Power amplifier nonlinear compensation method based on multi-layer memory depth BiLSTM

Abstract

The invention discloses a power amplifier nonlinear compensation method based on a multilayer memory depth BiLSTM, and the method comprises the steps: firstly obtaining the original data of a radar transmitting end and the output data of a radar power amplifier, constructing a multilayer memory depth BiLSTM neural network model, taking the original data of the radar transmitting end and the output data of the radar power amplifier as a data set, training the BiLSTM neural network model, and carrying out the nonlinear compensation of the radar power amplifier. And obtaining a trained multilayer memory depth BiLSTM neural network model, collecting data after radar power amplifier non-linear and Gaussian channel interference at a radar receiving end, and sending the data to the trained multilayer memory depth BiLSTM neural network model to complete radar power amplifier non-linear receiving end compensation. According to the scheme, the communication error rate can be effectively reduced, and compared with a traditional algorithm, the method has more excellent compensation performance under the conditions of low signal-to-noise ratio and strong radar power amplifier nonlinearity, and has better applicability.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of radar power amplifier linearization, and particularly relates to a power amplifier nonlinear compensation method based on a multi-layer memory deep BiLSTM. BACKGROUND

[0002] With the rapid development of information technology, the scarcity of spectrum resources is increasingly prominent, and the functional requirements of radar and communication systems are growing synchronously. At present, the working frequency bands of radar and communication systems begin to overlap, which promotes the rise of radar communication integration technology. This technology realizes the dual functions of target detection and communication transmission through sharing hardware devices. However, the power amplifier of the radar system is usually in a deep nonlinear working state in pursuit of high transmission power, and this operating mode will cause significant signal distortion, thereby causing demodulation failure or abnormal increase of the error rate at the receiving end. Specifically, the nonlinear characteristics of the power amplifier have become the main bottleneck restricting the high-speed communication performance of the integrated system. By implementing the amplifier linearization technology, the distortion degree of the signal can be effectively suppressed, the signal-to-noise ratio can be improved, and thus the error performance, transmission distance and overall reliability of the communication system can be improved.

[0003] Traditional nonlinear compensation technologies mainly include a predistortion scheme and a power amplifier nonlinear cancellation technology. The predistortion scheme needs to additionally configure a predistorter to suppress the nonlinear distortion of the power amplifier, but the hardware complexity and energy consumption are high, which is difficult to meet the application requirements of the radar system. The receiving-end nonlinear cancellation technology proposed by Tellado et al. decomposes and compensates the nonlinear components introduced by the power amplifier, but this method is harsh in the fitting accuracy of the power amplifier model parameters, and is only suitable for simple power amplifier structures without memory characteristics. SUMMARY

[0004] The power amplifier of the radar system is usually in a deep nonlinear working state in pursuit of high transmission power, and this operating mode will cause significant signal distortion, thereby causing demodulation failure or abnormal increase of the error rate at the receiving end. To solve the problem of high error rate of the distorted data of the memory-type power amplifier in the prior art, the purpose of the present application is to provide a power amplifier nonlinear compensation method based on a multi-layer memory deep BiLSTM. For the nonlinear problem of the power amplifier with memory effect, the long-term time sequence dependence characteristics in the trajectory data are captured in both directions, and the processing capacity for the memory-type nonlinear distortion is significantly improved. Experiments show that this method successfully solves the problem of serious signal distortion caused by the memory-type power amplifier, the signal-to-noise ratio of the receiving end is improved by 1.2 dB, and has the potential to be applied to the field of satellite-borne power amplifier nonlinear compensation.

[0005] The specific technical scheme for achieving the purpose of the present application is as follows:

[0006] A power amplifier nonlinear compensation method based on a multi-layer memory deep BiLSTM, comprising the following steps:

[0007] Step 1: Obtain the raw data from the radar transmitter and the output data from the radar power amplifier;

[0008] Step 2: Construct a multi-layer memory deep BiLSTM neural network model;

[0009] Step 3: Use the raw data from the radar transmitter and the output data from the radar power amplifier as a dataset to train the BiLSTM neural network model, and obtain the trained multi-layer memory deep BiLSTM neural network model.

[0010] Step 4: Collect data at the radar receiver after it has been subjected to radar power amplifier nonlinearity and Gaussian channel interference, and feed it into the trained multi-layer memory deep BiLSTM neural network model to complete the receiver compensation of radar power amplifier nonlinearity.

