Satellite-to-ground channel detection bandwidth extension method of auto-encoder

Abstract

The invention relates to the technical field of signal detection, in particular to a satellite-ground channel detection bandwidth extension method of an auto-encoder. The method comprises the following steps: preprocessing a time domain data signal to obtain power spectrum density data; power spectral density characteristics of the power spectral density data are extracted by using a preset frequency spectrum prediction network model, and the preset frequency spectrum prediction network model is obtained by training a full-connection neural network through broadband data; and based on the power spectrum density characteristics, predicting and complementing the missing part in the residual spectrum signal by using a preset spectrum prediction network model, and recovering the original spectrum shape. Therefore, the spectrum of broadband data is learned in offline training, and after spectrum features are extracted, residual spectrum detection and spectrum prediction complementation are realized by using the full-connection neural network, so that the problem of inaccurate spectrum prediction result caused by the fact that the prior art is limited by ADC receiving sampling rate, broadband resolution and other factors is solved; and the spectrum prediction precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of signal detection, and in particular to a method for extending bandwidth of satellite-to-ground channel detection of an autoencoder. Background Art

[0002] Driven by the global demand for information technology applications, the demand for data transmission rates in satellite communication hotspots such as satellite navigation, space-based Internet, global Internet of Things and 6G communication is increasing. However, achieving higher data transmission rates usually requires a corresponding increase in signal bandwidth. In order to effectively sample the output signal of the power amplifier (PA), a high-speed analog-to-digital converter (ADC) is usually required.

[0003] The ADC sampling rate of traditional signal transmitters needs to be several times the original signal bandwidth. The maximum receiving sampling rate of the ADC is limited by its hardware design. Once the signal bandwidth exceeds this range, the signal will suffer varying degrees of attenuation, which will affect the normal reception of the signal. Similarly, in the face of the need for broadband and high-resolution spectrum detection, existing spectrum sensing technologies can only be implemented within the maximum receiving sampling bandwidth of the ADC. For channels that exceed the ADC sampling bandwidth, it is often difficult to accurately obtain the signal status, or even impossible to sense. In the field of satellite communications, the ADC receiving bandwidth and power of onboard signal processing receivers are strictly limited, which poses a great challenge to the spectrum sensing of broadband signals in satellite communications and needs to be solved urgently. Summary of the invention

[0004] The present invention provides a satellite-to-ground channel detection bandwidth extension method for an autoencoder, so as to solve the problem that the existing technology is limited by factors such as ADC receiving sampling rate and broadband resolution, resulting in inaccurate spectrum prediction results, and improve the spectrum prediction accuracy.

[0005] To achieve the above object, a first aspect of the present invention provides a method for extending the bandwidth of a satellite-to-ground channel detection of an autoencoder, comprising the following steps: Acquire a time domain data signal, and preprocess the time domain data signal to obtain power spectrum density data; Extracting power spectrum density features of the power spectrum density data using a preset spectrum prediction network model, wherein the preset spectrum prediction network model is obtained by training a fully connected neural network using broadband data; Based on the power spectral density characteristics, the preset spectrum prediction network model is used to predict and complete the missing parts in the residual spectrum signal to restore the original spectrum shape.

[0006] According to one embodiment of the present invention, before extracting the power spectrum density features of the power spectrum density data using the preset spectrum prediction network model, the method further includes: Acquire the broadband data, and construct a training test sample data group using the broadband data, wherein the broadband data is sampled using an analog-to-digital converter ADC; Preprocessing the training and testing sample data group, and obtaining a training set and a validation set based on the preprocessed training and testing sample data group; Based on a preset loss function, the fully connected neural network is trained using the training set to obtain an initial neural network model, and the initial neural network model is verified using the verification set until the initial neural network model meets the preset criteria, and the iterative training of the fully connected neural network is terminated to obtain the preset spectrum prediction network model, otherwise the iterative training continues after adjusting the model parameters.

[0007] According to one embodiment of the present invention, preprocessing the training test sample data set and obtaining a training set and a validation set based on the preprocessed training test sample data set includes: The training test sample data group is divided into low-speed ADC sampling broadband data and high-speed ADC sampling broadband data at corresponding moments; Based on a preset frame length and a preset step length, the low-speed ADC sampling broadband data and the high-speed ADC sampling broadband data at the corresponding moment are respectively processed in segments, and the low-speed ADC sampling broadband data after the segment processing is used as training data, and the high-speed ADC sampling broadband data after the segment processing is used as verification data; Performing a fast Fourier transform process with a length of a first preset length on each row of data of each data group in the training data and the verification data to obtain a broadband spectrum; Taking a modulus value of the broadband spectrum to obtain a power spectrum density estimation result; The power spectrum density estimation result is divided into intervals of a second preset length to obtain the training set and the validation set.

