所述获取被测设备在有限个测试点的实际辐射性能参数包括：将被测设备放置于高精度转台和控制器，通过所述高精度转台和控制器对放置的被测设备在水平面上进行全方位的高精度转向；通过环形均匀分布在所述微波暗室的天线探头测量所述被测设备在空间各个方 向上的实际辐射性能参数。
在一些实施方式中，辐射性能参数包括辐射功率参数和接收灵敏度参数，所述全连接深度神经网络模型实现为：将有限个测试点的实际辐射功率参数和实际接收灵敏度参数作为训练样本分别代入到下式的随机梯度下降算法公式：

其中，FCDNNα(.Φα)表示全连接深度神经网络模型，Φα为其参数，下标α用以区分针对辐射功率参数的模型或针对接收灵敏度参数的模型，E[.]表示针对于随机变量x的数学期望，X表示所有测试点的集合，|X|表示所有测试点的数量,yα(x)表示被测设备在三维球面上任意一点的辐射性能参数，表示全连接深度神经网络推算出被测设备在x点上的辐射性能参数，p表示测试时天线探头的极化方式,θr表示球坐标系的仰角,表示球坐标系的方位角。
在一些实施方式中，所述方法还包括：基于预置的精度门限值动态评估判断全连接深度神经网络模型是否需要再次采集训练样本；若所述全连接深度神经网络模型的验证集中的均方误差小于所述预置的精度门限值，则所述全连接深度神经网络模型需要再次采集训练样本。
在一些实施方式中，所述方法还包括：对被测设备在有限个测试点的实际辐射性能参数进行归一化处理；基于经过归一化处理的实际辐射性能参数进行最小化的全连接深度神经的均方误差训练生成的全连接深度神经网络模型。
根据本发明的第二个方面，提供了一种基于深度学习的射频通信设备空间辐射测试系统，其特征在于，所述系统包括：包括多探头微波暗室、无线通信测试仪器、信道模拟器等设备等的射频辐射测试系统，用于获取被测设备在有限个测试点的实际辐射性能参数；
基于所述实际辐射性能参数进行最小化的均方误差训练生成全连接深度神经网络模型；预测模块，用于基于所述全连接深度神经网络模型推算被测设备在空间各方向的辐射性能参数。
在一些实施方式中，所述多探头微波暗室包括：高精度转台和控制器，用于对放置的被测设备在水平面上进行全方位的高精度转向；微波暗室；环形均匀分布在所述微波暗室的天线探头，用于测量所述被测设备在空间各个方向上的实际辐射性能参数；用于切换天线探头的射频开关箱；用于测量被测设备在指定方向上射频辐射性能的无线通信测试仪器。
在一些实施方式中，辐射性能参数包括辐射功率参数和接收灵敏度参数，所述全连接深度神经网络模型实现为：将有限个测试点的实际辐射功率参数和实际接收灵敏度参数作为训练样本分别代入到下式的随机梯度下降算法公式：