[0011] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0012] This invention addresses the nonlinearity problem of power amplifiers with memory effects. It employs a BiLSTM neural network model to compensate for the nonlinearity of radar power amplifiers at the receiver. By bidirectionally capturing long-term temporal dependence features in trajectory data, it significantly improves the ability to handle memory-type nonlinear distortion. This effectively suppresses the nonlinear distortion of radar power amplifiers, improves the signal-to-noise ratio, and thus enhances the error rate performance, transmission distance, and overall reliability of the communication system, thereby improving radar communication performance. Experiments show that this method successfully solves the problem of severe signal distortion caused by memory-type power amplifiers, improving the demodulation signal-to-noise ratio at the receiver by up to 1.2 dB, and has the potential for widespread application in the field of nonlinear compensation for spaceborne power amplifiers.

[0013] The present invention will be further described below with reference to specific embodiments. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating an embodiment of the present invention.

[0015] Figure 2 This is a schematic diagram of the experimental data acquisition and multi-layer memory deep BiLSTM training process in an embodiment of the present invention.

[0016] Figure 3 This is a schematic diagram of the internal structure of a single BiLSTM unit provided in an embodiment of the present invention;

[0017] Figure 4 A schematic diagram of a multi-layer memory depth BiLSTM model provided in an embodiment of the present invention. Detailed Implementation

[0018] Example

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not specifically refer to the singular and may also include the plural. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0021] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it need not be further discussed in subsequent drawings.

[0022] Combination Figure 1 A power amplifier nonlinearity compensation method based on multi-layer memory depth BiLSTM includes the following steps:

[0023] Step 1: Obtain the raw data from the radar transmitter and the output data from the radar power amplifier;

[0024] Combination Figure 2 The radar data was acquired using the experimental platform of KEYSIGHT. The raw data from the radar transmitter was passed through a signal generator (vector signal source M9384B VXG), then up-converted and input to the radar power amplifier. The output data of the radar power amplifier was passed through an attenuator and a coupler, then down-converted and then passed through a signal analyzer (vector signal analyzer N9042B UXA) to form the output data of the radar power amplifier.

[0025] In this embodiment, the input signal is a baseband OFDM signal;

[0026] Step 2: Construct a multi-layer memory deep BiLSTM neural network model;

[0027] The multi-layer memory deep BiLSTM model sequentially includes an input layer, a first BiLSTM layer, a BN layer, a tanh activation layer, a second BiLSTM layer, a BN layer, a tanh activation layer, a fully connected layer, and an output layer.

[0028] Wherein, the first BiLSTM includes The second BiLSTM layer includes one BiLSTM unit. BiLSTM units

[0029] The BiLSTM unit is a neural network formed by combining a forward LSTM and a backward LSTM:

[0030]

[0031]

[0032] In the formula, and These represent the forward and backward LSTM networks, respectively. Represents the predicted response. Represents the input to the LSTM network;

[0033] The LSTM network incorporates forget gates and remember gates to partially achieve the task of selecting model input information. These gate structures can emphasize or ignore the current cell state information, allowing for the retention or removal of network input information based on its importance. The sigmoid function is used as the activation function for the three gates; the forget gate uses historical data for decision-making. For the current moment The degree of influence, and determine the state quantity of historical data. Weight allocation, input gate through The output gate determines which state variables will be updated based on the output variable. And combined with the updated state variables Obtain the final output result :

[0034] Forgotten Gate:

[0035]

[0036] in, Represents the sigmoid function; and These represent the output of the previous time step and the input of the current time step, respectively. and These represent the parameter matrix and bias matrix of the forget gate, respectively; Indicates the output of the forget gate;

[0037] Input Gate:

[0038]

[0039] and These represent the parameter matrix and bias matrix of the input gate, respectively; Indicates the output of the input gate;

[0040] Output gate:

[0041]

[0042] and These represent the parameter matrix and bias matrix of the output gate, respectively; Indicates the output of the output gate;

[0043] Status Update:

[0044]

[0045]

[0046]

[0047] For parameter matrices; This is the bias matrix; express function; This indicates the output at the current moment; , as well as These represent the temporary cell state, the cell state at the previous time step, and the cell state at the current time step, respectively.

[0048] Step 3: Use the raw data from the radar transmitter and the output data from the radar power amplifier as a dataset to train the BiLSTM neural network model, and obtain the trained multi-layer memory deep BiLSTM neural network model.

[0049] The radar power amplifier output data is used as the input data of the BiLSTM neural network model, and the original data is used as the output data of the BiLSTM neural network model to train the BiLSTM neural network model.

[0050] Mean squared error is used as the loss function; the Adam optimizer based on cross-entropy loss function is employed.