[0008] According to one embodiment of the present invention, after obtaining the training set and the validation set based on the preprocessed training and testing sample data set, the method further includes: Normalizing the training set to obtain a normalized training set; Activating each training sample in the normalized training set to obtain a first feature vector of each training sample; Activating the channel dimensions of the first feature vector of each training sample respectively to obtain the second feature vector of each training sample; The first feature vector of each training sample and the second feature vector of the corresponding training sample are integrated to obtain a channel feature result of each training sample.

[0009] According to an embodiment of the present invention, after obtaining the channel characteristic result of each training sample, the method further includes: Activating the channel feature results of each training sample respectively to obtain a third feature vector of each training sample; Activating the channel dimension of the third eigenvector of each training sample respectively to obtain the fourth eigenvector of each training sample; Integrate and predict the third eigenvector of each training sample and the fourth eigenvector of the corresponding sample to obtain a predicted channel power spectral density of each training sample; Based on a preset splicing strategy, the predicted channel power spectrum density of each training sample is spliced ​​to obtain a complete power spectrum density result.

[0010] According to one embodiment of the present invention, the preset loss function is:

[0011] in, For the complete power spectral density result, is the validation set, is the preset frame length, is the number of rows of the power spectral density result.

[0012] According to an autoencoder satellite-to-ground channel detection bandwidth extension method proposed in an embodiment of the present invention, power spectrum density data can be obtained by acquiring a time domain data signal and preprocessing the time domain data signal; then, the power spectrum density features of the power spectrum density data can be extracted using a preset spectrum prediction network model; based on the power spectrum density features, the preset spectrum prediction network model is used to predict and complete the missing parts of the residual spectrum signal, thereby restoring the original spectrum shape. Thus, by learning the spectrum of broadband data in offline training, and after extracting the spectrum features, a fully connected neural network is used to realize residual spectrum detection and spectrum prediction completion, the problem of inaccurate spectrum prediction results caused by factors such as ADC reception sampling rate and broadband resolution in the prior art is solved, thereby improving the accuracy of spectrum prediction.

[0013] To achieve the above object, a second aspect of the present invention provides a satellite-to-ground channel detection bandwidth extension device for an autoencoder, comprising: An acquisition module, used for acquiring a time domain data signal and preprocessing the time domain data signal to obtain power spectrum density data; An extraction module, used to extract the power spectrum density features of the power spectrum density data by using a preset spectrum prediction network model, wherein the preset spectrum prediction network model is obtained by training a fully connected neural network with broadband data; The prediction module is used to predict and complete the missing parts in the residual spectrum signal based on the power spectral density characteristics and restore the original spectrum shape by using the preset spectrum prediction network model.

[0014] According to an embodiment of the present invention, before extracting the power spectrum density features of the power spectrum density data using the preset spectrum prediction network model, the extraction module includes: A construction unit, used to obtain the broadband data and construct a training test sample data group using the broadband data, wherein the broadband data is obtained by sampling using an analog-to-digital converter ADC; A preprocessing unit, used to preprocess the training test sample data group, and obtain a training set and a validation set based on the preprocessed training test sample data group; A training unit is used to train the fully connected neural network based on a preset loss function using the training set to obtain an initial neural network model, and to verify the initial neural network model using the verification set, until the initial neural network model meets the preset standard, and then terminate the iterative training of the fully connected neural network to obtain the preset spectrum prediction network model, otherwise, continue the iterative training after adjusting the model parameters.

[0015] According to one embodiment of the present invention, the preprocessing unit is specifically used for: The training test sample data group is divided into low-speed ADC sampling broadband data and high-speed ADC sampling broadband data at corresponding moments; Based on a preset frame length and a preset step length, the low-speed ADC sampling broadband data and the high-speed ADC sampling broadband data at the corresponding moment are respectively processed in segments, and the low-speed ADC sampling broadband data after the segment processing is used as training data, and the high-speed ADC sampling broadband data after the segment processing is used as verification data; Performing a fast Fourier transform process with a length of a first preset length on each row of data of each data group in the training data and the verification data to obtain a broadband spectrum; Taking a modulus value of the broadband spectrum to obtain a power spectrum density estimation result; The power spectrum density estimation result is divided into intervals of a second preset length to obtain the training set and the validation set.

[0016] According to one embodiment of the present invention, after obtaining the training set and the validation set based on the preprocessed training and testing sample data set, the preprocessing unit further includes: A processing subunit, used for performing normalization processing on the training set to obtain a normalized training set; A first activation subunit is used to activate each training sample in the normalized training set to obtain a first feature vector of each training sample; A second activation subunit is used to activate the channel dimension of the first feature vector of each training sample respectively to obtain the second feature vector of each training sample; The integration subunit is used to integrate the first feature vector of each training sample and the second feature vector of the corresponding training sample to obtain the channel feature result of each training sample.

[0017] According to an embodiment of the present invention, after obtaining the channel characteristic result of each training sample, the integration subunit is further used to: Activating the channel feature results of each training sample respectively to obtain a third feature vector of each training sample; Activating the channel dimension of the third eigenvector of each training sample respectively to obtain the fourth eigenvector of each training sample; Integrate and predict the third eigenvector of each training sample and the fourth eigenvector of the corresponding sample to obtain a predicted channel power spectral density of each training sample; Based on a preset splicing strategy, the predicted channel power spectrum density of each training sample is spliced ​​to obtain a complete power spectrum density result.