其中，FCDNNα(.Φα)表示全连接深度神经网络模型，Φα为其参数，下标α用以区分针对辐射功率参数的模型或针对接收灵敏度参数的模型，E[.]表示针对于随机变量x的数学期望，X表示所有测试点的集合，|X|表示所有测试点的数量,yα(x)表示被测设备在三维球面上任意一点的辐射性能参数，表示全连接深度神经网络推算出被测设备在x点上的辐射性能参数，p表示测试时天线探头的极化方式,θr表示球坐标系的仰角,表示球坐标系的方位角。
在一些实施方式中，所述系统还包括：动态检验模型，基于预置的精度门限值动态评估判断全连接深度神经网络模型是否需要再次采集训练样本；若所述全连接深度神经网络模型的验证集中的均方误差小于所述预置的精度门限值，则所述全连接深度神经网络模型需要再次采集训练样本。
在一些实施方式中，所述系统还包括：归一化模块，用于对被测设备在有限个测试点的实际辐射性能参数进行归一化处理；基于经过归一化处理的实际辐射性能参数进行最小化的全连接深度神经的均方误差训练生成的全连接深度神经网络模型。
与现有技术相比，本发明的有益效果在于：
实施本发明能够通过自主研制的多探头微波暗室辅助获取精准的被测设备的辐射特性参数数据，并引入深度学习方法，利用三维空间有限测量数据训练一个全连接深度神经网络(FCDNN)模型，从而估计被测射频通信系统在三维空间各个方向上的辐射性能。为了权衡训练FCDNN模型所需的测试点数量与模型预测结果的准确度，进一步提出了动态检验模型准确度,通过预置的精度门限值逐步提升训练测试点数量，直到模型精度达到预设要求的解决办法。在实验结果表明，相比于传统测试系统，本发明所提出的基于深度学习的测试系统只需约60％的测试点，就能精准重构被测射频通信设备的空间辐射性能，验证了本发明的准确性与高效性。并且，本发明所以出的系统与现有的系统相比，还能够推断是空间中任意点的辐射性能参数，而非局限于测试点上。
而且，所提出的系统不仅能大幅降低测试点数量及系统硬件复杂度，更能准确构建出三维空间任意方向角度的辐射性能预测模型，为行业提供了一种高效精确却低成本的空间辐射测试技术解决方案。
附图说明
图1为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试的系统示意图；
图2为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的多探头微波暗室系统示意图；
图3为本发明实施例公开的一种三维空间方向与单位半径的球坐标系示意图；
图4为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的神经网络模型训练流程示意图；
图5为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的被测设备所测量的30度采样间隔下的EIRP、EIS性能示意图；
图6为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的被测设备所测量的15度采样间隔下的EIRP、EIS性能示意图；
图7为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的一种训练集规模下的FCDNN模型所推断出的被测设备射频通信设备的空间辐射性能示意图；
图8为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的又一种训练集规模下的FCDNN模型所推断出的被测设备射频通信设备的空间辐射性能示意图；
图9为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的另一种训练集规模下的FCDNN模型所推断出的被测设备射频通信设备的空间辐射性能示意图；
图10为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试方法的流程示意图。
具体实施方式
为了更好地理解和实施，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本发明实施例的术语“包括”和“具有”以及他们的任何变形，意图在于覆盖不排他的包含，例如，包含了一系列步骤或模块的过程、方法、系统、产品或设备不必限于清楚地列出的那些步骤或模块，而是可包括没有清楚地列出的或对于这些过程、方法、产品或设备固有的其它步骤或模块。
本发明实施例公开了一种基于深度学习的射频通信设备空间辐射测试系统及方法，通过自主研制的多探头微波暗室辅助获取精准的被测设备的辐射特性参数数据，并引入深度学习方法，利用三维空间有限测量数据训练一个全连接深度神经网络(FCDNN)模型，从而估计被测射频通信系统在三维空间各个方向上的辐射性能。为了权衡训练FCDNN模型所需的测试点数量与模型预测结果的准确度，进一步提出了动态检验模型准确度,通过预置的精度门限值逐步提升训练测试点数量，直到模型精度达到预设要求的解决办法。在实验结果表明，相比于传统测试系统，本发明所提出的基于深度学习的测试系统只需约60％的测试点，就能精准重构被测射频通信设备的空间辐射性能，验证了本发明的准确性与高效性。并且，本发明所以出的系统与现有的系统相比，还能够推断是空间中任意点的辐射性能参数，而非局限于测试点上。
而且，所提出的系统不仅能大幅降低测试点数量及系统硬件复杂度，更能准确构建出三维空间任意方向角度的辐射性能预测模型，为行业提供了一种高效精确却低成本的空间辐射测试技术解决方案。
请参阅图1，图1为本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试系统的示意图。其中，该基于深度学习的射频通信设备空间辐射测试系统可以应用在无线通信系统，对于该系统的应用本发明实施例不做限制。如图1所示，该基于深度学习的射频通信设备空间辐射测试系统可以包括：
