A wireless channel characteristic parameter dynamic measurement device and a space migration method

Abstract

The application discloses a wireless channel characteristic parameter dynamic measurement device and a space domain migration method, wherein the device comprises a channel measurement unit, a characteristic parameter extraction unit, a characteristic parameter network training unit and a characteristic parameter space domain migration unit; an output interface of the channel measurement unit is connected with the characteristic parameter extraction unit through a PCIE bus, output interfaces of the characteristic parameter extraction unit are connected with input interfaces of the characteristic parameter network training unit and the characteristic parameter space domain migration unit respectively, and an output interface of the characteristic parameter network training unit is connected with an input interface of the characteristic parameter space domain migration unit. The application trains actual measurement data based on a network, acquires an accurate characteristic parameter model, and then realizes space domain migration of the channel characteristic parameter, so that the consumption of field measurement resources is greatly saved, and the efficiency of channel measurement and simulation reproduction is improved.

Description

Technical Field

[0001] This invention relates to a device for dynamic measurement of wireless channel characteristic parameters and a spatial migration method, which belongs to the field of wireless communication technology. Background Technology

[0002] With the rapid development of wireless communication technology, a wide variety of wireless communication devices, such as mobile phones, mobile computers, and drones, have emerged. To ensure the normal operation of these devices, the modeling and analysis of wireless channels has become increasingly important. Obtaining reliable wireless channel characteristic parameters is a crucial prerequisite for channel modeling and analysis, and is also particularly important for the design and optimization of wireless channel communication systems. However, faced with increasingly complex communication scenarios, traditional channel measurement techniques struggle to obtain comprehensive wireless channel characteristic parameters.

[0003] Due to the influence of reflection and scattering caused by the geographical environment, the signals received by ground receivers in wireless communication typically include components from the direct path, ground reflection, and scattering from surrounding buildings. To effectively acquire omnidirectional wireless channel characteristic parameters while reducing measurement costs and shortening the overall measurement task cycle, an efficient spatial migration method for wireless channel characteristic parameters is needed to achieve cross-spatial prediction and completion of wireless channel characteristic parameters, thereby achieving the goal of acquiring omnidirectional wireless channel characteristic parameters with less measurement work.

[0004] The key to spatial migration of wireless channel feature parameters is how to efficiently obtain a stable feature parameter training network and accurately obtain spatial migration feature parameters through this network. Traditional measurement methods require actual measurement at any point in the measured space, which has good real-time performance and accuracy, but it faces the problem of large-scale manpower and material resource requirements when measuring large-scale scenarios. Summary of the Invention

[0005] This invention provides a dynamic measurement device and spatial migration method for wireless channel characteristic parameters to address the problems existing in the prior art. It can dynamically measure and obtain channel impulse response data in rapidly changing and non-stationary scenarios, and use the data to train a neural network based on a propagation model to obtain a spatial migration model of channel characteristic parameters, thereby realizing the spatial migration of channel characteristic parameters. It is applicable to the dynamic measurement and spatial migration of wireless channel characteristic parameters in complex non-stationary scenarios, and can be used for hardware simulation of wireless channel propagation characteristics, thereby improving the performance testing and evaluation of communication equipment.

[0006] The technical solution adopted in this invention is: a dynamic measurement device for wireless channel feature parameters, comprising a channel measurement unit, a feature parameter extraction unit, a feature parameter network training unit, and a feature parameter spatial domain migration unit;

[0007] The output interface of the channel measurement unit is connected to the feature parameter extraction unit via a PCIe bus. The output interface of the feature parameter extraction unit is connected to the input interfaces of the feature parameter network training unit and the feature parameter spatial domain migration unit, respectively. The output interface of the feature parameter network training unit is connected to the input interface of the feature parameter spatial domain migration unit.

[0008] The present invention also employs the following technical solution: a method for spatial migration of wireless channel characteristic parameters, comprising the following steps:

[0009] The first step involves the user calibrating the measurement system, configuring the channel measurement unit's measurement parameters, and transmitting the channel measurement signal using the channel measurement transmitter module. The data is transmitted through a real channel environment to the channel measurement receiving module, ultimately obtaining wireless channel measurement data.

[0010] In the second step, the feature parameter extraction unit uses the preprocessing module to process the channel measurement data through noise cancellation and system response cancellation, and then inputs them into the power delay estimation module to obtain the delay of the signals received by different antennas. and power Then it is passed to the ray angle estimation module to obtain the multipath angle based on geometric geographic information. ;

[0011] The third step involves passing the time delay power and multipath angle parameters through the path discriminator module to construct networks according to different path types such as direct path, ground reflection path, and other scattering paths. These networks are then fed into the ray power training module and ray angle offset training module of the feature parameter network training unit to train and obtain a stable parameter estimation network.