[0018] According to one embodiment of the present invention, the preset loss function is:

[0019] in, For the complete power spectral density result, is the validation set, is the preset frame length, is the number of rows of the power spectral density result.

[0020] According to an embodiment of the present invention, a bandwidth extension device for satellite-to-ground channel detection of an autoencoder is proposed. By acquiring a time domain data signal and preprocessing the time domain data signal, power spectrum density data can be obtained; then, the power spectrum density features of the power spectrum density data can be extracted using a preset spectrum prediction network model; based on the power spectrum density features, the preset spectrum prediction network model is used to predict and complete the missing parts of the residual spectrum signal, thereby restoring the original spectrum shape. Therefore, by learning the spectrum of broadband data in offline training, and after extracting the spectrum features, a fully connected neural network is used to realize residual spectrum detection and spectrum prediction completion, the problem of inaccurate spectrum prediction results caused by factors such as ADC receiving sampling rate and broadband resolution in the prior art is solved, thereby improving the accuracy of spectrum prediction.

[0021] To achieve the above-mentioned purpose, the third aspect of the present invention proposes an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a method for extending the satellite-to-ground channel detection bandwidth of an autoencoder as described in the above-mentioned embodiment.

[0022] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a method for extending the bandwidth of a satellite-to-earth channel detection of an autoencoder as described in the above embodiment.

[0023] To achieve the above objectives, a fifth aspect of the present invention provides a computer program product, which includes a computer program. When the program is executed by a processor, it is used to implement a satellite-to-earth channel detection bandwidth extension method for an autoencoder as described in the above embodiment.

[0024] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 A flowchart of a method for extending bandwidth of satellite-to-ground channel detection of an autoencoder according to an embodiment of the present invention; Figure 2 A schematic diagram of the working process of a method for extending bandwidth of satellite-to-ground channel detection of an autoencoder according to an embodiment of the present invention; Figure 3 A flowchart of a training process of a spectrum prediction network model according to an embodiment of the present invention; Figure 4 A schematic diagram of the effect of extending the spectrum detection bandwidth according to an embodiment of the present invention; Figure 5 A schematic diagram of the effect of extending the spectrum detection bandwidth according to another embodiment of the present invention; Figure 6 A block diagram of a bandwidth extension device for satellite-to-ground channel detection of an autoencoder according to an embodiment of the present invention; Figure 7 It is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0026] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0027] A method for extending the satellite-to-ground channel detection bandwidth of an autoencoder according to an embodiment of the present invention is described below with reference to the accompanying drawings.

[0028] Figure 1 The present invention is a flowchart of a method for extending bandwidth of a satellite-to-ground channel detection of an autoencoder according to an embodiment of the present invention.

[0029] For example, Figure 1 As shown, the satellite-to-ground channel detection bandwidth extension method of an autoencoder comprises the following steps: In step S101, a time domain data signal is acquired and preprocessed to obtain power spectrum density data.

[0030] Specifically, first, time domain data signals can be collected from various sensors or data acquisition devices. Since these signals usually contain various noises and interferences, further processing can be performed to ensure the quality and accuracy of the data. Next, the acquired time domain data signals are preprocessed, for example, the signals are filtered to remove high-frequency noise or low-frequency interference, thereby improving the quality of the signals; the signals are normalized to be within a specific range for subsequent analysis and processing; the signals are detrended to eliminate long-term trends in the signals to make them more stable, etc. After these preprocessing steps, more reliable time domain data signals can be obtained. The preprocessed signals can be sampled using ADC to obtain broadband data, and the broadband data is fast Fourier transformed to convert them from the time domain to the frequency domain. The frequency components of the signal can be obtained by fast Fourier transform, and the power spectral density data can be calculated. The power spectral density data reflects the energy distribution of the signal at different frequencies, which lays the foundation for further signal analysis and processing.

[0031] In step S102, a preset spectrum prediction network model is used to extract power spectrum density features of the power spectrum density data, wherein the preset spectrum prediction network model is obtained by training a fully connected neural network using broadband data.

[0032] Among them, power spectral density (PSD) is a function that describes how signal power distribution changes with frequency; fully connected neural network refers to a network structure in deep learning, in which each neuron is connected to all neurons in the previous layer and can process complex data patterns.

[0033] That is to say, by using broadband data to train a fully connected neural network, a preset spectrum prediction network model can be obtained. The preset spectrum prediction network model is essentially an artificial intelligence algorithm. Through the fully connected spectrum feature extraction module in the preset spectrum prediction network model, the power spectrum density features in the power spectrum density data can be effectively extracted. Here, we mainly focus on the spectrum information that is severely attenuated in the power spectrum density, thereby accurately locating the position and shape of the spectrum signal before attenuation.