多探头微波暗室1、全连接深度神经网络模型2和预测模块3。在现有的技术方案中，往往为了提升射频辐射OTA测试系统的空间分辨力，提出使用更大的微波暗室，及部署更多的天线探头，从而通过在三维空间上测试更多点的射频辐射数据以获得更好的射频辐射空间分辨力。然而，这些方法会占用更大的场地空间，增加测试系统的复杂度，从而大幅增加测试时间和测试成本。此外，为了降低测试时间，提出使用更少的天线探头或更大的转台旋转步长，通过减少空间上的测试点数从而节省测试时间。这类方法以牺牲射频通信系统辐射性能的空间分辨力和测量精度来加快测试速度，适用于需要快速验证设计的场景。为了减少场地空间占用，提出使用紧缩场技术在近距离上实现射频通信设备空间辐射性能的远场测量，但这类测量方法仍无法有效推断被测设备在未测试点上的射频通信设备辐射性能，会引入较高的误差。
为此，本发明采取了利用自主设计的多探头微波暗室在三维空间中获取有限的测量数据即实际辐射性能参数，训练成一个全连接深度神经网络FCDNN，从而得到被测射频通信系统在整个三维空间中各个方向上的辐射性能。
具体地，多探头微波暗室1用于获取被测设备在有限个测试点的实际辐射性能参数。该多探头微波暗室1包括：高精度转台和控制器，用于对放置的被测设备在水平面上进行全方位的高精度转向；微波暗室；环形均匀分布在所述微波暗室的天线探头，用于测量被测设备在空间各个方向上的实际辐射性能参数。如图2所示，为一种具体应用中的多探头微波暗室系统示意图，在实际使用时，该系统主要由宽带无线通信测试仪(RF communication test instrument)、微波暗室(Anechoic chamber)、天线探头(Probe antennas)、高精度转台和控制器(Turn table andcontroller)，以及带自动测试软件(Automatic testequipment，ATE)的个人电脑组成。其中，为了避免射频信号在空间辐射测试中的反射以及回波，在微波暗室的内部可以采纳吸波材料，从而构建出理想的测试环境。在测试时，将被测设备(Device under test，DUT)放置于转台上，并通过转台控制器使被测设备在水平面上进行全方位的高精度转向。在垂直于水平转台的平面上，通常采纳环形均匀分布的架构在暗室中部署多个天线探头，以测量被测设备在空间各个方向上的射频通信设备辐射性能即被测设备的实际辐射性能参数。这些天线探头与射频转换开关以及空间信道模拟器相连，并由宽带无线通信测试仪进行控制，从而精准的测量被测设备在空间中不同方向、以及不同无线环境下的辐射性能。最后，宽带无线通信测试仪以及转台控制器通过一台个人电脑全局管控，由自主开发的自动测试软件提供用户接口，进一步实现高效的测试途径，以及可视化数据的输出和存储。
其中，射频通信系统在微波暗室中三维空间各个方向的辐射功率是衡量其上行传输性能的关键指标。而其作为接收机时，其下行传输则可以由接收灵敏度参数作为关键指标。由此，实际辐射性能参数中包含有辐射功率参数和接收灵敏度参数。
具体地，辐射功率参数表征天线在各个方向上辐射功率的大小，其定义为在某一特定极化方式下，天线的功率与给定方向上天线绝对增益的乘积。如图3的球坐标系所示，(θr，)，可表示为单位半径球坐标系中的一个点，即三维空间中一个特定方向上的测试点。其中，0≤θr≤π为球坐标系的仰角，可通过选取不同的测试天线探头进行调整。为球坐标系的方位角，可通过控制转台的角度进行调整。在实际测试中，可以令被测设备处于最大发射功率状态，调整控制转台的方向，并使天线探头分别采纳水平和垂直的极化方式，利用多个部署的探头从而衡量各个方向上被测设备的辐射功率参数的值。
在其他优先实施方式中，可使用等效全向辐射功率参数作为实际的辐射功率参数，等效全向辐射功率参数EIRP的测试给出了被测设备在空间各个方向上的辐射功率性能，而总辐射功率则反映了其整体辐射功率情况，定义为空间中各点的辐射功率参数的值在三维球面上的积分。由于实际系统所能承载的测试点数量往往受制于其硬件复杂度，如转台精度、天线探头数目等，总辐射功率通常采纳数值积分的方式进行计算。当在仰角坐标上采样N个点，即部署N个天线探头，方位角坐标上采样M个点，即水平转台可精确转向M个角度时，被测件的总辐射功率TRP可以表示为：
具体地，接收灵敏度参数的测试可以反应当发射天线探头在某一极化方式下，被测接收设备针对空间中特定方向上的灵敏度，在实际测试中， 首先将被测设备放置于转台上，并调整转台角度。接下来依次激活暗室系统中的各个天线探头，分别采纳水平以及垂直的极化方式，并逐步降低或者升高天线探头的发射功率，使得被测设备的下行接收误比特率(Bit ErrorRate，BER)趋近于一个给定门限，如10％。此时被测设备的接收信号强度，即满足给定误比特率条件下所能容忍的最小接收信号强度，代表其接收灵敏度。最后，通过结合考虑不同转台方向、不同天线探头、不同天线极化方式的组合进行测试，从而确立被测设备在三维空间各个方向的接收灵敏度。
在其他优先实施方式中，可使用总全向灵敏度参数TIS作为接收灵敏度参数，可以通过对三维球面上各点接收灵敏度EIS的值积分进行计算，被测设备的总全向灵敏度参数TIS可以表示为：
可见，TIS的计算方式不同意上述的TRP，在TIS的计算中采纳了了EIS的倒数进行积分。这是由于被测设备在空间各个方向上的灵敏度往往差异较大，若不采纳倒数，灵敏度较差的测试点，即接收灵敏度EIS值较高的项，会主导该式中积分的结果，将导致积分忽略了灵敏度较优的测试点对被测设备整体灵敏度的影响。
进一步地，在获取了实际辐射性能参数后，就可以搭建全连接深度神经网络模型，具体实现为将实际辐射功率参数和实际接收灵敏度参数作为测试点分别代入到下式的随机梯度下降算法公式：

