Transmit power calibration method, device and equipment for multi-element phased array system

Abstract

The present invention discloses a method, device, and apparatus for calibrating the transmit power of a multi-element phased array system. The method comprises: obtaining transmit signal parameters of the multi-element phased array system currently detected by a sensor system; performing signal feature extraction on the transmit signal parameters to obtain transmit signal features; performing feature fusion on the transmit signal features to obtain fused features; inputting the fused features into a pre-trained transmit power prediction model to predict the current potential deviation of the transmit power of the multi-element phased array system; and calculating the corrected transmit power of the multi-element phased array system based on the transmit power and the potential deviation of the transmit power. The present invention can achieve real-time and accurate calibration of the transmit power of a multi-element phased array system.

Description

Technical Field

[0001] The present invention relates to the field of signal processing technology, and in particular to a method, device and equipment for calibrating the transmission power of a multi-element phased array system. Background Art

[0002] In the field of phased array systems, especially multi-element phased array radars or communications systems, accurate control of transmit power is a key factor in ensuring system performance. Traditionally, transmit power calibration for multi-element phased array systems has been based on static preset values or simple feedback loops. This approach often overlooks the complexity and dynamic changes in the actual operating environment, such as temperature fluctuations, hardware aging, and electromagnetic interference, which can cause deviations between the actual transmit power and the preset value. Furthermore, some traditional calibration methods typically involve periodic manual calibration, comparative measurements using a standard signal source, or adjustments via simple built-in calibration circuits. For example, in some multi-element phased array systems, fixed compensation coefficients are used to correct for known power losses, but this only partially addresses the problem and cannot adapt in real time to various unknown or nonlinear power attenuation conditions.

[0003] Therefore, the main problem with existing correction schemes for the transmit power of multi-element phased array systems is that they lack the ability to independently correct dynamic changes in transmit power. They cannot accurately reflect the power deviation of the multi-element phased array system under actual working conditions and cannot accurately correct the transmit power of the multi-element phased array system, thereby affecting the operating reliability of the multi-element phased array system, especially in applications requiring high precision and stability. Summary of the Invention

[0004] Embodiments of the present invention provide a method, apparatus, and device for calibrating the transmit power of a multi-element phased array system, which can achieve real-time and accurate calibration of the transmit power of the multi-element phased array system.

[0005] An embodiment of the present invention provides a method for calibrating transmit power of a multi-element phased array system, comprising:

[0006] Obtaining the transmission signal parameters of the multi-element phased array system currently detected by the sensor system; the transmission signal parameters include transmission power, power spectrum density, energy concentration ratio and phase difference;

[0007] Extracting signal features from the transmission signal parameters to obtain transmission signal features; the transmission signal features include transmission power features, power spectrum density features, energy concentration ratio features, and phase difference features;

[0008] Performing feature fusion on the emission signal features to obtain fusion features;

[0009] Inputting the fusion features into a pre-trained transmit power prediction model to predict the current potential deviation of the transmit power of the multi-element phased array system;

[0010] The corrected transmit power of the multi-element phased array system is calculated according to the transmit power and the potential transmit power deviation.

[0011] As an improvement to the above solution, the power spectrum density is: perform fast Fourier transform (FFT) on the signal to obtain the spectrum, and then calculate the square of its amplitude: S(f) is the power spectral density of signal s(t) at frequency f; is the Fourier transform of the signal s(t), which represents the representation of the signal s(t) in the frequency domain; It is the square of the modulus of the Fourier transform result, which represents the energy or power of the signal at frequency f;

[0012] The energy concentration ratio is: ECR is the energy concentration ratio, which reflects the degree to which the signal energy is concentrated within the main frequency bandwidth; l and f h are the lower and upper frequency limits of the main frequency bandwidth respectively; is the energy or power of the signal within the main frequency bandwidth; is the total energy or power of the signal s(t) in the entire frequency range;

[0013] The phase difference is: Δφ ij =∠s i -∠s j ;Δφ ij is the phase difference between the i-th array element and the j-th array element; ∠s i and ∠s j are the phase angles of the signals of the i-th array element and the j-th array element respectively.

[0014] As an improvement to the above solution, the step of fusing the emission signal features to obtain fused features includes:

[0015] Let x p is the transmit power eigenvector, x psd is the power spectrum density eigenvector, x ecr is the energy concentration ratio eigenvector, x pd is the phase difference eigenvector;

[0016] Scaling the feature vectors of each of the emission signal features using a maximum and minimum scaling method to ensure that all features are at the same magnitude;

[0017] By calculating the contribution of the feature vectors of each of the scaling-processed emission signal features to the overall data variability, the fusion weight of each of the emission signal features is dynamically adjusted; the calculation formula is: I i is the importance score of the i-th feature in the transmitted signal features; the importance score is calculated by a feature importance scoring method;

[0018] performing high-order statistical processing on the eigenvectors of the scaled transmitted signal features to obtain statistics; the high-order statistical processing includes calculating the mean as a first-order moment, the variance as a second-order moment, the skewness as a third-order moment, and the kurtosis as a fourth-order moment of each eigenvector to capture the central tendency, dispersion, skewness, and peakedness of each transmitted signal feature;

[0019] All processed feature vectors and statistics are fused and calculated using the following feature fusion formula:

[0020] Represents the concatenation operation of vectors; x′ i is the eigenvector X i After scaling, the result x i is x p 、x psd 、x ecr 、x pd ; i represents the subscript of p, psd, ecr, pd; I i is the importance score of the i-th feature in the emission signal features; w i is the fusion weight of the i-th feature in the emission signal features;

[0021] γ 1，i , γ 2，i are the mean, variance, skewness and kurtosis of the i-th feature in the emission signal features respectively; x fusion is the fused feature vector.