[0034] In step S103, based on the power spectral density characteristics, a preset spectrum prediction network model is used to predict and complete the missing parts of the residual spectrum signal to restore the original spectrum shape.

[0035] It can be understood that after obtaining the power spectral density characteristics, the fully connected spectrum feature prediction extension module in the preset spectrum prediction network model is further called to infer the missing data based on the known spectrum information, thereby restoring a complete original spectrum shape and realizing the prediction generation and completion of the residual spectrum signal.

[0036] For ease of understanding, the training process of the preset spectrum prediction network model is described in detail below.

[0037] As a possible implementation method, in some embodiments, before using a preset spectrum prediction network model to extract the power spectrum density characteristics of the power spectrum density data, it also includes: obtaining broadband data, and using the broadband data to construct a training test sample data group, wherein the broadband data is sampled using an analog-to-digital converter ADC; preprocessing the training test sample data group, and obtaining a training set and a validation set based on the preprocessed training test sample data group; based on a preset loss function, using the training set to train a fully connected neural network to obtain an initial neural network model, and using the validation set to verify the initial neural network model until the initial neural network model meets the preset criteria, and then terminating the iterative training of the fully connected neural network to obtain the preset spectrum prediction network model, otherwise adjusting the model parameters and continuing the iterative training.

[0038] Specifically, in order to train the preset spectrum prediction network model, first, the broadband data sampled by the ADC can be used to construct a training test sample data group for input into the fully connected neural network. These sample data groups are the basis for subsequent analysis and model training. Next, these training test sample data groups are preprocessed, such as data cleaning, normalization, denoising and other operations to ensure the quality and consistency of the data. After the preprocessing is completed, an optimized training set and validation set can be obtained. Based on the training set and validation set obtained after these preprocessing, a preset loss function can be used to guide the training process of the model, that is, the fully connected neural network is used as the basic model and trained using the training set. During the training process, the model will continuously adjust its internal parameters and update the model to minimize the value of the loss function.

[0039] In order to verify the effectiveness and accuracy of the model, we will use the validation set to evaluate the initial neural network model obtained during the training process. If the performance of the model meets the preset standards, such as the loss function value is lower than a certain threshold or the accuracy is higher than a certain level, then it can be considered that the model has achieved the expected performance, thus ending the iterative training process. However, if the performance of the model fails to meet the preset standards, it is necessary to adjust the model parameters, such as the learning rate, number of layers, number of neurons, etc., and then continue iterative training until the performance of the model meets the preset standards, thereby obtaining a spectrum prediction network model with excellent performance.

[0040] In some embodiments, the preset loss function is:

[0041] in, For the complete power spectral density result, is the validation set, is the preset frame length, is the number of rows of the power spectral density result.

[0042] In the embodiment of the present invention, in order to ensure that the power spectrum results generated in step S102 and step S103 are correct and stable, the above loss function can be used to update the model during the model training process. The result obtained by inputting the training set into the model for training each time Compare and calculate the mean square error loss result , and use The data updates the model parameters. As it approaches 0, the model is updated more and more accurately. As the data set is learned, the model's ability to generate accurate power spectra is further improved. Among them, the input and output of the fully connected neural network are both length A one-dimensional vector of .

[0043] The following is a detailed description of how to obtain the training set and validation set.

[0044] As a possible implementation method, in some embodiments, the training test sample data group is preprocessed, and a training set and a validation set are obtained based on the preprocessed training test sample data group, including: dividing the training test sample data group into low-speed ADC sampling broadband data and high-speed ADC sampling broadband data at corresponding moments; based on a preset frame length and a preset step length, the low-speed ADC sampling broadband data and the high-speed ADC sampling broadband data at corresponding moments are segmented, and the segmented low-speed ADC sampling broadband data is used as training data, and the segmented high-speed ADC sampling broadband data is used as validation data; each row of data of each data group in the training data and the validation data is fast Fourier transform-processed with a length of a first preset length to obtain a broadband spectrum; the broadband spectrum is modulo-valued to obtain a power spectrum density estimation result; the power spectrum density estimation result is divided into intervals of a second preset length to obtain a training set and a validation set.

[0045] Specifically, the training test sample data group is first divided into two main categories, namely, broadband data sampled by low-speed ADC and broadband data sampled by high-speed ADC at the corresponding time; next, based on the preset frame length and step length Parameters, segmentation operations are performed on these two types of data respectively, where Need to The division starting point is set at the first sampling point of the broadband data sampled by the low-speed ADC, and the total frame length intercepted from the starting point is The data of length is taken as the first row of training data in the data set, and then the starting point of the division is moved to the back end of the data by the step length sampling points, with the The frame length is intercepted backward from the starting point. The data is used as the second row of training data in the dataset, and the training data in the subsequent datasets is Behavior The sampling points are taken as the starting point, and the frame length intercepted backward is until all the broadband data sampled by the low-speed ADC is divided, thus obtaining the training data The segmentation operation of the broadband data sampled by the high-speed ADC is the same as the segmentation operation of the broadband data sampled by the low-speed ADC, thereby obtaining the verification data .