其中，FCDNNα(.Φα)表示全连接深度神经网络模型，Φα为其参数，下标α用以区分针对辐射功率参数的模型或针对接收灵敏度参数的模型，E[.]表示针对于随机变量x的数学期望，X表示所有测试点的集合，|X|表示所有测试点的数量,yα(x)表示被测设备在三维球面上任意一点的辐射性能参数，表示全连接深度神经网络推算出被测设备在x点上的辐射性能参数，p表示测试时天线探头的极化方式,θr表示球坐标系的仰角,表示球坐标系的方位角。
该随机梯度下降算法即SGD能够实现监督学习的方式优化该全连接深度神经网络模型的神经网络参数帮助构建全连接的神经网络模型。
进一步地，在得到了全连接的神经网络模型后，预测模块3就能够基于全连接深度神经网络模型直接推算被测设备在空间各方向的辐射性能参数，由此基于有限个测试点，就能精准重构被测射频通信设备的空间辐射性能，及准确推断出在未测试点上的测试数据，从而大幅降低了测试时间和测试成本。
作为一种其他优先实施方式中，虽然采纳越密集的测试点往往能训练出越准确的模型，但采纳过于密集的测试点通常很难有效提高模型的准确度，且将引入实际测试系统难以承载的测试次数以及时间成本。为了权衡训练FCDNN模型所需的测试点数量与模型预测结果的准确度，将进一步提出通过动态检验模型准确度,并逐步提升训练测试点数量，直到模型精度达到预设要求的解决办法，最终构建出基于深度学习的射频通信设备空间辐射测试系统。由此，该系统还包括：动态检验模型4，该动态检验模型4用于基于预置的精度门限值动态评估判断全连接深度神经网络模型是否需要再次采集训练样本。
具体地，如图4所示，所设计的射频通信设备辐射测试系统将先在三维空间上随机选取Ntest个测试点进行测量并将数据{Ntest,X,yE}，并将其存储于验证池中用以后续所有的评估以及检验。接下来，将在排除验证集中测试点的条件下，额外随机选取Ntrain个测试点采集数据{Ntrain,X,yE}，围绕上述的随机梯度下降算法公式采纳监督学习的方式优化FCDNN模型的神经网络参数。最后，将针对验证池中的样本评估训练后FCDNN模型的准确度，只有当其在验证集中的均方误差表现优于给定预置的精度门限值即容忍门限时才最终确立模型，否则将进一步采集训练数据，并重复以上训练、检验步骤。
在其他优先实施方式中，为了获得更好的训练模型，本系统还包括归一化模块(图中未示)，用于对被测设备的实际辐射性能参数进行归一化处理，基于经过归一化处理的实际辐射性能参数进行最小化的全连接深度神经的均方误差训练生成的全连接深度神经网络模型。
具体地，作为本发明的一种验证实施例，请参阅图5和图6。为了验证所提出的基于深度学习的射频通信设备空间辐射测试系统的可行性、高效性，以及准确性，接下来将以手机作为被测设备，测量其802.11n Wi-Fi射频通信设备的空间辐射性能。在以下的实验中，考虑被测设备采纳如下设置：接入信道为2.4GHz频段下的1号信道，其载波中心频率为2412MHz；系统带宽为20MHz；采用的调制与编码策略为MCS3，即物理层速率为26Mbps。
首先，基于传统射频通信设备空间辐射性能的测试方式测量被测设备的EIRP以及EIS，作为后续实验性能比对的基准。在测试点密度方面，分别考虑以30度为间隔(图5)，和以15度为间隔的两种方式(图6)，针对图3球坐标系的仰角以及方位角采样确立测试点。结合考虑天线探头水平与垂直的极化方式，两种不同测试点密度下的总测量点数分别为120个、528个。被测设备所测量的EIRP、EIS性能如图5和图6所示。可以看出被测设备在空间中不同方向上的辐射性能存在明显差异。此外，越密集的测试点所刻画出的空间辐射性能越精确，但同时所需的测试点数目也较高，将导致较高的测试时间以及系统复杂度。
接下来，通过本发明的基于深度学习的射频通信设备空间辐射测试系统法测量同一被测设备的空间辐射性能。首先，将随机采样Ntest＝60个测量点的测试数据以构成实验的验证集。在训练集方面，将分别额外随机采样140、280、420个测试点上的测量数据，构建出三种不同规模的训练集，并比较他们所训练模型的预测准确度。在模型的超参方面，针对EIRP以及EIS分别采纳了一个四层的FCDNN架构，各层的神经元数量分别为3、64、64、1，每层神经网络的激活函数均为修正线性单元(ReLu)。此外，为了获得更好的训练模型，对采集的数据进行了下式的归一化处理：
其中，a代表原始数据，表示数据的均值，σa则代表数据的标准差。最后，采用SGD算法，并以0.02作为学习率(learning rate)训练各个FCDNN模型，训练的迭代次数(epoch)为15000。每个模型的训练均在十秒内完成，不会成为测试系统的瓶颈。
下表所示为不同训练样本数目所训练的FCDNN模型在测试集中的MSE误差性能比对。
随着训练样本数的增大，FCDNN模型能更准确的预测被测设备在未测量点上的空间辐射性能。此外，当训练样本数从140增加到280时，模型预测的准确度获得了大幅提升，而当训练样本数从280增加到420时，准确度只取得了小幅的增加。因此，为了减少训练模型所需的测试样本，可以采纳所提出的通过动态检验模型准确度，并逐渐提升训练样本数的方法。