[0022] As an improvement to the above solution, the pre-trained transmit power prediction model includes:

[0023] The convolution layer extracts the fusion features of the input data through a set of learnable filter kernels. It is connected to the ReLU activation function to process the results of the convolution operation. The convolution operation is expressed as: Y(i, j) = ∑ m，nX(i+m, j+n)*K(m, n); X is the input tensor of the fused feature; K is the convolution kernel; * represents the multiplication operation, which refers to the point-by-point multiplication of the convolution kernel and the elements at the corresponding position of the input tensor; m and n are the indices of the convolution kernel, and the convolution kernel will traverse the entire input tensor and perform convolution calculations; Y(i, j) is the value of the output tensor at position (i, j) after the convolution operation;

[0024] The LSTM recurrent layer takes as input the sequence data output by the convolutional layer and outputs a hidden state vector that contains summary information of the time series data. It includes an input gate, a forget gate, an output gate, and a cell state.

[0025] The fully connected layer, whose input is the last hidden state of the LSTM recurrent layer, is expressed as: [l] =W [l] a [l-1] +b [l] ;z [l] is the linear combination output of the lth layer; W [l] is the weight matrix of the lth layer; a [l-1] is the activation output of the l-1 layer, which serves as the input of the l layer; b [l] is the bias vector of the lth layer; the activation function of the fully connected layer is: a [l] =g(z [l] );a [l] is the activation output of the lth layer, which outputs the predicted potential deviation of the current transmission power of the multi-element phased array system, that is, z after the activation function g [l] ; g is the ReLU, sigmoid or tanh activation function.

[0026] Another embodiment of the present invention provides a transmission power calibration device for a multi-element phased array system, including:

[0027] An acquisition module is used to obtain the transmission signal parameters of the multi-element phased array system currently detected by the sensor system; the transmission signal parameters include transmission power, power spectrum density, energy concentration ratio and phase difference;

[0028] A feature extraction module is used to extract signal features from the transmission signal parameters to obtain transmission signal features; the transmission signal features include transmission power features, power spectrum density features, energy concentration ratio features and phase difference features;

[0029] A feature fusion module, configured to perform feature fusion on the emission signal features to obtain fused features;

[0030] A power deviation prediction module is used to input the fusion feature into a pre-trained transmit power prediction model to predict the current potential deviation of the transmit power of the multi-element phased array system;

[0031] A power correction module is used to calculate the corrected transmit power of the multi-element phased array system according to the transmit power and the potential deviation of the transmit power.

[0032] As an improvement to the above solution, the power spectrum density is: perform fast Fourier transform (FFT) on the signal to obtain the spectrum, and then calculate the square of its amplitude: S(f) is the power spectral density of signal s(t) at frequency f; is the Fourier transform of the signal s(t), which represents the representation of the signal s(t) in the frequency domain; It is the square of the modulus of the Fourier transform result, which represents the energy or power of the signal at frequency f;

[0033] The energy concentration ratio is: ECR is the energy concentration ratio, which reflects the degree to which the signal energy is concentrated within the main frequency bandwidth; l and f h are the lower and upper frequency limits of the main frequency bandwidth respectively; is the energy or power of the signal within the main frequency bandwidth; is the total energy or power of the signal s(t) in the entire frequency range;

[0034] The phase difference is: Δφ ij =∠s i -∠s j ;Δφ ij is the phase difference between the i-th array element and the j-th array element; ∠s i and ∠s j are the phase angles of the signals of the i-th array element and the j-th array element respectively.

[0035] As an improvement to the above solution, the feature fusion module is specifically used to:

[0036] Let x p is the transmit power eigenvector, x psd is the power spectrum density eigenvector, x ecr is the energy concentration ratio eigenvector, x pd is the phase difference eigenvector;

[0037] Scaling the feature vectors of each of the emission signal features using a maximum and minimum scaling method to ensure that all features are at the same magnitude;

[0038] By calculating the contribution of the feature vectors of each of the scaling-processed emission signal features to the overall data variability, the fusion weight of each of the emission signal features is dynamically adjusted; the calculation formula is: I iis the importance score of the i-th feature in the transmitted signal features; the importance score is calculated by a feature importance scoring method;

[0039] performing high-order statistical processing on the eigenvectors of the scaled transmitted signal features to obtain statistics; the high-order statistical processing includes calculating the mean as a first-order moment, the variance as a second-order moment, the skewness as a third-order moment, and the kurtosis as a fourth-order moment of each eigenvector to capture the central tendency, dispersion, skewness, and peakedness of each transmitted signal feature;

[0040] All processed feature vectors and statistics are fused and calculated using the following feature fusion formula:

[0041] Represents the concatenation operation of vectors; x′ i is the eigenvector x i After scaling, the result x i is x p 、x psd 、x ecr 、x pd ; i represents the subscript of p, psd, ecr, pd; Ii is the importance score of the i-th feature in the emission signal feature; is the fusion weight of the i-th feature in the emission signal feature; μ i , γ 1，i , γ 2，i are the mean, variance, skewness and kurtosis of the i-th feature in the emission signal features respectively; x fusion is the fused feature vector.