[0012] The fourth step involves training a stable set of tap coefficients in the ray power training module and a stable set of angle bias parameters in the ray angle bias training module. These are then fed into the ray power transfer module and ray angle transfer module of the feature parameter spatial transfer unit, respectively. The coordinates of the two points to be predicted are obtained through the geometric geographic coordinate extraction module. The time delay parameters of the required predicted coordinates are then obtained through the time delay extraction module and fed into the ray power transfer module. The mean ray angle is obtained through the ray angle mean extraction module and fed into the ray angle transfer module. Finally, the ray power and ray angle parameters between any two point coordinates are obtained, which is the final output result.

[0013] Furthermore, the steps for obtaining the feature parameters in the second step are as follows:

[0014] 2.1) Obtain various delays The corresponding angle parameters and power, where power is directly extracted from channel measurement data after eliminating the influence of the hardware system itself. , No. Path Angle of arrival, azimuth, and pitch at any given time Obtained using the following formula:

[0015]

[0016] in, Initialize the matrix for the angle parameters. For the first The coherent moment, the first Each receiving antenna array element The nth discrete time delay value of the channel impulse response matrix at time step;

[0017] 2.2) Initialize the channel feature parameter set ,in, , The first Power, time delay, angle of arrival (azimuth), and angle of arrival (pitch) of the path;

[0018] 2.3). Using the aforementioned channel feature parameter set The first element is the cluster center, and the cluster center and the channel feature parameter set are calculated. The distance between other elements in the set is used to merge elements that are less than a threshold into a new subset. The distance is calculated using the following formula:

[0019]

[0020] in, , and The first The element and the first The relationship between the elements regarding the angle of arrival Distance, regarding latency Distance and power distance, It is a positive integer;

[0021] 2.4) Delete the aforementioned channel feature parameter set. In and the subset For the same elements, update the channel feature parameter set. Repeat step 2.3 until the updated channel feature parameter set is obtained. If the set is empty, denote the set of all obtained subsets as the subset set. ,in The largest subset number;

[0022] 2.5). Delete the subset set. The subsets whose average power is less than the power threshold are deleted, and the resulting subset set is denoted as the subset set. .

[0023] Furthermore, the specific implementation steps of the third step are as follows:

[0024] 3.1). The specific implementation method of the ray power training module is as follows:

[0025] 1) Based on the subset set The parameter set is divided into three categories: direct path, ground reflection path, and other scattering paths;

[0026] 2) Determine the neural network structure and scale. The neural network includes an input layer, which receives a subset of time delay parameters. Following the input layer, the neural network consists of layers with... The neural network includes a hidden layer of neurons, and further includes an output layer following the hidden layer, the output layer being represented as follows:

[0027]

[0028] in, This represents the time delay parameters for each path. This represents the weight of each path. Let be the activation function of the output layer. is the activation function for the hidden layer.

[0029] 3) Determine the error function of the ray power training module, wherein the error function is:

[0030]

[0031] 4) Adjust the weights using the gradient descent method. When the error function value tends to stabilize and reaches its minimum, the network training is stable.

[0032] 3.2). The specific implementation method of the ray angle offset training module is as follows:

[0033] 1) Based on the subset set The parameter set is divided into three categories: direct path, ground reflection path, and other scattering paths;

[0034] 2) Obtain the mean angle of multipath based on geometric geographic relationships. First, obtain the latitude and longitude coordinates of points A and B using the geometric geographic coordinate extraction module. Then calculate the mean angle, the azimuth angle of point B relative to point A. The calculation is as follows:

[0035]

[0036] in, Let A be the angle between two points relative to the Earth's center, and let B be the pitch angle of point A relative to point B. The calculation is as follows:

[0037]

[0038] Where O represents the Earth's center. Add the Earth's radius to the altitude of point A.

[0039] 3) Calculate the angle offset within all sets by combining the true estimated ray angle with the mean ray angle. Use the ray angle offset training module to train the true multipath angle offset of the above subset sets to obtain the statistical distribution of angle offset of different multipaths, thereby generating more data that conforms to the same statistical law.

[0040] The network loss function for the ray angle bias training module is as follows:

[0041]

[0042] in, It is the expectation function.