图7、图8和图9所示为不同训练集规模下的三个FCDNN模型所推断出的被测设备射频通信设备的空间辐射性能。相比于图5和图6中传统测试方法在高密度测试点下的测量结果，可以看出随着训练样本数的增加，所提出的基于深度学习测试方法能够准确的刻画出被测设备的空间辐射性能。此外，当训练样本数为280时，所提出的方法得到的测量结果趋近于传统测试方法采纳528个测试点时的结果。因此，这一观测结果验证了所提基于深度学习测试方法的可行性、高效性与准确性。
进一步地，下表所示为传统测试方法与所提出的基于深度学习的测试方法针对TRP、TIS测试结果与测试点数的比对。
在传统测试方法中，分别针对三维球面采样了528个点以及384个点进行EIRP、EIS的测量并计算出TRP、TIS。在所提出的测试方法中，采纳构建测试集，并采纳作为系统每次采集训练样本的个数。此外，以6作为系统评估模型时的MSE门限。在实验中，所设计的系统只需采集两轮训练数据，即280个测试点上的数据，就可达到所需的测试集MSE门限要求。考虑系统需采集60个测试点上的数据作为测试集，因而所提出的基于深度学习的测试方法共需要采集340个测试点上的数据用以训练、测试。最后，当模型完成训练后，将利用训练好的模型推断出被测设备在空间各个方向上的EIRP、EIS值，并基于推断的结果计算设备的TRP、TIS。实验结果表明，相比于传统的测试方法，所提出的基于深度学习的测试方法能得到非常精准的TRP、TIS测量结果，却只需约60％的测试点，从而大幅提升了测试效率并降低了系统硬件复杂度。
请参与图10，图10为基于深度学习的射频通信设备空间辐射测试方法的流程图，该方法包括：
701、获取被测设备在有限个测试点的实际辐射性能参数。
利用自主设计的多探头微波暗室在三维空间中获取有限的测量数据即实际辐射性能参数，训练成一个全连接深度神经网络FCDNN，从而得到被测射频通信系统在整个三维空间中各个方向上的辐射性能。具体地，探头微波暗室用于获取被测设备的实际辐射性能参数。该多探头微波暗室包括：高精度转台和控制器，用于对放置的被测设备在水平面上进行全方位的高精度转向；微波暗室；环形均匀分布在所述微波暗室的天线探头，用于测量被测设备在空间各个方向上的实际辐射性能参数。
其中，射频通信系统在微波暗室中三维空间各个方向的辐射功率是衡量其上行传输性能的关键指标。而其作为接收机时，其下行传输则可以由接收灵敏度参数作为关键指标。由此，实际辐射性能参数中包含有辐射功率参数和接收灵敏度参数。
具体地，辐射功率参数表征天线在各个方向上辐射功率的大小，其定义为在某一特定极化方式下，天线的功率与给定方向上天线绝对增益的乘积。如图3的球坐标系所示，(θr，)，可表示为单位半径球坐标系中的一个点，即三维空间中一个特定方向上的测试点。其中，0≤θr≤π为球坐标系的仰角，可通过选取不同的测试天线探头进行调整。为球坐标系的方位角，可通过控制转台的角度进行调整。在实际测试中，可以令被测设备处于最大发射功率状态，调整控制转台的方向，并使天线探头分别采纳水平和垂直的极化方式，利用多个部署的探头从而衡量各个方向上被测设备的辐射功率参数的值。
在其他优先实施方式中，可使用等效全向辐射功率参数作为实际的辐射功率参数，等效全向辐射功率参数EIRP的测试给出了被测设备在空间各个方向上的辐射功率性能，而总辐射功率则反映了其整体辐射功率情况，定义为空间中各点的辐射功率参数的值在三维球面上的积分。由于实际系统所能承载的测试点数量往往受制于其硬件复杂度，如转台精度、天线探头数目等，总辐射功率通常采纳数值积分的方式进行计算。当在仰角坐标上采样N个点，即部署N个天线探头，方位角坐标上采样M个点，即水平转台可精确转向M个角度时，被测件的总辐射功率TRP可以表示为：
具体地，接收灵敏度参数的测试可以反应当发射天线探头在某一极化方式下，被测接收设备针对空间中特定方向上的灵敏度，在实际测试中，首先将被测设备放置于转台上，并调整转台角度。接下来依次激活暗室系统中的各个天线探头，分别采纳水平以及垂直的极化方式，并逐步降低或者升高天线探头的发射功率，使得被测设备的下行接收误比特率(Bit ErrorRate，BER)趋近于一个给定门限，如10％。此时被测设备的接收信号强度，即满足给定误比特率条件下所能容忍的最小接收信号强度，代表其接收灵敏度。最后，通过结合考虑不同转台方向、不同天线探头、不同天线极化方式的组合进行测试，从而确立被测设备在三维空间各个方向的接收灵敏度。
在其他优先实施方式中，可使用总全向灵敏度参数TIS作为接收灵敏度参数，可以通过对三维球面上各点接收灵敏度EIS的值积分进行计算，被测设备的总全向灵敏度参数TIS可以表示为：
可见，TIS的计算方式不同意上述的TRP，在TIS的计算中采纳了了EIS的倒数进行积分。这是由于被测设备在空间各个方向上的灵敏度往往差异较大，若不采纳倒数，灵敏度较差的测试点，即接收灵敏度EIS值较高的项，会主导该式中积分的结果，将导致积分忽略了灵敏度较优的测试点对被测设备整体灵敏度的影响。
702、基于实际辐射性能参数进行最小化的均方误差训练生成全连接深度神经网络模型。
在获取了实际辐射性能参数后，就可以搭建全连接深度神经网络模型，具体实现为将实际辐射功率参数和实际接收灵敏度参数作为测试点分别代入到下式的随机梯度下降算法公式：