[0042] As an improvement to the above solution, the pre-trained transmit power prediction model includes:

[0043] The convolution layer extracts the fusion features of the input data through a set of learnable filter kernels. It is connected to the ReLU activation function to process the results of the convolution operation. The convolution operation is expressed as: Y(i,j) = ∑ m,n X(i+m,j+n)*K(m,n); X is the input tensor of the fused feature; K is the convolution kernel; * represents the multiplication operation, which refers to the point-by-point multiplication of the convolution kernel and the elements at the corresponding position of the input tensor; m and n are the indices of the convolution kernel, and the convolution kernel will traverse the entire input tensor and perform convolution calculations; Y(i,j) is the value of the output tensor at position (i,j) after the convolution operation;

[0044] The LSTM recurrent layer takes as input the sequence data output by the convolutional layer and outputs a hidden state vector that contains summary information of the time series data. It includes an input gate, a forget gate, an output gate, and a cell state.

[0045] The fully connected layer, whose input is the last hidden state of the LSTM recurrent layer, is expressed as: [l] =W [l] a [l-1] +b [l] ;z [l] is the linear combination output of the lth layer; W [l] is the weight matrix of the lth layer; a [l-1] is the activation output of the l-1 layer, which serves as the input of the l layer; b [l] is the bias vector of the lth layer; the activation function of the fully connected layer is: a [l] =g(z [l] );a [l] is the activation output of the lth layer, which outputs the predicted potential deviation of the current transmission power of the multi-element phased array system, that is, z after the activation function g [l] ; g is the ReLU, sigmoid or tanh activation function.

[0046] Another embodiment of the present invention provides a transmit power calibration device for a multi-element phased array system, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the transmit power calibration method for the multi-element phased array system described in the above-mentioned embodiment of the invention is implemented.

[0047] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0048] First, the system acquires signal parameters of a multi-element phased array transmitter in real time, including key metrics such as transmit power, power spectral density, energy concentration ratio, and phase difference. Next, it performs feature extraction on these parameters, refining features closely related to transmit power, such as power fluctuation patterns, energy distribution, and phase coherence between elements. Subsequently, an innovative feature fusion algorithm integrates these extracted multi-dimensional features into a comprehensive fused feature, fully accounting for the correlations and weights between features to enhance the model's predictive capabilities. These fused features are then input into a deep learning model, trained with extensive historical data, capable of predicting the potential deviation in transmit power of a multi-element phased array system under current conditions. Finally, based on the predicted deviation, the system automatically adjusts transmit power to ensure that actual output is consistent with expected output, thereby achieving precise power control. Compared to the passive and static nature of traditional correction methods, the embodiments of the present invention, through real-time analysis and intelligent decision-making, can proactively adapt to environmental changes. This effectively addresses the low accuracy and slow response speed of power correction in existing technologies, enabling precise correction of transmit power for multi-element phased array systems in complex and dynamic environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 This is a flow chart of a method for calibrating transmit power of a multi-element phased array system provided by one embodiment of the present invention;

[0050] Figure 2 1 is a schematic structural diagram of a transmission power calibration device for a multi-element phased array system provided by one embodiment of the present invention;

[0051] Figure 3 The present invention is a schematic structural diagram of a transmission power calibration device for a multi-element phased array system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0053] See also Figure 1 , is a flow chart of a method for calibrating the transmit power of a multi-element phased array system provided by one embodiment of the present invention. The method for calibrating the transmit power of a multi-element phased array system comprises steps S10 to S14:

[0054] S10, obtaining transmission signal parameters of the multi-element phased array system currently detected by the sensor system; the transmission signal parameters include transmission power, power spectrum density, energy concentration ratio and phase difference;

[0055] S11, performing signal feature extraction on the transmission signal parameters to obtain transmission signal features; the transmission signal features include transmission power features, power spectrum density features, energy concentration ratio features and phase difference features;

[0056] S12, performing feature fusion on the emission signal features to obtain fused features;

[0057] S13, inputting the fusion feature into a pre-trained transmit power prediction model to predict the current potential deviation of the transmit power of the multi-element phased array system;

[0058] S14: Calculate the corrected transmit power of the multi-element phased array system according to the transmit power and the potential transmit power deviation.

[0059] In this embodiment of the present invention, the system first acquires signal parameters of a multi-element phased array transmitter in real time, including key metrics such as transmit power, power spectral density, energy concentration ratio, and phase difference. Next, it performs feature extraction on these parameters, refining features closely related to transmit power, such as power fluctuation patterns, energy distribution, and phase coherence between elements. Subsequently, an innovative feature fusion algorithm integrates the extracted multi-dimensional features into a comprehensive fused feature, fully considering the correlations and weights between features to enhance the model's predictive capabilities. The fused feature is then input into a deep learning model, which has been trained with extensive historical data and can predict the potential deviation in transmit power of a multi-element phased array system under current conditions. Finally, based on the predicted deviation, the system automatically adjusts the transmit power to ensure that the actual output is consistent with the expected value, thereby achieving precise power control. Compared to the passive and static nature of traditional correction methods, this embodiment of the present invention, through real-time analysis and intelligent decision-making, can actively adapt to environmental changes. This effectively addresses the low accuracy and slow response speed of power correction in existing technologies, achieving precise correction of the transmit power of a multi-element phased array system in complex dynamic environments.

[0060] Specifically, the corrected transmit power of the multi-element phased array system is calculated based on the sum of the transmit power and the potential transmit power deviation. Furthermore, based on the calculated correction amount, a specific power correction instruction is generated to achieve overall power balance and optimization. The correction instruction is sent to the control unit of the multi-element phased array system to perform power adjustment. If the potential transmit power deviation is a negative value, it indicates that the current power is too high and the transmit power needs to be reduced. Conversely, if the potential transmit power deviation is a positive value, the transmit power needs to be increased.