[0043] Furthermore,

[0044] 4.1) Input the coordinates between the two points to be predicted into the ray angle mean extraction module and the time delay extraction module respectively. The ray angle mean is calculated as shown in equations (5) and (6); the time delay is calculated as follows:

[0045]

[0046]

[0047]

[0048] in, For the Earth's radius, It is the speed of light.

[0049] 4.2). Calculate the average ray angle. Input the ray angle migration module to obtain the ray angle.

[0050]

[0051] The calculated time delay data is input into the ray power transfer module to obtain the power.

[0052] . .

[0053] The present invention has the following beneficial effects:

[0054] (1) This invention realizes channel measurement and spatial migration of characteristic parameters in fast time-varying non-stationary scenarios, and is particularly suitable for applications such as base station deployment, wireless network optimization, and hardware simulation and reproduction of wireless channels.

[0055] (2). This invention obtains an accurate feature parameter model by training the measured data on the network, and then realizes the spatial migration of the channel feature parameters, which greatly saves the consumption of field measurement resources and improves the efficiency of channel measurement and simulation reproduction. Attached Figure Description

[0056] Figure 1 This is a schematic diagram of the overall implementation of an exemplary dynamic measurement device for wireless channel characteristic parameters according to an embodiment of the present invention.

[0057] Figure 2 This is a simplified flowchart illustrating an exemplary wireless channel feature parameter spatial migration method according to an embodiment of the present invention. Detailed Implementation

[0058] The invention will now be further described with reference to the accompanying drawings.

[0059] The wireless channel feature parameter dynamic measurement device of the present invention includes a channel measurement unit 1-1, a feature parameter extraction unit 1-2, a feature parameter network training unit 1-3, and a feature parameter spatial domain migration unit 1-4.

[0060] The channel measurement unit 1-1 is used to acquire channel measurement data. The feature parameter extraction unit 1-2 is used to preprocess the channel measurement data, then extract feature parameters and pass them to the feature parameter network training unit 1-3 and the feature parameter spatial domain transfer unit 1-4. The feature parameter network training unit 1-3 uses the feature parameters as training data for the parameter network. Based on the data passed from the feature parameter extraction unit 1-2, it trains the ray power training module and the ray angle offset training module, and finally obtains a stable set of tap coefficients and angle offset parameters, which are then passed to the feature parameter spatial domain transfer unit 1-4. The feature parameter spatial domain transfer unit 1-4 is used to achieve spatial domain transfer of feature parameters between any two points in space based on geometric geographic coordinates, trained stable network tap coefficients, and the set of angle offset parameters.

[0061] The channel measurement unit 1-1 includes a channel measurement transmitting module and a channel measurement receiving module. The channel measurement transmitting module is used to send a channel measurement signal according to the input channel measurement parameters. The channel measurement receiving module is used to receive the channel measurement signal and obtain the raw channel measurement data after analog-to-digital conversion.

[0062] The feature parameter extraction unit 1-2 includes a preprocessing module, a power delay estimation module, a ray angle estimation module, a geometric geographic coordinate extraction module, and a path discriminator module. The preprocessing module performs signal processing on the channel measurement data to eliminate additive noise and system response coefficients, and classifies and labels different antennas and different times, then passes this data to the power delay estimation module for parameter estimation. The power delay estimation module extracts delay power parameters based on the preprocessed channel measurement data. The geometric geographic coordinate extraction module obtains the geographic coordinates of the transceiver and receiver, and uses a digital map to obtain the geographic coordinates of any point in space. The ray angle estimation module estimates the ray angle based on the power delay parameters and geometric geographic coordinates, obtaining the ray angle parameters. The path discriminator module determines the network model type based on the above parameters, achieving classification modeling.

[0063] The feature parameter network training units 1-3 include a ray power training module and a ray angle bias training module. The ray power training module is used to establish networks for parameters of different path types based on the time delay power feature parameters and the discrimination results of the path discriminator module, training a stable time delay power relationship model, and then transferring the model tap coefficients to the feature parameter spatial domain transfer units 1-4. The ray angle bias training module is used to establish networks for parameters of different path types based on the ray angle feature parameters and the discrimination results of the path discriminator module, training a stable angle bias statistical distribution model, obtaining a stable ray angle bias parameter set, and transferring it to the feature parameter spatial domain transfer units 1-4.