[0046] In order to extract the signal spectrum features and make the data in the data set conform to the designed model size, after the segmentation processing is completed, the training data in each data group can be obtained. , Verify data For the data in a single data group, there are several rows with frame lengths of The data can be used for training data and validation data Each row of data in each data group in the data set is subjected to fast Fourier transform processing to obtain a Fourier transform result in complex form, i.e., a broadband spectrum; then, a modulus operation is performed on the broadband spectrum to convert the complex spectrum into a real power spectrum, thereby obtaining a power spectrum density estimation result; finally, the power spectrum density estimation result is transmitted to a second preset length Divide by interval, and the number of columns can be obtained as The matrix of the training set and validation set , and save it for subsequent model training and verification processes.

[0047] Through the above operations, not only the accuracy and consistency of data processing are ensured, but also a high-quality data set is provided for model training and verification.

[0048] Furthermore, in some embodiments, after obtaining the training set and the validation set based on the preprocessed training test sample data group, it also includes: normalizing the training set to obtain a normalized training set; activating each training sample in the normalized training set to obtain a first eigenvector of each training sample; activating the channel dimension of the first eigenvector of each training sample respectively to obtain a second eigenvector of each training sample; integrating the first eigenvector of each training sample and the second eigenvector of the corresponding training sample to obtain a channel feature result of each training sample.

[0049] Next, we build a fully connected spectrum feature extraction module and use it to extract the training set Processing is performed to extract features of severely attenuated signals in the spectrum, and further generate spectrum features of spectrum information.

[0050] Specifically, the broadband spectrum is divided into channels, and only one of them is processed at a time. Taking one of the channels as an example, first, the training data in the first training data group Perform normalization to obtain the corresponding normalized training data .in, is the number of rows of the power spectral density result of the input neural network, is the number of signal channels, Represents mathematical multiplication, function Then, the input training data is processed by using the module E1 in the fully connected spectrum feature extraction module, which includes a fully connected layer, a normalization layer (Batch Normalization, BN for short) and an activation layer. Activate and extract the spectral features in the channel to obtain the first eigenvector , and then use module E2 containing a fully connected layer, a normalization layer, and an activation layer to transform the first feature vector Activate the channel dimension to get the second eigenvector Finally, the signal of the channel dimension is integrated and generated through the feature enhancement generation module E3 containing the fully connected layer to obtain the enhanced channel feature result , Represents the extracted enhanced features of the current channel, which is used for subsequent recovery and generation of channel feature content reference.

[0051] in,

[0052]

[0053]

[0054] Among them, the detailed structure of module E1 is: the dimension is The detailed structure of module E2 is: a fully connected layer, a normalization layer, a RELU (Rectified Linear Unit) activation function layer, and a Dropout layer (a technology used to reduce overfitting of neural networks); the detailed structure of module E2 is: a fully connected layer, a normalization layer, a RELU activation function layer, and a Dropout layer with a dimension equal to the output dimension of the E1 module; the detailed structure of module E3 is: a fully connected layer with an input dimension equal to the output dimension of the E2 module and an output dimension of N.

[0055] Furthermore, in some embodiments, after obtaining the channel characteristic results of each training sample, it also includes: activating the channel characteristic results of each training sample respectively to obtain the third eigenvector of each training sample; activating the channel dimension of the third eigenvector of each training sample respectively to obtain the fourth eigenvector of each training sample; integrating and predicting the third eigenvector of each training sample and the fourth eigenvector of the corresponding sample to obtain the predicted channel power spectral density of each training sample; based on a preset splicing strategy, splicing the predicted channel power spectral density of each training sample to obtain a complete power spectral density result.

[0056] Next, we build a fully connected spectrum feature prediction extension module and use it to enhance the channel feature results. Processing is performed to analyze the severely attenuated signal characteristics in the spectrum, and the spectrum characteristics of the spectrum information are generated in a predictive manner.

[0057] Specifically, taking one of the channels as an example, the fully connected spectral feature prediction extension module includes a fully connected layer, a normalized layer, and an activation layer. The module D1 is used to enhance the feature data of the input channel. Activate and extract the enhanced channel spectrum features to obtain the third eigenvector , and then use the module D2 containing a fully connected layer, a normalized layer, and an activation layer to transform the third feature vector Activate the channel dimension to get the fourth eigenvector Finally, the signal in the channel dimension is integrated and predicted through the spectrum generation module D3 containing a fully connected layer to obtain the predicted channel power spectrum density , It represents the predicted power spectrum density result of the current channel, which is used for spectrum recognition and detection in subsequent modules.

[0058] in,

[0059]

[0060]

[0061] Among them, the detailed structure of module D1 is: a fully connected layer, a normalization layer, a RELU activation function layer and a Dropout layer with a dimension of N; the detailed structure of module D2 is: a fully connected layer, a normalization layer, a RELU activation function layer and a Dropout layer with a dimension of the output dimension of the D1 module; the detailed structure of module D3 is a fully connected layer with an input dimension of the output dimension of D2.