[0061] For example, transmit power refers to the actual transmit power level of each array element in the system. Power sensors measure the instantaneous power of each array element and then aggregate this data to understand the power distribution of the entire system. Power spectral density (PSD): Power spectral density provides the energy distribution of a signal in the frequency domain. A spectrum analyzer estimates the PSD by performing a fast Fourier transform (FFT) on the signal, revealing the power levels of different frequency components. The PSD helps identify the degree of power concentration at different frequencies. Energy concentration ratio (ECR): ECR measures the proportion of signal energy concentrated within the primary frequency bandwidth. It is calculated by comparing the total power within the primary frequency bandwidth with the total power of the entire signal spectrum and helps assess the signal's spectral purity and power efficiency. Phase difference: Phase difference is a measure of the signal phase relationship between different array elements, reflecting phase coherence. Phase detectors can measure the phase difference between adjacent elements or selected pairs of elements, which is crucial for maintaining phase synchronization and power combining in phased array systems.

[0062] Specifically, the power spectrum density is: perform fast Fourier transform (FFT) on the signal to obtain the spectrum, and then calculate the square of its amplitude: S(f) is the power spectral density of signal s(t) at frequency f; is the Fourier transform of the signal s(t), which represents the representation of the signal s(t) in the frequency domain; It is the square of the modulus of the Fourier transform result, which represents the energy or power of the signal at frequency f;

[0063] The energy concentration ratio is: ECR is the energy concentration ratio, which reflects the degree to which the signal energy is concentrated within the main frequency bandwidth; l and f h are the lower and upper frequency limits of the main frequency bandwidth respectively; is the energy or power of the signal within the main frequency bandwidth; is the total energy or power of the signal s(t) in the entire frequency range;

[0064] The phase difference is: Δφ ij =∠s i -∠s j ;Δφ ij is the phase difference between the i-th array element and the j-th array element; ∠s i and ∠s j are the phase angles of the signals of the i-th array element and the j-th array element respectively.

[0065] After collecting the above-mentioned transmission signal parameters, the next step is to extract signal features from these parameters to convert them into higher-level feature representations to facilitate subsequent deep learning model processing. The goal of signal feature extraction is to extract the most representative features that best reflect the system status from the raw data. Specifically: Transmit power features: Extract features such as instantaneous power or power fluctuations from the transmit power data. These features reflect the overall level and stability of the power. Power spectral density features: Extract features such as the main frequency, bandwidth, and power at the peak frequency from the PSD. These features provide in-depth insights into the signal spectrum characteristics. Energy concentration ratio features: Calculate the ECR value and the possible ECR change rate to reflect the stability and consistency of the signal energy distribution. Phase difference features: Quantify the statistical characteristics of the phase difference, such as the average phase difference, the standard deviation of the phase difference, or the distribution of the phase difference. These features help diagnose phase synchronization problems between array elements.

[0066] The signal feature extraction process can utilize existing feature extraction methods, including signal preprocessing (such as denoising and filtering), feature engineering (such as statistical analysis and spectral analysis), and feature selection (identifying the most informative features). These features are then fused to form a comprehensive feature vector, which serves as input to the transmit power prediction model. This approach enables accurate monitoring and prediction of the transmit power status of multi-element phased array systems, even in complex and dynamic environments, providing a solid foundation for subsequent power correction.

[0067] As an example, the performing feature fusion on the emission signal features to obtain fused features includes:

[0068] Let x p is the transmit power eigenvector, x psd is the power spectrum density eigenvector, x ecr is the energy concentration ratio eigenvector, x pd is the phase difference eigenvector;

[0069] Scaling the feature vectors of each of the emission signal features using a maximum and minimum scaling method to ensure that all features are at the same magnitude;

[0070] By calculating the contribution of the feature vectors of each of the scaling-processed emission signal features to the overall data variability, the fusion weight of each of the emission signal features is dynamically adjusted; the calculation formula is: I i is the importance score of the i-th feature in the transmitted signal features; the importance score is calculated by a feature importance scoring method;

[0071] performing high-order statistical processing on the eigenvectors of the scaled transmitted signal features to obtain statistics; the high-order statistical processing includes calculating the mean as a first-order moment, the variance as a second-order moment, the skewness as a third-order moment, and the kurtosis as a fourth-order moment of each eigenvector to capture the central tendency, dispersion, skewness, and peakedness of each transmitted signal feature;

[0072] All processed feature vectors and statistics are fused and calculated using the following feature fusion formula:

[0073] Represents the concatenation operation of vectors; x′ i is the eigenvector x i After scaling, the result x i is x p 、x psd 、x ecr 、x pd ; i represents the subscript of p, psd, ecr, pd; I i is the importance score of the i-th feature in the emission signal features; w i is the fusion weight of the i-th feature in the emission signal feature; μ i , γ 1,i ,γ 2,i are the mean, variance, skewness and kurtosis of the i-th feature in the emission signal features respectively; x fusion is the fused feature vector.

[0074] In this embodiment, a comprehensive feature fusion mechanism is constructed to transform the transmit signal parameters of a multi-element phased array system into fused features that can be effectively processed by a deep learning model, enabling accurate prediction of potential transmit power deviation. Specifically, the extracted transmit power, power spectral density, energy concentration ratio, and phase difference features are first represented as vectors. These features are then scaled to their magnitudes through maximum and minimum scaling, ensuring balanced influence between different physical quantities. Next, a dynamic weight adjustment strategy is employed to calculate the importance score of each feature based on its contribution to the overall data variability. Fusion weights are then adjusted to assign higher weights to features that contribute more significantly to model prediction, enhancing the model's decision-making basis. Subsequently, high-order statistical processing is used to further explore the deeper information of the feature vectors, calculating the mean, variance, skewness, and kurtosis to comprehensively reflect the statistical properties of the transmitted signal. Finally, a feature fusion formula is used to integrate the scaled feature vectors, dynamically adjusted fusion weights, and high-order statistics into a fused feature vector, which serves as input to the deep learning model for predicting potential transmit power deviation. This embodiment not only solves the problems of inconsistent feature magnitudes, unreasonable weight distribution, and low information utilization in traditional power correction methods, but also effectively improves the model's ability to recognize complex signal patterns, ensures the accuracy and real-time performance of transmit power correction of multi-element phased array systems in dynamic environments, and enhances the stability and performance of the system in practical applications.