[0064] The feature parameter spatial domain transfer units 1-4 include a ray angle mean extraction module, a time delay extraction module, a ray angle transfer module, and a ray power transfer module. The angle mean extraction module calculates the ray angle mean based on the transmitter / receiver parameters and all scatterer geographic coordinate parameters input from the geometric geographic coordinate extraction module, and outputs it to the ray angle transfer module. The time delay extraction module calculates the theoretical time delay parameter based on the transmitter / receiver parameters and all scatterer geographic coordinate parameters input from the geometric geographic coordinate extraction module, and outputs it to the ray power transfer module. The ray angle transfer module calculates and outputs the ray angle transfer value based on the ray angle mean extracted from the ray angle mean module and the ray angle bias parameter set input from the ray angle bias training module. The ray power transfer module calculates and outputs the power transfer value based on the time delay parameter input from the time delay extraction module and the model tap coefficients input from the ray power training module, using network forward propagation.

[0065] A method for spatial migration of wireless channel characteristic parameters includes the following steps:

[0066] The first step involves the user calibrating the measurement system, configuring the measurement parameters for channel measurement unit 1-1, and transmitting channel measurement signals using the channel measurement transmitter module. The data is transmitted through a real channel environment to the channel measurement receiving module, ultimately obtaining wireless channel measurement data.

[0067] In the second step, the feature parameter extraction unit 1-2 uses the preprocessing module to process the channel measurement data through noise cancellation and system response cancellation, and then inputs them into the power delay estimation module to obtain the delay of the signals received by different antennas. and power Then it is passed to the ray angle estimation module to obtain the multipath angle based on geometric geographic information. ;

[0068] The third step involves passing the time delay power and multipath angle parameters through the path discriminator module, constructing networks according to different path types such as direct path, ground reflection path, and other scattering paths, and then inputting them into the ray power training module and ray angle offset training module of feature parameter network training units 1-3. Finally, a stable parameter estimation network is trained to obtain the network.

[0069] The fourth step involves training a stable set of tap coefficients in the ray power training module and a stable set of angle bias parameters in the ray angle bias training module. These are then fed into the ray power transfer module and ray angle transfer module of the feature parameter spatial transfer units 1-4, respectively. The coordinates of the two points to be predicted are obtained through the geometric geographic coordinate extraction module. The time delay parameters of the required predicted coordinates are then obtained through the time delay extraction module and fed into the ray power transfer module. The mean ray angle is obtained through the ray angle mean extraction module and fed into the ray angle transfer module. Finally, the ray power and ray angle parameters between any two point coordinates are obtained, which is the final output result.

[0070] Furthermore, the steps for obtaining the feature parameters in the second step are as follows:

[0071] 2.1) Obtain various delays The corresponding angle parameters and power, where power is directly extracted from channel measurement data after eliminating the influence of the hardware system itself. , No. Path Angle of arrival, azimuth, and pitch at any given time Obtained using the following formula:

[0072]

[0073] in, Initialize the matrix for the angle parameters. For the first The coherent moment, the first Each receiving antenna array element The nth discrete time delay value of the channel impulse response matrix at time step;

[0074] 2.2) Initialize the channel feature parameter set ,in, , The first Power, time delay, angle of arrival (azimuth), and angle of arrival (pitch) of the path;

[0075] 2.3) Using the aforementioned channel feature parameter set The first element is the cluster center, and the cluster center and the channel feature parameter set are calculated. The distance between other elements in the set is used to merge elements that are less than a threshold into a new subset. The distance is calculated using the following formula:

[0076]

[0077] in, , and The first The element and the first The relationship between the elements regarding the angle of arrival Distance, regarding latency Distance and power distance, It is a positive integer;

[0078] 2.4) Delete the aforementioned channel feature parameter set In and the subset For the same elements, update the channel feature parameter set. Repeat step 2.3 until the updated channel feature parameter set is obtained. If the set is empty, denote the set of all obtained subsets as the subset set. ,in The largest subset number;

[0079] 2.5) Delete the subset set The subsets whose average power is less than the power threshold are deleted, and the resulting subset set is denoted as the subset set. .

[0080] Furthermore, the specific implementation steps of the third step are as follows:

[0081] 3.1) The specific implementation method of the ray power training module is as follows:

[0082] 1) Based on the subset set The parameter set is divided into three categories: direct path, ground reflection path, and other scattering paths;

[0083] 2) Determine the neural network structure and scale. The neural network includes an input layer, which receives a subset of time delay parameters. Following the input layer, the neural network consists of layers with... The neural network includes a hidden layer of neurons, and further includes an output layer following the hidden layer, the output layer being represented as follows:

[0084]

[0085] in, This represents the time delay parameters for each path. This represents the weight of each path. Let be the activation function of the output layer. is the activation function for the hidden layer.