[0062] The fully connected spectrum feature prediction extension module generates a wideband spectrum for each processing The power spectral density of one of the channels, therefore, needs to be processed segment data, and based on the preset splicing strategy, the channel power spectrum density predicted each time is Change the order of dimensions and splice them into a complete power spectral density result ,in, is the length of the continuous signal before division, that is, the frame length .

[0063] like Figure 2 As shown, Figure 2 (a) is a schematic diagram of the broadband spectrum in the input model; Figure 2 (b) Schematic diagram of the fully connected spectral feature extraction module and the fully connected spectral feature prediction expansion module; Figure 2 (c) Schematic diagram of the complete power spectral density results output by the model.

[0064] To facilitate those skilled in the art to further understand the satellite-to-ground channel detection bandwidth extension method of an autoencoder proposed in an embodiment of the present invention, the following is a description of the method. Figure 3~Figure 5 For further elaboration.

[0065] The bandwidth extension method of satellite-to-ground channel detection by an autoencoder mainly adopts a spectrum prediction network model based on end-to-end training, learns the attenuated signal spectrum outside the ADC sampling bandwidth in offline training, and uses a fully connected neural network to realize residual spectrum detection and spectrum prediction completion after the autoencoder neural network extracts the spectrum features. Figure 3 As shown, the training process of the spectrum prediction network model may include the following steps: Step S301, constructing an ADC sampling broadband signal data training set.

[0066] Step S302: construct a fully connected spectrum feature extraction module.

[0067] Step S303: construct a fully connected spectrum feature prediction extension module.

[0068] Step S304: training a spectrum prediction network model.

[0069] Step S305, train and verify the model effect multiple times.

[0070] Furthermore, if Figure 4 As shown, Figure 4 (a) is a schematic diagram of the spectrum detection result after the input model (the two ends of the frequency band are suppressed). Figure 4 (b) Based on Figure 4 (a) Schematic diagram of the ideal restoration result, Figure 4 (c) is a schematic diagram of the recovery result output by the embodiment of the present invention; Figure 5 same Figure 4 , to avoid redundancy, it will not be repeated here.

[0071] It can be seen that the satellite-to-ground channel detection bandwidth extension method of an autoencoder proposed in an embodiment of the present invention has the following advantages: (1) An embodiment of the present invention proposes a method for extending the bandwidth of satellite-to-ground channel detection using an autoencoder. By using a generative neural network to restore severely attenuated spectrum information, the position and shape of the spectrum signal before attenuation can be accurately located, and the prediction generation and completion of the residual spectrum signal can be achieved. This improves the spectrum prediction accuracy and realizes broadband spectrum perception under low-speed ADC sampling conditions.

[0072] (2) An embodiment of the present invention proposes a method for extending the bandwidth of a satellite-to-ground channel detection using an autoencoder, by decomposing the spectrum information into The channel is decomposed and processed, which reduces the required network processing depth and significantly reduces the complexity of network implementation.

[0073] (3) The embodiment of the present invention proposes a method for extending the bandwidth of a satellite-to-ground channel detection using an autoencoder. By using a preprocessing method for segmenting and intercepting ADC sampled broadband data, the types of data sets are enriched, and the amount of original data required for neural network training is greatly reduced.

[0074] According to an autoencoder satellite-to-ground channel detection bandwidth extension method proposed in an embodiment of the present invention, power spectrum density data can be obtained by acquiring a time domain data signal and preprocessing the time domain data signal; then, the power spectrum density features of the power spectrum density data can be extracted using a preset spectrum prediction network model; based on the power spectrum density features, the preset spectrum prediction network model is used to predict and complete the missing parts of the residual spectrum signal, thereby restoring the original spectrum shape. Thus, by learning the spectrum of broadband data in offline training, and after extracting the spectrum features, a fully connected neural network is used to realize residual spectrum detection and spectrum prediction completion, the problem of inaccurate spectrum prediction results caused by factors such as ADC reception sampling rate and broadband resolution in the prior art is solved, thereby improving the accuracy of spectrum prediction.

[0075] Next, a satellite-to-ground channel detection bandwidth extension device of an autoencoder according to an embodiment of the present invention is described with reference to the accompanying drawings.

[0076] Figure 6 It is a block diagram of a bandwidth extension device for satellite-to-ground channel detection of an autoencoder according to an embodiment of the present invention.

[0077] like Figure 6 As shown, the satellite-to-ground channel detection bandwidth extension device 10 of the autoencoder includes: an acquisition module 100, an extraction module 200 and a prediction module 300.

[0078] The acquisition module 100 is used to acquire a time domain data signal and pre-process the time domain data signal to obtain power spectrum density data; An extraction module 200 is used to extract power spectrum density features of power spectrum density data using a preset spectrum prediction network model, wherein the preset spectrum prediction network model is obtained by training a fully connected neural network using broadband data; The prediction module 300 is used to predict and complete the missing parts of the residual spectrum signal based on the power spectrum density characteristics and restore the original spectrum shape by using a preset spectrum prediction network model.