[0075] It can be understood that feature importance scoring is a method for evaluating the contribution of each feature in a machine learning model to the prediction target. Feature importance scoring is used to dynamically adjust the fusion weights to ensure that the features that contribute most to the overall data variability and prediction results have a more important position in the fusion process. As an example, the feature importance score can be calculated as follows:

[0076] Gradient-based feature importance scoring: The importance of features is evaluated by calculating the gradient of the model output with respect to the input features. Specifically: Calculate the gradient: For each input feature, calculate the partial derivative of the model output (such as the predicted transmit power deviation) with respect to the feature, which can be done during the model training process using the backpropagation algorithm. The absolute value or square of the gradient: The absolute value or square of the gradient is usually used as a measure of feature importance, because a large gradient means that the model output is sensitive to small changes in the corresponding feature, that is, the feature has a greater impact on the model output. Averaging or integration: In order to obtain a stable feature importance score, the gradient measures of all samples can be averaged or integrated.

[0077] Feature importance scores based on tree models: For example, they are calculated based on Random Forest or Gradient Boosting Trees. In tree models, feature importance can be calculated in two ways:

[0078] Based on the reduction of impurity (Gini Importance or Mean Decrease Impurity): Each time a feature is used to split a node, it reduces the impurity of the child node (such as Gini impurity or entropy). The feature importance is the average of the impurity reduction of the feature in all trees.

[0079] Based on the decrease in prediction accuracy (Mean Decrease Accuracy or Permutation Importance): By randomly perturbing the value of a single feature and then re-evaluating the model's prediction accuracy. The feature importance is the decrease in accuracy before and after the permutation.

[0080] As an improvement to the above solution, the pre-trained transmit power prediction model includes:

[0081] The convolution layer extracts the fusion features of the input data through a set of learnable filter kernels. It is connected to the ReLU activation function to process the results of the convolution operation. The convolution operation is expressed as: Y(i, j) = ∑ m，n X(i+m, j+n)*K(m, n); X is the input tensor of the fused feature; K is the convolution kernel; * represents the multiplication operation, which refers to the point-by-point multiplication of the convolution kernel and the elements at the corresponding position of the input tensor; m and n are the indices of the convolution kernel, and the convolution kernel will traverse the entire input tensor and perform convolution calculations; Y(i, j) is the value of the output tensor at position (i, j) after the convolution operation;

[0082] The LSTM recurrent layer takes as input the sequence data output by the convolutional layer and outputs a hidden state vector that contains summary information of the time series data. It includes an input gate, a forget gate, an output gate, and a cell state.

[0083] The fully connected layer, whose input is the last hidden state of the LSTM recurrent layer, is expressed as: [l] =W [l] a [l-1] +b [l] ;z [l] is the linear combination output of the lth layer; W [l] is the weight matrix of the lth layer; a [l-1] is the activation output of the l-1 layer, which serves as the input of the l layer; b [l]is the bias vector of the lth layer; the activation function of the fully connected layer is: a [l] =g(z [l] );a [l] is the activation output of the lth layer, which outputs the predicted potential deviation of the current transmission power of the multi-element phased array system, that is, z after the activation function g [l] ; g is the ReLU, sigmoid or tanh activation function.

[0084] In this embodiment, a multi-stage deep learning model is constructed to predict the potential deviation of the transmit power of a multi-element phased array system, thereby improving the accuracy and adaptability of power control. Specifically, the model combines convolutional layers, LSTM recurrent layers, and fully connected layers. First, the convolutional layer extracts key patterns in the spatial or frequency domain from the fused features through a set of learnable filters, and uses the ReLU activation function to enhance the model's nonlinear expression capabilities and effectively capture the complex structure of the features. Next, the LSTM recurrent layer receives the output of the convolutional layer and uses its powerful time series processing capabilities to learn and retain long-term dependencies in the sequence data, generating a hidden state vector containing time series information, providing rich context for subsequent predictions. Finally, the fully connected layer uses the hidden state of the LSTM recurrent layer as input and maps the learned abstract features to the predicted value of the potential deviation of the transmit power through linear combination and nonlinear activation functions, thereby achieving accurate prediction of the transmit power deviation. Therefore, the embodiments of the present invention, through the multi-layer structure of the deep learning model, can not only automatically learn and extract multi-level information from the transmission signal characteristics, but also effectively handle timing dependencies. This allows for high-precision prediction of potential deviations in the transmission power of a multi-element phased array system under dynamic and complex operating conditions, thereby improving the real-time performance and accuracy of power correction.

[0085] It is understood that the specific function structure of the LSTM recurrent layer can refer to the existing technology. The training method of the transmit power prediction model can refer to the existing model training technology.