[0086] 3) Determine the error function of the ray power training module, wherein the error function is:

[0087]

[0088] 4) Adjust the weights using the gradient descent method. When the error function value tends to stabilize and reaches its minimum, the network training is stable.

[0089] 3.2) The specific implementation method of the ray angle offset training module is as follows:

[0090] 1) Based on the subset set The parameter set is divided into three categories: direct path, ground reflection path, and other scattering paths;

[0091] 2) Obtain the mean angle of multipath based on geometric geographic relationships. First, obtain the latitude and longitude coordinates of points A and B using the geometric geographic coordinate extraction module. Then calculate the mean angle, the azimuth angle of point B relative to point A. The calculation is as follows:

[0092]

[0093] in, Let A be the angle between two points relative to the Earth's center, and let B be the pitch angle of point A relative to point B. The calculation is as follows:

[0094]

[0095] Where O represents the Earth's center. Add the Earth's radius to the altitude of point A.

[0096] 3) Calculate the angle offset within all sets by combining the true estimated ray angle with the mean ray angle. Use the ray angle offset training module to train the true multipath angle offset of the above subset sets to obtain the statistical distribution of angle offset of different multipaths, thereby generating more data that conforms to the same statistical law.

[0097] The network loss function for the ray angle bias training module is as follows:

[0098]

[0099] in, It is the expectation function.

[0100] Furthermore, the specific implementation steps of the fourth step are as follows:

[0101] 4.1) Input the coordinates between the two points to be predicted into the ray angle mean extraction module and the time delay extraction module respectively. The ray angle mean is calculated as shown in equations (5) and (6); the time delay is calculated as follows:

[0102]

[0103]

[0104]

[0105] in, For the Earth's radius, It is the speed of light.

[0106] 4.2) Calculate the average value of the ray angles. Input the ray angle migration module to obtain the ray angle.

[0107]

[0108] The calculated time delay data is input into the ray power transfer module to obtain the power.

[0109] .

[0110] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions are described more clearly and completely below. The measurement frequency is set to 27 GHz, the measurement bandwidth to 1 GHz, and the detection signal sequence length to 4096.

[0111] The specific implementation steps are as follows:

[0112] The first step is for the user to configure the measurement parameters of the channel measurement unit 1-1, including the measurement frequency. Signal sampling rate Measure channel sequence length Channel measurement signals are transmitted using a channel measurement transmission module. The channel measurement receiving module obtains channel measurement data and transmits the measurement data to the feature parameter extraction unit 1-2 via the PCIE bus;

[0113] In the second step, feature parameter extraction units 1-2 use the preprocessing module to organize the channel measurement data and then input it into the power delay estimation module to obtain the delay of the signals received by different antennas. power Then, the data is passed to the angle estimation module, which combines the geometric coordinates obtained by the geometric geographic coordinate extraction module to obtain the multipath angle. ;

[0114] Furthermore, the steps for obtaining the feature parameters in the second step are as follows:

[0115] 2.1) Obtain the ray angle parameters and power corresponding to each time delay. Power is directly extracted from channel measurement data after eliminating the influence of the hardware system itself. Path Angle of arrival, azimuth, and pitch at any given time According to equation (25):

[0116] The time delay, power, and ray angle parameters of a single slice of the channel impulse response are shown in Table 1 below;

[0117] Table 1 Characteristic parameters of channel impulse response

[0118]

[0119]

[0120] 2.2) Initialize the channel feature parameter set ,in, , The first Power, time delay, angle of arrival (azimuth), and angle of arrival (pitch) of the path;

[0121] 2.3). Using the aforementioned channel feature parameter set The first element is the cluster center, and the cluster center and the channel feature parameter set are calculated. The distance between other elements in the set is used to merge elements that are less than a threshold into a new subset. The distance is calculated according to the following formula (26);

[0122] 2.4) Delete the aforementioned channel feature parameter set. In and the subset For the same elements, update the channel feature parameter set. Repeat step 2.3 until the updated channel feature parameter set is obtained. If the set is empty, denote the set of all obtained subsets as the subset set. ;

[0123] 2.5). Delete the subset set. The subsets whose average power is less than the power threshold are deleted, and the resulting subset set is denoted as the subset set. The number of subsets can be obtained after processing. As shown in Table 2:

[0124] Table 2 Characteristic parameters of channel impulse response

[0125]

[0126]

[0127]

[0128]

[0129] The third step involves passing the time delay power and multipath angle parameters through the path discriminator module, constructing networks according to different path types such as direct path, ground reflection path, and other scattering paths, and then inputting them into the ray power training module and ray angle offset training module of feature parameter network training units 1-3. Finally, a stable parameter estimation network is trained to obtain the network.