[0079] Furthermore, in some embodiments, before extracting the power spectrum density features of the power spectrum density data using the preset spectrum prediction network model, the extraction module 200 includes: A construction unit is used to obtain broadband data and construct a training test sample data group using the broadband data, wherein the broadband data is obtained by sampling using an analog-to-digital converter ADC; A preprocessing unit, used to preprocess the training and testing sample data group, and obtain a training set and a validation set based on the preprocessed training and testing sample data group; The training unit is used to train the fully connected neural network based on a preset loss function using a training set to obtain an initial neural network model, and to verify the initial neural network model using a verification set until the initial neural network model meets the preset standard, and then terminate the iterative training of the fully connected neural network to obtain a preset spectrum prediction network model, otherwise, continue the iterative training after adjusting the model parameters.

[0080] Furthermore, in some embodiments, the preprocessing unit is specifically configured to: The training test sample data group is divided into low-speed ADC sampling broadband data and high-speed ADC sampling broadband data at corresponding moments; Based on a preset frame length and a preset step length, the low-speed ADC sampling broadband data and the high-speed ADC sampling broadband data at the corresponding moment are respectively processed in segments, and the low-speed ADC sampling broadband data after the segment processing is used as training data, and the high-speed ADC sampling broadband data after the segment processing is used as verification data; Performing a fast Fourier transform process with a length of a first preset length on each row of data of each data group in the training data and the verification data to obtain a broadband spectrum; Taking the modulus value of the broadband spectrum, the power spectrum density estimation result is obtained; The power spectral density estimation result is divided into intervals of a second preset length to obtain a training set and a validation set.

[0081] Furthermore, in some embodiments, after obtaining the training set and the validation set based on the preprocessed training and testing sample data set, the preprocessing unit further includes: A processing subunit, used for normalizing the training set to obtain a normalized training set; A first activation subunit is used to activate each training sample in the normalized training set to obtain a first feature vector of each training sample; A second activation subunit is used to activate the channel dimension of the first feature vector of each training sample respectively to obtain a second feature vector of each training sample; The integration subunit is used to integrate the first feature vector of each training sample and the second feature vector of the corresponding training sample to obtain the channel feature result of each training sample.

[0082] Furthermore, in some embodiments, after obtaining the channel characteristic result of each training sample, the integration subunit is further used to: Activate the channel feature results of each training sample respectively to obtain the third feature vector of each training sample; Activate the channel dimension of the third eigenvector of each training sample respectively to obtain the fourth eigenvector of each training sample; Integrate and predict the third eigenvector of each training sample and the fourth eigenvector of the corresponding sample to obtain the predicted channel power spectrum density of each training sample; Based on the preset splicing strategy, the predicted channel power spectrum density of each training sample is spliced ​​to obtain a complete power spectrum density result.

[0083] Furthermore, in some embodiments, the preset loss function is:

[0084] in, For the complete power spectral density result, is the validation set, is the preset frame length, is the number of rows of the power spectral density result.

[0085] It should be noted that the above explanation of the embodiment of a method for extending the satellite-to-ground channel detection bandwidth of an autoencoder is also applicable to the device for extending the satellite-to-ground channel detection bandwidth of an autoencoder in this embodiment, which will not be repeated here.

[0086] According to an embodiment of the present invention, a bandwidth extension device for satellite-to-ground channel detection of an autoencoder is proposed. By acquiring a time domain data signal and preprocessing the time domain data signal, power spectrum density data can be obtained; then, the power spectrum density features of the power spectrum density data can be extracted using a preset spectrum prediction network model; based on the power spectrum density features, the preset spectrum prediction network model is used to predict and complete the missing parts of the residual spectrum signal, thereby restoring the original spectrum shape. Therefore, by learning the spectrum of broadband data in offline training, and after extracting the spectrum features, a fully connected neural network is used to realize residual spectrum detection and spectrum prediction completion, the problem of inaccurate spectrum prediction results caused by factors such as ADC receiving sampling rate and broadband resolution in the prior art is solved, thereby improving the accuracy of spectrum prediction.

[0087] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. The electronic device may include: A memory 701 , a processor 702 , and a computer program stored in the memory 701 and executable on the processor 702 .

[0088] When the processor 702 executes the program, the satellite-to-ground channel detection bandwidth extension method of the autoencoder provided in the above embodiment is implemented.

[0089] Furthermore, the electronic device further comprises: The communication interface 703 is used for communication between the memory 701 and the processor 702 .

[0090] The memory 701 is used to store computer programs that can be executed on the processor 702 .

[0091] The memory 701 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0092] If the memory 701, the processor 702 and the communication interface 703 are implemented independently, the communication interface 703, the memory 701 and the processor 702 can be connected to each other through a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0093] Optionally, in a specific implementation, if the memory 701, the processor 702 and the communication interface 703 are integrated on a chip, the memory 701, the processor 702 and the communication interface 703 can communicate with each other through an internal interface.

[0094] The processor 702 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention.