[0086] See also Figure 2 , is a schematic diagram of the structure of a transmission power calibration device for a multi-element phased array system provided by one embodiment of the present invention. The transmission power calibration device for a multi-element phased array system includes:

[0087] An acquisition module 10 is configured to acquire transmission signal parameters of a multi-element phased array system currently detected by the sensor system; the transmission signal parameters include transmission power, power spectrum density, energy concentration ratio, and phase difference;

[0088] The feature extraction module 11 is used to extract signal features from the transmission signal parameters to obtain transmission signal features; the transmission signal features include transmission power features, power spectrum density features, energy concentration ratio features and phase difference features;

[0089] A feature fusion module 12 is used to perform feature fusion on the emission signal features to obtain fused features;

[0090] A power deviation prediction module 13 is configured to input the fusion feature into a pre-trained transmit power prediction model to predict the current potential deviation of the transmit power of the multi-element phased array system;

[0091] The power correction module 14 is configured to calculate the corrected transmit power of the multi-element phased array system according to the transmit power and the potential transmit power deviation.

[0092] As an improvement to the above solution, the power spectrum density is: perform fast Fourier transform (FFT) on the signal to obtain the spectrum, and then calculate the square of its amplitude: S(f) is the power spectral density of signal s(t) at frequency f; is the Fourier transform of the signal s(t), which represents the representation of the signal s(t) in the frequency domain; It is the square of the modulus of the Fourier transform result, which represents the energy or power of the signal at frequency f;

[0093] The energy concentration ratio is: ECR is the energy concentration ratio, which reflects the degree to which the signal energy is concentrated within the main frequency bandwidth; l and f h are the lower and upper frequency limits of the main frequency bandwidth respectively; is the energy or power of the signal within the main frequency bandwidth; is the total energy or power of the signal s(t) in the entire frequency range;

[0094] The phase difference is: Δφ ij =∠s i -∠s j ;Δφ ij is the phase difference between the i-th array element and the j-th array element; ∠s i and ∠s j are the phase angles of the signals of the i-th array element and the j-th array element respectively.

[0095] As an improvement to the above solution, the feature fusion module is specifically used to:

[0096] Let x p is the transmit power eigenvector, x psd is the power spectrum density eigenvector, x ecris the energy concentration ratio eigenvector, x pd is the phase difference eigenvector;

[0097] Scaling the feature vectors of each of the emission signal features using a maximum and minimum scaling method to ensure that all features are at the same magnitude;

[0098] By calculating the contribution of the feature vectors of each of the scaling-processed emission signal features to the overall data variability, the fusion weight of each of the emission signal features is dynamically adjusted; the calculation formula is: I i is the importance score of the i-th feature in the transmitted signal features; the importance score is calculated by a feature importance scoring method;

[0099] performing high-order statistical processing on the eigenvectors of the scaled transmitted signal features to obtain statistics; the high-order statistical processing includes calculating the mean as a first-order moment, the variance as a second-order moment, the skewness as a third-order moment, and the kurtosis as a fourth-order moment of each eigenvector to capture the central tendency, dispersion, skewness, and peakedness of each transmitted signal feature;

[0100] All processed feature vectors and statistics are fused and calculated using the following feature fusion formula:

[0101] Represents the concatenation operation of vectors; x′ i is the eigenvector x i After scaling, the result x i is x p 、x psd 、x ecr 、x pd ; i represents the subscript of p, psd, ecr, pd; I i is the importance score of the i-th feature in the emission signal features; w i is the fusion weight of the i-th feature in the emission signal feature; μ i , γ 1,i ,γ 2,i are the mean, variance, skewness and kurtosis of the i-th feature in the emission signal features respectively; x fusion is the fused feature vector.

[0102] As an improvement to the above solution, the pre-trained transmit power prediction model includes:

[0103] The convolution layer extracts the fusion features of the input data through a set of learnable filter kernels. It is connected to the ReLU activation function to process the results of the convolution operation. The convolution operation is expressed as: Y(i,j) = ∑ m,n X(i+m,j+n)*K(m,n); X is the input tensor of the fused feature; K is the convolution kernel; * represents the multiplication operation, which refers to the point-by-point multiplication of the convolution kernel and the elements at the corresponding position of the input tensor; m and n are the indices of the convolution kernel, and the convolution kernel will traverse the entire input tensor and perform convolution calculations; Y(i,j) is the value of the output tensor at position (i,j) after the convolution operation;

[0104] The LSTM recurrent layer takes as input the sequence data output by the convolutional layer and outputs a hidden state vector that contains summary information of the time series data. It includes an input gate, a forget gate, an output gate, and a cell state.

[0105] The fully connected layer, whose input is the last hidden state of the LSTM recurrent layer, is expressed as: [l] =W [l] a [l-1] +b [l] ;z [l] is the linear combination output of the lth layer; W [l] is the weight matrix of the lth layer; a [l-1] is the activation output of the l-1 layer, which serves as the input of the l layer; b [l] is the bias vector of the lth layer; the activation function of the fully connected layer is: a [l] =g(z [l] );a [l] is the activation output of the lth layer, which outputs the predicted potential deviation of the current transmission power of the multi-element phased array system, that is, z after the activation function g [l] ; g is the ReLU, sigmoid or tanh activation function.

[0106] See also Figure 3 , is a schematic diagram of a transmit power calibration device for a multi-element phased array system provided in one embodiment of the present invention. The transmit power calibration device for a multi-element phased array system in this embodiment includes: a processor 100, a memory 101, and a computer program stored in the memory 101 and executable on the processor 100, such as a transmit power calibration program for a multi-element phased array system. When the processor 100 executes the computer program, it implements the steps in each of the aforementioned embodiments of the transmit power calibration method for a multi-element phased array system. Alternatively, when the processor 100 executes the computer program, it implements the functions of each module / unit in each of the aforementioned apparatus embodiments.