[0130] Furthermore, the specific implementation steps of the third step are as follows:

[0131] 3.1). The specific implementation method of the ray power training module is as follows:

[0132] 1) Based on the subset set, set the parameter set as follows: For direct shot diameter, Ground reflection radius, For other scattering paths;

[0133] 2) Determine the structure and scale of the neural network, which includes an input layer, the input layer being fed a time delay parameter of a subset of inputs, a hidden layer with 20 neurons following the input layer, and an output layer following the hidden layer.

[0134] 3) Determine the error function of the neural network as shown in equation (28);

[0135] 4) Adjust the weights using the gradient descent method. When the error function value tends to stabilize and reaches Network training is stable.

[0136] 3.2). The specific implementation method of the ray angle offset training module is as follows:

[0137] 1) Based on the subset set, set the parameter set as follows: For direct shot diameter, Ground reflection radius, For other scattering paths;

[0138] 2) Obtain the mean angle of multipath based on geometric geographic relationships. First, obtain the latitude and longitude coordinates of points A and B using the geometric geographic coordinate extraction module. Then calculate the mean angle (as shown in Table 3), the azimuth angle of point B relative to point A. Calculate as shown in equations (29) and (30);

[0139] Table 3 Mean Angles

[0140]

[0141] 3) Calculate the angle offset within all sets by combining the true estimated ray angle with the mean ray angle. Use the ray angle offset training module to train the true multipath angle offset of the above subset sets to obtain the statistical distribution of angle offset of different multipaths, and then generate more data that conform to the same statistical law as shown in Table 4.

[0142] Table 4 Angle Offset Parameter Set

[0143]

[0144]

[0145]

[0146] The fourth step involves training the ray power training module to obtain stable tap coefficients and the ray angle offset training module to obtain a stable set of angle offset parameters. These are then fed into the ray power transfer module and the ray angle transfer module of the feature parameter spatial transfer units 1-4, respectively. The coordinates of the two points to be predicted are obtained through the geometric geographic coordinate extraction module. The time delay parameters of the required predicted coordinates are then obtained through the time delay extraction module and fed into the ray power transfer module. The ray angle mean is obtained through the ray angle mean extraction module and fed into the ray angle transfer module. This process yields the time delay power and ray angle parameters between any two coordinates.

[0147] Furthermore, the specific implementation steps of the fourth step are as follows:

[0148] 4.1) Predict the coordinates of points A and B to be predicted. and the coordinates of the scattering point The values ​​are passed to the ray angle mean extraction module and the time delay extraction module respectively. The ray angle mean is calculated as shown in equations (29) and (30), and the results are obtained.

[0149]

[0150] The delay is calculated as shown in equations (32), (33), and (34), from which we can obtain...

[0151]

[0152] 4.2) Input the calculated average ray angle into the ray angle migration module to obtain the ray angle.

[0153]

[0154] The calculated time delay data is input into the ray power transfer module to obtain the power.

[0155] Model distribution Delay (s) Power (dB) direct caliber 1.374694462406294e-06 -28.374000000000 Ground reflection path 1.377803994712944e-06 -29.655000000000 Scattering path 2.104619615400989e-06 -43.857000000000