[0095] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the above-mentioned method for extending the bandwidth of satellite-to-ground channel detection of an autoencoder is implemented.

[0096] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the program is executed by a processor, it is used to implement the above-mentioned satellite-to-earth channel detection bandwidth extension method of an autoencoder.

[0097] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0098] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0099] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for extending bandwidth of satellite-to-ground channel detection of an autoencoder, characterized in that: The following steps are involved: Acquire a time domain data signal, and preprocess the time domain data signal to obtain power spectrum density data; Extracting power spectrum density features of the power spectrum density data using a preset spectrum prediction network model, wherein the preset spectrum prediction network model is obtained by training a fully connected neural network using broadband data; Based on the power spectral density characteristics, the preset spectrum prediction network model is used to predict and complete the missing parts in the residual spectrum signal to restore the original spectrum shape.

2. The method for extending the bandwidth of satellite-to-ground channel detection of an autoencoder according to claim 1, characterized in that: Before extracting the power spectrum density features of the power spectrum density data by using the preset spectrum prediction network model, the method further includes: Acquire the broadband data, and construct a training test sample data group using the broadband data, wherein the broadband data is sampled using an analog-to-digital converter ADC; Preprocessing the training and testing sample data group, and obtaining a training set and a validation set based on the preprocessed training and testing sample data group; Based on a preset loss function, the fully connected neural network is trained using the training set to obtain an initial neural network model, and the initial neural network model is verified using the verification set until the initial neural network model meets the preset criteria, and the iterative training of the fully connected neural network is terminated to obtain the preset spectrum prediction network model, otherwise the iterative training continues after adjusting the model parameters.

3. The method for extending the bandwidth of satellite-to-ground channel detection of an autoencoder according to claim 2, characterized in that: The preprocessing of the training test sample data set and obtaining a training set and a validation set based on the preprocessed training test sample data set includes: The training test sample data group is divided into low-speed ADC sampling broadband data and high-speed ADC sampling broadband data at corresponding moments; Based on a preset frame length and a preset step length, the low-speed ADC sampling broadband data and the high-speed ADC sampling broadband data at the corresponding moment are respectively processed in segments, and the low-speed ADC sampling broadband data after the segment processing is used as training data, and the high-speed ADC sampling broadband data after the segment processing is used as verification data; Performing a fast Fourier transform process with a length of a first preset length on each row of data of each data group in the training data and the verification data to obtain a broadband spectrum; Taking a modulus value of the broadband spectrum to obtain a power spectrum density estimation result; The power spectrum density estimation result is divided into intervals of a second preset length to obtain the training set and the validation set.

4. The method for extending the bandwidth of satellite-to-ground channel detection of an autoencoder according to claim 2, characterized in that: After obtaining the training set and the validation set based on the preprocessed training and testing sample data set, the method further includes: Normalizing the training set to obtain a normalized training set; Activating each training sample in the normalized training set to obtain a first feature vector of each training sample; Activating the channel dimensions of the first feature vector of each training sample respectively to obtain the second feature vector of each training sample; The first feature vector of each training sample and the second feature vector of the corresponding training sample are integrated to obtain a channel feature result of each training sample.

5. The method for extending the bandwidth of satellite-to-ground channel detection of an autoencoder according to claim 4, characterized in that: After obtaining the channel characteristic result of each training sample, the method further includes: Activating the channel feature results of each training sample respectively to obtain a third feature vector of each training sample; Activating the channel dimension of the third eigenvector of each training sample respectively to obtain the fourth eigenvector of each training sample; Integrate and predict the third eigenvector of each training sample and the fourth eigenvector of the corresponding sample to obtain a predicted channel power spectral density of each training sample; Based on a preset splicing strategy, the predicted channel power spectrum density of each training sample is spliced ​​to obtain a complete power spectrum density result.

6. The method for extending the bandwidth of satellite-to-ground channel detection of an autoencoder according to claim 2, characterized in that: The preset loss function is: in, For the complete power spectral density result, is the validation set, is the preset frame length, is the number of rows of the power spectral density result.

7. A bandwidth extension device for satellite-to-ground channel detection of an autoencoder, characterized in that: include: An acquisition module, used for acquiring a time domain data signal and preprocessing the time domain data signal to obtain power spectrum density data; An extraction module, used to extract the power spectrum density features of the power spectrum density data by using a preset spectrum prediction network model, wherein the preset spectrum prediction network model is obtained by training a fully connected neural network with broadband data; The prediction module is used to predict and complete the missing parts in the residual spectrum signal based on the power spectral density characteristics and restore the original spectrum shape by using the preset spectrum prediction network model.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a method for extending bandwidth of satellite-to-ground channel detection of an autoencoder as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement a bandwidth extension method for satellite-to-ground channel detection of an autoencoder as described in any one of claims 1-6.

10. A computer program product, characterized in that It includes a computer program, which, when executed by a processor, is used to implement the satellite-to-ground channel detection bandwidth extension method of an autoencoder as described in any one of claims 1-6.

Images