[0107] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program in the transmit power calibration device of the multi-element phased array system.

[0108] The transmit power calibration device for the multi-element phased array system can be a control device for the multi-element phased array system, or a computing device such as a desktop computer, laptop, PDA, or cloud server that is communicatively connected to the multi-element phased array system. The transmit power calibration device for the multi-element phased array system may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the schematic diagram is merely an example of a transmit power calibration device for a multi-element phased array system and does not limit the transmit power calibration device for a multi-element phased array system. The transmit power calibration device may include more or fewer components than shown in the diagram, or a combination of certain components, or different components. For example, the transmit power calibration device for the multi-element phased array system may also include input / output devices, network access devices, buses, etc.

[0109] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor. The processor serves as the control center of the transmit power calibration device of the multi-element phased array system, and connects various parts of the transmit power calibration device of the multi-element phased array system using various interfaces and lines.

[0110] The memory can be used to store the computer programs and / or modules. The processor implements the various functions of the transmit power calibration device of the multi-element phased array system by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store an operating system and at least one application required for a function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created based on the use of the mobile phone (such as audio data, a phone book, etc.). In addition, the memory may include high-speed random access memory and non-volatile memory, such as a hard disk, internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0111] If the integrated modules / units of the transmit power calibration device for the multi-element phased array system are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present invention can implement all or part of the processes in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practices in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practices, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0112] It should be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art can understand and implement the present invention without inventive effort.

[0113] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for calibrating the transmit power of a multi-element phased array system, characterized in that: include: Obtaining the transmission signal parameters of the multi-element phased array system currently detected by the sensor system; the transmission signal parameters include transmission power, power spectrum density, energy concentration ratio and phase difference; Extracting signal features from the transmission signal parameters to obtain transmission signal features; the transmission signal features include transmission power features, power spectrum density features, energy concentration ratio features, and phase difference features; Performing feature fusion on the emission signal features to obtain fusion features; Inputting the fusion features into a pre-trained transmit power prediction model to predict the current potential deviation of the transmit power of the multi-element phased array system; Calculating a corrected transmit power of a multi-element phased array system according to the transmit power and the potential transmit power deviation; The power spectral density is: perform a fast Fourier transform (FFT) on the signal to obtain the spectrum, and then calculate the square of its amplitude: S(f) is the power spectral density of signal s(t) at frequency f; is the Fourier transform of the signal s(t), which represents the representation of the signal s(t) in the frequency domain; It is the square of the modulus of the Fourier transform result, which represents the energy or power of the signal at frequency f; The energy concentration ratio is: ECR is the energy concentration ratio, which reflects the degree to which the signal energy is concentrated within the main frequency bandwidth; l and f h are the lower and upper frequency limits of the main frequency bandwidth respectively; is the energy or power of the signal within the main frequency bandwidth; is the total energy or power of the signal s(t) in the entire frequency range; The phase difference is: Δφ ij =∠s i -∠s j ;Δφ ij is the phase difference between the i-th array element and the j-th array element; ∠s i and ∠s j are the phase angles of the signals of the i-th array element and the j-th array element respectively; The performing feature fusion on the emission signal features to obtain fused features includes: Let x p is the transmit power eigenvector, x psd is the power spectrum density eigenvector, x ecr is the energy concentration ratio eigenvector, x pd is the phase difference eigenvector; Scaling the feature vectors of each of the emission signal features using a maximum and minimum scaling method to ensure that all features are at the same magnitude; By calculating the contribution of the feature vectors of each of the scaling-processed emission signal features to the overall data variability, the fusion weight of each of the emission signal features is dynamically adjusted; the calculation formula is: I i is the importance score of the i-th feature in the transmitted signal features; the importance score is calculated by a feature importance scoring method; performing high-order statistical processing on the eigenvectors of the scaled transmitted signal features to obtain statistics; the high-order statistical processing includes calculating the mean as a first-order moment, the variance as a second-order moment, the skewness as a third-order moment, and the kurtosis as a fourth-order moment of each eigenvector to capture the central tendency, dispersion, skewness, and peakedness of each transmitted signal feature; All processed feature vectors and statistics are fused and calculated using the following feature fusion formula: Represents the concatenation operation of vectors; x′ i is the eigenvector x i After scaling, the result x i is x p 、x psd 、x ecr 、x pd ; i represents the subscript of p, psd, ecr, pd; I i is the importance score of the i-th feature in the emission signal features; w i is the fusion weight of the i-th feature in the emission signal features; are the mean, variance, skewness and kurtosis of the i-th feature in the emission signal features respectively; x fusion is the fused feature vector.

2. The method for calibrating transmit power of a multi-element phased array system according to claim 1, wherein: The pre-trained transmit power prediction model includes: The convolution layer extracts the fusion features of the input data through a set of learnable filter kernels. It is connected to the ReLU activation function to process the results of the convolution operation. The convolution operation is expressed as: Y(i,j) = ∑ m,n X(i+m,j+n)*K(m,n); X is the input tensor of the fused feature; K is the convolution kernel; * represents the multiplication operation, which refers to the point-by-point multiplication of the convolution kernel and the elements at the corresponding position of the input tensor; m and n are the indices of the convolution kernel, and the convolution kernel will traverse the entire input tensor and perform convolution calculations; Y(i,j) is the value of the output tensor at position (i,j) after the convolution operation; The LSTM recurrent layer takes as input the sequence data output by the convolutional layer and outputs a hidden state vector that contains summary information of the time series data. It includes an input gate, a forget gate, an output gate, and a cell state. The fully connected layer, whose input is the last hidden state of the LSTM recurrent layer, is expressed as: [l] =W [l] a [l-1] +b [l] ;z [l] is the linear combination output of the lth layer; W [l] is the weight matrix of the lth layer; a [l-1] is the activation output of the l-1 layer, which serves as the input of the l layer; b [l] is the bias vector of the lth layer; the activation function of the fully connected layer is: a [l] =g(z [l] );a [l] is the activation output of the lth layer, which outputs the predicted potential deviation of the current transmission power of the multi-element phased array system, that is, z after the activation function g [l] ; g is the ReLU, sigmoid or tanh activation function.