[0156] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A method for spatial migration of wireless channel characteristic parameters, characterized in that: Includes the following steps: The first step involves the user calibrating the measurement system, configuring the measurement parameters of the channel measurement unit (1-1), and transmitting the channel measurement signal using the channel measurement transmission module. The data is transmitted through a real channel environment to the channel measurement receiving module, ultimately obtaining wireless channel measurement data. In the second step, the feature parameter extraction unit (1-2) uses the preprocessing module to process the channel measurement data through noise cancellation and system response cancellation, and then inputs them into the power delay estimation module to obtain the delay of the signals received by different antennas. and power Then it is passed to the ray angle estimation module to obtain the multipath angle based on geometric geographic information. ; The third step involves passing the time delay power and multipath angle parameters through the path discriminator module to construct networks according to different path types such as direct path, ground reflection path, and other scattering paths. These networks are then fed into the ray power training module and ray angle offset training module of the feature parameter network training unit (1-3) to train and obtain a stable parameter estimation network. The fourth step involves training the ray power training module to obtain stable tap coefficients and the ray angle offset training module to obtain a stable set of angle offset parameters. These are then fed into the ray power transfer module and the ray angle transfer module of the feature parameter spatial transfer unit (1-4), respectively. The coordinates of the two points to be predicted are obtained through the geometric geographic coordinate extraction module. The time delay parameters of the required predicted coordinates are then obtained through the time delay extraction module and fed into the ray power transfer module. The ray angle mean is obtained through the ray angle mean extraction module and fed into the ray angle transfer module. This process yields the ray power and ray angle parameters between any two coordinates, resulting in the final output. The specific steps for the third step are as follows: 3.1). The specific implementation method of the ray power training module is as follows: 1) Based on channel characteristic parameters The parameter set is divided into three categories: direct path, ground reflection path, and other scattering paths; 2) Determine the neural network structure and scale. The neural network includes an input layer, which receives a subset of time delay parameters. Following the input layer, the neural network consists of layers with... The neural network includes a hidden layer of neurons, and further includes an output layer following the hidden layer, the output layer being represented as follows: in, This represents the time delay parameters for each path. This represents the weight of each path. Let be the activation function of the output layer. Let be the activation function of the hidden layer; 3) Determine the error function of the ray power training module, wherein the error function is: 4) Adjust the weights using the gradient descent method. When the error function value tends to stabilize and reaches its minimum, the network training is stable; 3.2). The specific implementation method of the ray angle offset training module is as follows: 1) Based on channel characteristic parameters The parameter set is divided into three categories: direct path, ground reflection path, and other scattering paths; 2) Obtain the mean angle of multipath based on geometric geographic relationships. First, obtain the latitude and longitude coordinates of points A and B using the geometric geographic coordinate extraction module. Then calculate the mean angle, the azimuth angle of point B relative to point A. The calculation is as follows: in, Let A be the angle between two points relative to the Earth's center, and let B be the pitch angle of point A relative to point B. The calculation is as follows: Where O represents the Earth's center. Add the Earth's radius to the altitude of point A; 3) Calculate the angle offset within all sets by combining the true estimated ray angle with the mean ray angle. Use the ray angle offset training module to train the true multipath angle offset of the above subset sets to obtain the statistical distribution of angle offset of different multipaths, thereby generating more data that conforms to the same statistical law. The network loss function for the ray angle bias training module is as follows: in, It is the expectation function.

2. A method for spatial migration of wireless channel characteristic parameters as described in claim 1, characterized in that: The steps to obtain the feature parameters in the second step are as follows: 2.1) Obtain various delays The corresponding angle parameters and power, where power is directly extracted from channel measurement data after eliminating the influence of the hardware system itself. , No. Path Angle of arrival, azimuth, and pitch at any given time Obtained using the following formula: in, Initialize the matrix for the angle parameters. For the first The coherent moment, the first Each receiving antenna array element The nth discrete time delay value of the channel impulse response matrix at time step; 2.2) Initialize the channel feature parameter set ,in, , The first Power, time delay, angle of arrival (azimuth), and angle of arrival (pitch) of the path; 2.3). Using the aforementioned channel feature parameter set The first element is the cluster center, and the cluster center and the channel feature parameter set are calculated. The distance between other elements in the set is used to merge elements that are less than a threshold into a new subset. The distance is calculated using the following formula: in, , and The first The element and the first The relationship between the elements regarding the angle of arrival Distance, regarding latency Distance and power distance, It is a positive integer; 2.4) Delete the aforementioned channel feature parameter set. In and the subset For the same elements, update the channel feature parameter set. Repeat step 2.3 until the updated channel feature parameter set is obtained. If the set is empty, denote the set of all obtained subsets as the subset set. ,in The largest subset number; 2.5). Delete the subset set. The subsets whose average power is less than the power threshold are deleted, and the resulting subset set is denoted as the subset set. .