[0051] Delay processing of training data Each unit is used to generate the memory depth. The data, after delay processing, is fed into the network structure, and the memory depth is... The input data matrix has a dimension of The output matrix has a dimension of Input data and output data The format is:

[0052]

[0053]

[0054] in, The input signal is represented by the first... The real part of each sampling point The input signal is represented by the first... The imaginary part of each sampling point The output signal is represented by the first... The real part of each sampling point The output signal is represented by the first... The imaginary part of each sampling point;

[0055] In the training of the BiLSTM neural network model, the following methods are adopted: The dataset is used for neural network training. Group data is used for cross-validation, where .

[0056] Based on the data characteristics, a multi-dimensional to two-dimensional input / output mode is adopted; the minimum batch size is... The maximum number of iterations is The learning rate is .

[0057] Step 4: Collect data at the radar receiver after passing through the nonlinearity of the radar power amplifier and Gaussian channel interference, and feed it into the trained multi-layer memory deep BiLSTM neural network model to complete the receiver compensation of the nonlinearity of the radar power amplifier.

[0058] Data is collected at the radar receiver after being subjected to radar power amplifier nonlinearity and Gaussian channel interference, and then fed into a trained multi-layer memory deep BiLSTM neural network model. The output of the multi-layer memory deep BiLSTM neural network model is the radar received data after radar power amplifier nonlinearity compensation.

[0059] The compensation algorithm described in this invention can effectively reduce the communication bit error rate. Compared with traditional algorithms, it has better compensation performance under low signal-to-noise ratio and strong radar power amplifier nonlinearity, and has good applicability.

[0060] This invention also provides a power amplifier nonlinearity compensation system based on multilayer memory depth BiLSTM, comprising the following steps:

[0061] Data acquisition module: used to acquire raw data from the radar transmitter and output data from the radar power amplifier;

[0062] BiLSTM Neural Network Model Building Module: Used to build a multi-layer memory deep BiLSTM neural network model, and use the original data from the radar transmitter and the output data from the radar power amplifier as the dataset to train the BiLSTM neural network model, and obtain the trained multi-layer memory deep BiLSTM neural network model.

[0063] Radar power amplifier nonlinearity compensation module: This module is used to collect data at the radar receiver after it has been subjected to radar power amplifier nonlinearity and Gaussian channel interference, and then feed it into a trained multi-layer memory deep BiLSTM neural network model to complete the receiver-side compensation for radar power amplifier nonlinearity.

[0064] This solution also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0065] Step 1: Obtain the raw data from the radar transmitter and the output data from the radar power amplifier;

[0066] Step 2: Construct a multi-layer memory deep BiLSTM neural network model;

[0067] Step 3: Use the raw data from the radar transmitter and the output data from the radar power amplifier as a dataset to train the BiLSTM neural network model, and obtain the trained multi-layer memory deep BiLSTM neural network model.

[0068] Step 4: Collect data at the radar receiver after it has been subjected to radar power amplifier nonlinearity and Gaussian channel interference, and feed it into the trained multi-layer memory deep BiLSTM neural network model to complete the receiver compensation of radar power amplifier nonlinearity.

[0069] This solution also provides a computer-readable storage medium on which a computer program is stored, wherein the computer program, when executed by a processor, performs the following steps:

[0070] Step 1: Obtain the raw data from the radar transmitter and the output data from the radar power amplifier;

[0071] Step 2: Construct a multi-layer memory deep BiLSTM neural network model;

[0072] Step 3: Use the raw data from the radar transmitter and the output data from the radar power amplifier as a dataset to train the BiLSTM neural network model, and obtain the trained multi-layer memory deep BiLSTM neural network model.

[0073] Step 4: Collect data at the radar receiver after it has been subjected to radar power amplifier nonlinearity and Gaussian channel interference, and feed it into the trained multi-layer memory deep BiLSTM neural network model to complete the receiver compensation of radar power amplifier nonlinearity.

[0074] The embodiments described above are merely one implementation method of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A power amplifier nonlinearity compensation method based on multi-layer memory depth BiLSTM, characterized in that, Includes the following steps: Step 1: Obtain the raw data from the radar transmitter and the output data from the radar power amplifier; Step 2: Construct a multi-layer memory deep BiLSTM neural network model; Step 3: Use the raw data from the radar transmitter and the output data from the radar power amplifier as a dataset to train the BiLSTM neural network model, and obtain the trained multi-layer memory deep BiLSTM neural network model. Step 4: Collect data at the radar receiver after it has been subjected to radar power amplifier nonlinearity and Gaussian channel interference, and feed it into the trained multi-layer memory deep BiLSTM neural network model to complete the receiver compensation of radar power amplifier nonlinearity.