3. A transmission power calibration device for a multi-element phased array system, characterized in that: include: An acquisition module is used to obtain the transmission signal parameters of the multi-element phased array system currently detected by the sensor system; the transmission signal parameters include transmission power, power spectrum density, energy concentration ratio and phase difference; A feature extraction module is used to extract signal features from the transmission signal parameters to obtain transmission signal features; the transmission signal features include transmission power features, power spectrum density features, energy concentration ratio features and phase difference features; A feature fusion module, configured to perform feature fusion on the emission signal features to obtain fused features; A power deviation prediction module is used to input the fusion feature into a pre-trained transmit power prediction model to predict the current potential deviation of the transmit power of the multi-element phased array system; a power correction module, configured to calculate a corrected transmit power of a multi-element phased array system based on the transmit power and the potential transmit power deviation; The power spectral density is: perform a fast Fourier transform (FFT) on the signal to obtain the spectrum, and then calculate the square of its amplitude: S(f) is the power spectral density of signal s(t) at frequency f; is the Fourier transform of the signal s(t), which represents the representation of the signal s(t) in the frequency domain; It is the square of the modulus of the Fourier transform result, which represents the energy or power of the signal at frequency f; The energy concentration ratio is: ECR is the energy concentration ratio, which reflects the degree to which the signal energy is concentrated within the main frequency bandwidth; l and f h are the lower and upper frequency limits of the main frequency bandwidth respectively; is the energy or power of the signal within the main frequency bandwidth; is the total energy or power of the signal s(t) in the entire frequency range; The phase difference is: Δφ ij =∠s i -∠s j ;Δφ ij is the phase difference between the i-th array element and the j-th array element; ∠s i and ∠s j are the phase angles of the signals of the i-th array element and the j-th array element respectively; The feature fusion module is specifically used for: Let x p is the transmit power eigenvector, x psd is the power spectrum density eigenvector, x ecr is the energy concentration ratio eigenvector, x pd is the phase difference eigenvector; Scaling the feature vectors of each of the emission signal features using a maximum and minimum scaling method to ensure that all features are at the same magnitude; Dynamically adjusting the fusion weight of each of the emission signal features by calculating the contribution of the feature vector of each of the emission signal features subjected to scaling processing to the overall data variability; The calculation formula is: I i is the importance score of the i-th feature in the transmitted signal features; the importance score is calculated by a feature importance scoring method; performing high-order statistical processing on the eigenvectors of the scaled transmitted signal features to obtain statistics; the high-order statistical processing includes calculating the mean as a first-order moment, the variance as a second-order moment, the skewness as a third-order moment, and the kurtosis as a fourth-order moment of each eigenvector to capture the central tendency, dispersion, skewness, and peakedness of each transmitted signal feature; All processed feature vectors and statistics are fused and calculated using the following feature fusion formula: Represents the concatenation operation of vectors; x′ i is the eigenvector x i After scaling, the result x i is x p 、x psd 、x ecr 、x pd ; i represents the subscript of p, psd, ecr, pd; I i is the importance score of the i-th feature in the emission signal features; w i is the fusion weight of the i-th feature in the emission signal features; are the mean, variance, skewness and kurtosis of the i-th feature in the emission signal features respectively; x fusion is the fused feature vector.

4. The transmission power calibration device for a multi-element phased array system according to claim 3, wherein: The pre-trained transmit power prediction model includes: The convolution layer extracts the fusion features of the input data through a set of learnable filter kernels. It is connected to the ReLU activation function to process the results of the convolution operation. The convolution operation is expressed as: Y(i,j) = ∑ m,n X(i+m,j+n)*K(m,n); X is the input tensor of the fused feature; K is the convolution kernel; * represents the multiplication operation, which refers to the point-by-point multiplication of the convolution kernel and the elements at the corresponding position of the input tensor; m and n are the indices of the convolution kernel, and the convolution kernel will traverse the entire input tensor and perform convolution calculations; Y(i,j) is the value of the output tensor at position (i,j) after the convolution operation; The LSTM recurrent layer takes as input the sequence data output by the convolutional layer and outputs a hidden state vector that contains summary information of the time series data. It includes an input gate, a forget gate, an output gate, and a cell state. The fully connected layer, whose input is the last hidden state of the LSTM recurrent layer, is expressed as: [l] =W [l] a [l-1] +b [l] ;z [l] is the linear combination output of the lth layer; W [l] is the weight matrix of the lth layer; a [l-1] is the activation output of the l-1 layer, which serves as the input of the l layer; b [l] is the bias vector of the lth layer; the activation function of the fully connected layer is: a [l] =g(z [l] );a [l] is the activation output of the lth layer, which outputs the predicted potential deviation of the current transmission power of the multi-element phased array system, that is, z after the activation function g [l] ; g is the ReLU, sigmoid or tanh activation function.

5. A transmission power calibration device for a multi-element phased array system, characterized in that: The system comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for calibrating the transmit power of a multi-element phased array system according to any one of claims 1 to 2 is implemented.

Images