3. A method for spatial migration of wireless channel characteristic parameters as described in claim 2, characterized in that: 4.1) Input the coordinates between the two points to be predicted into the ray angle mean extraction module and the time delay extraction module respectively. The ray angle mean is calculated as shown in equations (5) and (6); the time delay is calculated as follows: in, For the Earth's radius, The speed of light; 4.2). Calculate the average ray angle. Input the ray angle migration module to obtain the ray angle. The calculated time delay data is input into the ray power transfer module to obtain the power. . 。 4. A dynamic measurement device for wireless channel characteristic parameters used in the spatial migration method for wireless channel characteristic parameters as described in claim 1, characterized in that: It includes a channel measurement unit (1-1), a feature parameter extraction unit (1-2), a feature parameter network training unit (1-3), and a feature parameter spatial domain migration unit (1-4). The output interface of the channel measurement unit (1-1) is connected to the feature parameter extraction unit (1-2) via a PCIE bus. The output interface of the feature parameter extraction unit (1-2) is connected to the input interfaces of the feature parameter network training unit (1-3) and the feature parameter spatial domain migration unit (1-4), respectively. The output interface of the feature parameter network training unit (1-3) is connected to the input interface of the feature parameter spatial domain migration unit (1-4).

5. The dynamic measurement device for wireless channel characteristic parameters as described in claim 4, characterized in that: The channel measurement unit (1-1) is used to acquire channel measurement data. The feature parameter extraction unit (1-2) is used to preprocess the channel measurement data, then extract feature parameters and pass them to the feature parameter network training unit (1-3) and the feature parameter spatial domain migration unit (1-4). The feature parameter network training unit (1-3) is used to use the feature parameters as the training set data of the parameter network. Based on the data passed in by the feature parameter extraction unit (1-2), it trains the ray power training module and the ray angle offset training module, and finally obtains a stable set of tap coefficients and angle offset parameters and passes them to the feature parameter spatial domain migration unit (1-4). The feature parameter spatial domain migration unit (1-4) is used to realize the spatial domain migration of feature parameters between any two points in space according to geometric geographic coordinates, trained stable network tap coefficients and angle offset parameter sets.

6. The dynamic measurement device for wireless channel characteristic parameters as described in claim 5, characterized in that: The channel measurement unit (1-1) includes a channel measurement transmitting module and a channel measurement receiving module. The channel measurement transmitting module is used to send a channel measurement signal according to the input channel measurement parameters. The channel measurement receiving module is used to receive the channel measurement signal and obtain the original channel measurement data after analog-to-digital conversion.

7. The dynamic measurement device for wireless channel characteristic parameters as described in claim 6, characterized in that: The feature parameter extraction unit (1-2) includes a preprocessing module, a power delay estimation module, a ray angle estimation module, a geometric geographic coordinate extraction module, and a path discriminator module. The preprocessing module is used to eliminate additive noise and system response coefficients in the channel measurement data based on signal processing. The power delay estimation module is used to extract delay power parameters based on the preprocessed channel measurement data. The ray angle estimation module is used to estimate the ray angle based on the power delay parameters and geometric geographic coordinates to obtain the ray angle parameters. The path discriminator module determines the network model type based on the above parameters and realizes classification modeling.

8. The dynamic measurement device for wireless channel characteristic parameters as described in claim 7, characterized in that: The feature parameter network training unit (1-3) includes a ray power training module and a ray angle bias training module. The ray power training module is used to establish networks for parameters of different path types according to the time delay power parameters and the discrimination results of the path discriminator module, train a stable time delay power relationship model, and transfer the model tap coefficients to the feature parameter spatial domain transfer unit (1-4). The ray angle bias training module is used to establish networks for parameters of different path types according to the ray angle parameters and the discrimination results of the path discriminator module, train a stable angle bias statistical distribution model, obtain a stable ray angle bias parameter set, and transfer it to the feature parameter spatial domain transfer unit (1-4).

9. The dynamic measurement device for wireless channel characteristic parameters as described in claim 8, characterized in that: The feature parameter spatial migration unit (1-4) includes a ray angle mean extraction module, a time delay extraction module, a ray angle migration module, and a ray power migration module. The angle mean extraction module is used to calculate the ray angle mean based on the transmit / receive end and all scatterer geographic coordinate parameters passed from the geometric geographic coordinate extraction module, and output it to the ray angle migration module. The time delay extraction module is used to calculate the theoretical time delay parameter based on the transmit / receive end and all scatterer geographic coordinate parameters passed from the geometric geographic coordinate extraction module, and output it to the ray power migration module. The ray angle migration module is used to calculate and output the ray angle migration value based on the ray angle mean extracted from the ray angle mean extraction module and the ray angle offset parameter set passed from the ray angle offset training module. The ray power migration module is used to calculate and output the power migration value based on the time delay parameter passed from the time delay extraction module and the model tap coefficients passed from the ray power training module, and the network forward propagation.

Images