2. The power amplifier nonlinearity compensation method based on multilayer memory depth BiLSTM according to claim 1, characterized in that, The radar power amplifier data is as follows: The raw data from the radar transmitter is generated by a signal generator, then up-converted and input to the radar power amplifier. The output data from the radar power amplifier is then down-converted after passing through an attenuator and a coupler, and finally analyzed by a signal analyzer to form the output data of the radar power amplifier.

3. The power amplifier nonlinearity compensation method based on multilayer memory depth BiLSTM according to claim 1, characterized in that, The multi-layer memory deep BiLSTM model sequentially includes an input layer, a first BiLSTM layer, a BN layer, a tanh activation layer, a second BiLSTM layer, a BN layer, a tanh activation layer, a fully connected layer, and an output layer. Wherein, the first BiLSTM includes The second BiLSTM layer includes one BiLSTM unit. One BiLSTM unit.

4. The power amplifier nonlinearity compensation method based on multilayer memory depth BiLSTM according to claim 3, characterized in that, The BiLSTM unit is a neural network formed by combining a forward LSTM and a backward LSTM: ； ； In the formula, and These represent the forward and backward LSTM networks, respectively. Represents the predicted response. Represents the input to the LSTM network; The LSTM network includes forget gates and remember gates to retain or remove network input information based on its importance. Forgotten Gate: ； in, Represents the sigmoid function; and These represent the output of the previous time step and the input of the current time step, respectively. and These represent the parameter matrix and bias matrix of the forget gate, respectively; Indicates the output of the forget gate; Input Gate: ； and These represent the parameter matrix and bias matrix of the input gate, respectively; Indicates the output of the input gate; Output gate: ； and These represent the parameter matrix and bias matrix of the output gate, respectively; Indicates the output of the output gate; Status Update: ； ； ； For parameter matrices; This is the bias matrix; express function; This indicates the output at the current moment; , as well as These represent the temporary cell state, the cell state at the previous time step, and the cell state at the current time step, respectively.

5. The power amplifier nonlinearity compensation method based on multilayer memory depth BiLSTM according to claim 3, characterized in that, The training of the BiLSTM neural network model in step 3 specifically involves: The radar power amplifier output data is used as the input data of the BiLSTM neural network model, and the original data is used as the output data of the BiLSTM neural network model to train the BiLSTM neural network model. Mean squared error is used as the loss function; the Adam optimizer based on cross-entropy loss function is employed. Delay processing of training data Each unit is used to generate the memory depth. The data, after delay processing, is fed into the network structure, and the memory depth is... The input data matrix has a dimension of The output matrix has a dimension of Input data and output data The format is: ； ； in, The input signal is represented by the first... The real part of each sampling point The input signal is represented by the first... The imaginary part of each sampling point The output signal is represented by the first... The real part of each sampling point The output signal is represented by the first... The imaginary part of each sampling point.

6. The power amplifier nonlinearity compensation method based on multilayer memory depth BiLSTM according to claim 5, characterized in that, In the training of the BiLSTM neural network model, the following methods are adopted: The dataset is used for neural network training. Group data is used for cross-validation, where .

7. The power amplifier nonlinearity compensation method based on multilayer memory depth BiLSTM according to claim 1, characterized in that, The receiver compensation for the nonlinearity of the radar power amplifier in step 4 is specifically as follows: Data is collected at the radar receiver after being subjected to radar power amplifier nonlinearity and Gaussian channel interference, and then fed into a trained multi-layer memory deep BiLSTM neural network model. The output of the multi-layer memory deep BiLSTM neural network model is the radar received data after radar power amplifier nonlinearity compensation.

8. A power amplifier nonlinearity compensation system based on multi-layer memory depth BiLSTM, characterized in that, Includes the following steps: Data acquisition module: used to acquire raw data from the radar transmitter and output data from the radar power amplifier; BiLSTM Neural Network Model Building Module: Used to build a multi-layer memory deep BiLSTM neural network model, and use the original data from the radar transmitter and the output data from the radar power amplifier as the dataset to train the BiLSTM neural network model, and obtain the trained multi-layer memory deep BiLSTM neural network model. Radar power amplifier nonlinearity compensation module: This module is used to collect data at the radar receiver after it has been subjected to radar power amplifier nonlinearity and Gaussian channel interference, and then feed it into a trained multi-layer memory deep BiLSTM neural network model to complete the receiver-side compensation for radar power amplifier nonlinearity.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-7.

10. A computer-storable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

Images