其中，FCDNNα(.Φα)表示全连接深度神经网络模型，Φα为其参数，下标α用以区分针对辐射功率参数的模型或针对接收灵敏度参数的模型，E[.]表示针对于随机变量x的数学期望，X表示所有测试点的集合，|X|表示所有测试点的数量,yα(x)表示被测设备在三维球面上任意一点的辐射性能参数，表示全连接深度神经网络推算出被测设备在x点上的辐射性能参数，p表示测试时天线探头的极化方式,θr表示球坐标系的仰角,表示球坐标系的方位角。
该随机梯度下降算法即SGD能够实现监督学习的方式优化该全连接深度神经网络模型的神经网络参数帮助构建全连接的神经网络模型。
703、根据全连接深度神经网络模型推算被测设备在空间各方向的辐射性能参数。
由此基于有限个测试点，就能精准重构被测射频通信设备的空间辐射性能，及准确推断出在未测试点上的测试数据，从而大幅降低了测试时间和测试成本。
作为一种其他优先实施方式中，虽然采纳越密集的测试点往往能训练出越准确的模型，但采纳过于密集的测试点通常很难有效提高模型的准确度，且将引入实际测试系统难以承载的测试次数以及时间成本。为了权衡训练FCDNN模型所需的测试点数量与模型预测结果的准确度，将进一步提出通过动态检验的准确度,并逐步提升训练测试点数量，直到模型精度达到预设要求的解决办法，最终构建出基于深度学习的射频通信设备空间辐射测试系统。由此，还包括筛选满足预置的精度门限值的全连接深度神经网络模型。
在其他优先实施方式中，为了获得更好的训练模型，还包括对被测设备的实际辐射性能参数进行归一化处理，基于经过归一化处理的实际辐射性能参数进行最小化的全连接深度神经的均方误差训练生成的全连接深度神经网络模型。
对于上述步骤的实现方式请参考系统部门的实现方式，在此不进行赘述。
本发明实施例公开了一种计算机可读存储介质，其存储用于电子数据交换的计算机程序，其中，该计算机程序使得计算机执行所描述的基于深度学习的射频通信设备空间辐射测试方法。
本发明实施例公开了一种计算机程序产品，该计算机程序产品包括存储了计算机程序的非瞬时性计算机可读存储介质，且该计算机程序可操作来使计算机执行所描述的基于深度学习的射频通信设备空间辐射测试方法。
以上所描述的实施例仅是示意性的，其中所述作为分离部件说明的模块可以是或者也可以不是物理上分开的，作为模块显示的部件可以是或者也可以不是物理模块，即可以位于一个地方，或者也可以分布到多个网络模块上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。本领域普通技术人员在不付出创造性的劳动的情况下，即可以理解并实施。
通过以上的实施例的具体描述，本领域的技术人员可以清楚地了解到各实施方式可借助软件加必需的通用硬件平台的方式来实现，当然也可以通过硬件。基于这样的理解，上述技术方案本质上或者说对现有技术做出贡献的部分可以以软件产品的形式体现出来，该计算机软件产品可以存储在计算机可读存储介质中,存储介质包括只读存储器(Read-Only Memory，ROM)、随机存储器(Random Access Memory，RAM)、可编程只读存储器(Programmable Read-only Memory，PROM)、可擦除可编程只读存储器(Erasable Programmable Read Only Memory，EPROM)、一次可编程只读存储器(One-time Programmable Read-Only Memory，OTPROM)、电子抹除式可复写只读存储器(Electrically-Erasable Programmable Read-Only Memory，EEPROM)、只读光盘(Compact Disc Read-Only Memory，CD-ROM)或其他光盘存储器、磁盘存储器、磁带存储器、或者能够用于携带或存储数据的计算机可读的任何其他介质。
最后应说明的是：本发明实施例公开的一种基于深度学习的射频通信设备空间辐射测试方法及系统所揭露的仅为本发明较佳实施例而已，仅用于说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解；其依然可以对前述各项实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或替换，并不使相应的技术方案的本质脱离本发明各项实施例技术方案的精神和范围。

Claims (8)

  1. 一种基于深度学习的射频通信设备空间辐射测试方法，其特征在于，所述方法包括：
    获取被测设备在有限个测试点的实际辐射性能参数；
    基于所述实际辐射性能参数进行最小化的均方误差训练生成全连接深度神经网络模型；
    根据所述全连接深度神经网络模型推算被测设备在空间各方向的辐射性能参数；
    辐射性能参数包括辐射功率参数和接收灵敏度参数，所述全连接深度神经网络模型实现为：
    将有限个测试点的实际辐射功率参数和实际接收灵敏度参数作为训练样本，并采纳随机梯度下降算法依据下式的损失函数更新神经网络参数：

    其中，FCDNNα(Φα)表示全连接深度神经网络模型，Φα为其参数，下标α用以区分针对辐射功率参数的模型或针对接收灵敏度参数的模型，E[.]表示针对于随机变量x的数学期望，X表示所有测试点的集合，|X|表示所有测试点的数量,yα(x)表示被测设备在三维球面上任意一点的辐射性能参数，表示全连接深度神经网络推算出被测设备在x点上的辐射性能参数，p表示测试时天线探头的极化方式,θr表示球坐标系的仰角,表示球坐标系的方位角。
  2. 根据权利要求1所述的基于深度学习的射频通信设备空间辐射测试方法，其特征在于，所述获取被测设备在有限个测试点的实际辐射性能参数包括：
    将被测设备放置于高精度转台和控制器，通过所述高精度转台和控制器对放置的被测设备在水平面上进行全方位的高精度转向；
    通过环形均匀分布在微波暗室的天线探头测量所述被测设备在空间各个方向上的实际辐射性能参数。
  3. 根据权利要求1所述的基于深度学习的射频通信设备空间辐射测试方法，其特征在于，所述方法还包括：
    基于预置的精度门限值动态评估判断全连接深度神经网络模型是否需要再次采集训练样本；
    若所述全连接深度神经网络模型的验证集中的均方误差小于所述预置的精度门限值，则所述全连接深度神经网络模型需要再次采集训练样本。
  4. 根据权利要求1-3任一项所述的基于深度学习的射频通信设备空间辐射测试方法，其特征在于，所述方法还包括：
    对被测设备在有限个测试点的实际辐射性能参数进行归一化处理；
    基于经过归一化处理的实际辐射性能参数进行最小化的全连接深度神经的均方误差训练生成的全连接深度神经网络模型。
  5. 一种基于深度学习的射频通信设备空间辐射测试系统，其特征在于，所述系统包括：
    多探头微波暗室，用于获取被测设备在有限个测试点的实际辐射性能参数；
    基于所述实际辐射性能参数进行最小化的均方误差训练生成全连接深度神经网络模型；
    预测模块，用于基于所述全连接深度神经网络模型推算被测设备在空间各方向的辐射性能参数；
    辐射性能参数包括辐射功率参数和接收灵敏度参数，所述全连接深度神经网络模型实现为：
    将有限个测试点的实际辐射功率参数和实际接收灵敏度参数作为训练样本，并采纳随机梯度下降算法依据下式的损失函数更新神经网络参数：

    其中，FCDNNα(Φα)表示全连接深度神经网络模型，Φα为其参数，下标α用以区分针对辐射功率参数的模型或针对接收灵敏度参数的模型，E[.]表示针对于随机变量x的数学期望，X表示所有测试点的集合，|X|表示所有测试点的数量,yα(x)表示被测设备在三维球面上任意一点的辐射性能参数，表示全连接深度神经网络推算出被测设备在x点上的辐射性能参数，p表示测试时天线探头的极化方式,θr表示球坐标系的仰角,表示球坐标系的方位角。
  6. 根据权利要求5所述的基于深度学习的射频通信设备空间辐射测试系统，其特征在于，多探头微波暗室包括：
    高精度转台和控制器，用于对放置的被测设备在水平面上进行全方位的高精度转向；
    微波暗室；
    环形均匀分布在所述微波暗室的天线探头，用于测量所述被测设备在空间各个方向上的实际辐射性能参数。
  7. 根据权利要求6所述的基于深度学习的射频通信设备空间辐射测试系统，其特征在于，所述系统还包括：
    动态检验模型，基于预置的精度门限值动态评估判断全连接深度神经网络模型是否需要再次采集训练样本；
    若所述全连接深度神经网络模型的验证集中的均方误差小于所述预置的精度门限值，则所述全连接深度神经网络模型需要再次采集训练样本。
  8. 根据权利要求5-7任一项所述的基于深度学习的射频通信设备空间辐射测试系统，其特征在于，所述系统还包括：
    归一化模块，用于对被测设备在有限个测试点的实际辐射性能参数进行归一化处理；
    基于经过归一化处理的实际辐射性能参数进行最小化的全连接深度神经的均方误差训练生成的全连接深度神经网络模型。
PCT/CN2023/087808 2022-09-19 2023-04-12 基于深度学习的射频通信设备空间辐射测试系统及方法 WO2024060592A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211136481.X 2022-09-19
CN202211136481.XA CN115378518B (zh) 2022-09-19 2022-09-19 基于深度学习的射频通信设备空间辐射测试系统及方法

Publications (1)

Publication Number Publication Date
WO2024060592A1 true WO2024060592A1 (zh) 2024-03-28

Family

ID=84070668

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/087808 WO2024060592A1 (zh) 2022-09-19 2023-04-12 基于深度学习的射频通信设备空间辐射测试系统及方法

Country Status (2)

Country Link
CN (1) CN115378518B (zh)
WO (1) WO2024060592A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115378518B (zh) * 2022-09-19 2023-04-07 深圳市中承科技有限公司 基于深度学习的射频通信设备空间辐射测试系统及方法
CN116973819B (zh) * 2023-09-22 2023-12-12 上海优立检测技术股份有限公司 一种微波电磁参数三维测试方法、系统及储存介质
CN117560098B (zh) * 2024-01-11 2024-04-02 为准（北京）电子科技有限公司 预测射频通信设备的辐射性能参数的方法及装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110095950A1 (en) * 2008-05-15 2011-04-28 Zhong Yu Method and system for testing the radiation performance of wireless terminal based on data mode
US20120275506A1 (en) * 2010-01-19 2012-11-01 Tiantian Ding Method and Apparatus for Testing Total Isotropic Sensitivity in Multi-Antenna MIMO System
CN105406922A (zh) * 2014-09-04 2016-03-16 盛和联（苏州）电子科技有限公司 一种空间辐射性能测试方法
CN113162706A (zh) * 2020-01-22 2021-07-23 深圳市通用测试系统有限公司 无线设备的射频性能测试方法及系统
US20220260625A1 (en) * 2019-06-14 2022-08-18 Zte Corporation Method, device, system and terminal for measuring total radiation power of array antenna and computer storage medium
CN115378518A (zh) * 2022-09-19 2022-11-22 深圳市中承科技有限公司 基于深度学习的射频通信设备空间辐射测试系统及方法

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101399621B (zh) * 2007-09-24 2012-07-04 中兴通讯股份有限公司 空间射频性能测试方法
CN101769967A (zh) * 2008-12-30 2010-07-07 中兴通讯股份有限公司 辐射性能测试方法和系统

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110095950A1 (en) * 2008-05-15 2011-04-28 Zhong Yu Method and system for testing the radiation performance of wireless terminal based on data mode
US20120275506A1 (en) * 2010-01-19 2012-11-01 Tiantian Ding Method and Apparatus for Testing Total Isotropic Sensitivity in Multi-Antenna MIMO System
CN105406922A (zh) * 2014-09-04 2016-03-16 盛和联（苏州）电子科技有限公司 一种空间辐射性能测试方法
US20220260625A1 (en) * 2019-06-14 2022-08-18 Zte Corporation Method, device, system and terminal for measuring total radiation power of array antenna and computer storage medium
CN113162706A (zh) * 2020-01-22 2021-07-23 深圳市通用测试系统有限公司 无线设备的射频性能测试方法及系统
CN115378518A (zh) * 2022-09-19 2022-11-22 深圳市中承科技有限公司 基于深度学习的射频通信设备空间辐射测试系统及方法

Also Published As

Publication number Publication date
CN115378518B (zh) 2023-04-07
CN115378518A (zh) 2022-11-22

Similar Documents

Publication Publication Date Title
WO2024060592A1 (zh) 基于深度学习的射频通信设备空间辐射测试系统及方法
CN108234036B (zh) Mimo无线终端的无线性能测试方法
JP5886947B2 (ja) アンテナ、携帯電話及び他の無線端末の性能を測定するための改良された方法と装置
CN113225147B (zh) 一种阵列天线总辐射功率的测量方法、装置和系统
JP4625087B2 (ja) 無線デバイスの放射性能を決定するためのシステム、方法及び装置
CN113162706B (zh) 无线设备的射频性能测试方法及系统
JP2010048708A (ja) 方向推定システム、方向推定方法及び方向推定プログラム
CN102340360B (zh) 测试移动终端总全向灵敏度的方法和装置
US11012163B1 (en) Apparatus and methods for fast and accurate near-field measurement
CN107819529A (zh) 一种tis快速测量的方法
CN102237934A (zh) 一种终端数据传输性能的测试方法、系统及装置
CN117250412A (zh) 一种车载5g天线的测试方法及测试系统
CN110768734B (zh) 一种测量方法和装置
CN107957518A (zh) 一种天线方向性测试方法及系统
CN107911180A (zh) 一种大尺寸终端产品的空间射频性能测试方法
Kildal et al. Direct coupling as a residual error contribution during OTA measurements of wireless devices in reverberation chamber
CN107817391A (zh) 一种tis快速测量的方法
CN116232494A (zh) 多天线终端的性能测试方法及装置
Goulianos et al. A novel path-loss model for UWB off-body propagation
TWI781451B (zh) 用以基於一或多個近場測量結果獲得受測裝置性能度量之設備、系統及方法，用以基於自動測試裝備中之具有一或多個天線的受測裝置之近場中之空中測試判定受測裝置之性能度量之方法，以及相關之電腦程式
CN115913421A (zh) 射频指标测量方法、装置、系统、电子设备及存储介质
CN113242101A (zh) 无线设备的射频性能测试方法及系统
KR100589475B1 (ko) Cdma, pcs 및 gsm 방식의 핸드폰성능테스트 방법 및 장치
CN112865840B (zh) Mimo无线终端的测试方法、装置以及系统
Boussad et al. Evaluating smartphone performance for cellular power measurement

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23866889

Country of ref document: EP

Kind code of ref document: A1