Pilot frequency design scheme in satellite scene, OTFS joint channel estimation and data detection method and model based on deep learning

Abstract

The invention discloses an OTFS joint channel estimation and data detection method and model based on deep learning in a satellite scene. The method comprises the steps of constructing a data set, designing a pilot frequency design scheme, constructing a joint channel estimation and data detection deep learning model, repeatedly training the deep learning model, obtaining channel state information and data symbols through prediction data, comparing a network prediction result with the corresponding channel state information and data symbol labels, and determining the channel state information and the data symbols. Evaluating the output of the deep learning model through the minimum mean square error and the cross entropy loss; the obtained optimized deep learning model is used for obtaining channel state information and data symbols through verification data. According to the method, OTFS received signals are learned through the deep learning network so as to complete channel estimation and data detection tasks at the same time, a pilot frequency design scheme and a data slicing processing scheme are adopted at the same time, high performance is kept, the accuracy of channel estimation and data detection is remarkably improved, and the overhead of pilot frequency resources is reduced at the same time.

Description

Technical Field

[0001] The present invention belongs to the technical field of OTFS modulation communication channel estimation, and specifically relates to a pilot design scheme in a satellite scenario, and an OTFS joint channel estimation and data detection method and model based on deep learning. Background Art

[0002] Satellite communications are an important part of the integrated space-ground network, providing wide coverage and reliability under different geographical conditions. Traditional OFDM (Orthogonal Frequency Division Multiplexing) modulation can achieve higher communication rates through subcarrier multiplexing in ground-side communications. However, in satellite scenarios, especially low-orbit satellites, the high Doppler frequency deviation will cause the orthogonality of subcarriers in OFDM modulation to be lost, resulting in reduced performance in terms of computational efficiency and accuracy.

[0003] OTFS (Orthogonal Time Frequency Space Modulation), as an OFDM evolution modulation technology, can transform the time-frequency domain time-varying channel into a time-independent channel in the delay-Doppler domain through ISFFT (Inverse SymplecticFinite Fourier Transform) and SFFT (SymplecticFinite Fourier Transform), so that all transmitted data obtain the same channel gain to improve performance in high-speed mobile scenarios. OTFS modulation has been proven to provide communication rate and bit error rate performance that exceeds OFDM modulation in high-mobility scenarios.

[0004] In the field of wireless communications, efficient and accurate channel estimation and data detection have become the key to improving communication quality and data transmission rate. However, traditional OFDM channel estimation and data detection methods are not suitable for OTFS. Therefore, some scholars have designed a pilot design scheme based on the guard interval according to the relevant characteristics of OTFS, and used a threshold-based method to complete the channel estimation of OTFS, and used the message passing algorithm for data detection tasks. However, due to the introduction of the guard interval, the pilot design scheme based on the guard interval has an obvious problem of low spectrum utilization, so this method is difficult to apply in power-limited satellite scenarios. At the same time, some scholars have proposed a superimposed pilot design scheme, which superimposes pilot symbols with data symbols and cancels the setting of the guard interval, thereby improving spectrum utilization and reducing pilot resource overhead. However, due to the cancellation of the guard interval, the algorithm complexity of the channel estimation and signal demodulation at the receiving end increases and the performance is degraded. With the increasing requirements of communication technology for communication performance and resource overhead, the limitations of traditional methods are becoming increasingly prominent. Therefore, it is necessary to design novel pilot design methods, channel estimation and data detection methods to improve performance indicators while improving spectrum utilization as much as possible.

[0005] In the computer field, artificial intelligence represented by deep learning networks has shown great potential in the fields of image and data prediction; however, applying it to the fields of channel estimation and data detection, and using its nonlinear characteristics to learn the received signal and then perform channel estimation and data detection, is a brand-new research method. At present, there are also some related studies on using deep learning networks for OTFS communication, such as: patent application CN117424782A, which uses convolutional neural networks to perform noise reduction in the protection interval area, and then performs channel estimation on the noise-reduced data. It mainly focuses on using convolutional networks for noise reduction, and focuses more on the noise reduction process. After obtaining the noise-reduced data, an additional channel estimation algorithm is required to obtain the channel state information, which has a high computational complexity and cannot achieve data detection tasks. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a pilot design scheme in a satellite scenario, an OTFS joint channel estimation and data detection method and model based on deep learning, which solves the problems of low satellite communication spectrum utilization and high channel estimation calculation complexity in the prior art.

[0007] The present invention adopts the following technical solutions to solve the above technical problems:

[0008] The pilot design scheme in the satellite scenario adds a guard interval at the pilot symbol position, which is specifically expressed by the following formula:

[0009]

[0010] in, represents the transmitted signal in the delay-Doppler domain, denote Doppler and delay indices respectively, They represent the pilot symbol and the Doppler index respectively. , the delay index is The data symbol, They represent the Doppler and delay indexes of the pilot symbols respectively.

[0011] The OTFS joint channel estimation and data detection method based on deep learning in satellite scenarios includes the following steps:

[0012] S1. According to the pilot design scheme of claim 1, constructing a data set including a training set and a test set;

[0013] S2. Construct a deep learning model for joint channel estimation and data detection, which includes a coarse channel estimation network model, an iterative data detection network model, and a data-assisted channel estimation network model;

[0014] S3, inputting the training data of different networks in the training set into the corresponding network models respectively, training the feature extraction function of the deep learning model, and enabling the deep learning model to predict the channel state information and data symbol output from the input data;

[0015] S4, recovering an equivalent channel matrix from the output channel state information, comparing the equivalent channel matrix with its corresponding channel state information label, and evaluating the output of the coarse channel estimation network model and the data-assisted channel estimation network model;

[0016] S5, comparing the output data symbol with the label of its transmitted signal, and evaluating the output of the iterative data detection network model;

[0017] S6. After repeating steps S3 to S5 for a set number of times, an optimized joint channel estimation and data detection deep learning model is obtained;

[0018] S7. After the received signal obtained by the receiving antenna in real time is input into the trained model, the channel state information and data detection results are obtained.

[0019] The dataset is constructed as follows:

[0020] Firstly, a two-dimensional matrix of size M×N is used to simulate the OTFS data frame in the delay-Doppler domain, which represents the data symbols sent from the transmitting antenna to the receiving antenna in the actual communication system.

[0021] Next, data symbols and pilot symbols are inserted according to the pilot design scheme; the inserted matrix is ​​subjected to ISFFT transformation and Heisenberg transformation in turn, and the delay-Doppler domain signal is converted into a time-frequency domain signal and a time domain signal in turn;

[0022] After the transmitting antenna data is constructed, the channel parameters are set and initialized according to the NTN-TDL scenario in the 3GPP standard to complete the channel model construction;

[0023] The transmission signal of the transmitting antenna is input into the channel model to further obtain the receiving signal at the receiving antenna affected by the channel and environmental noise;

[0024] After the receiving antenna obtains the output result of the channel model, it transforms the time domain signal into the time-frequency domain signal and the delay-Doppler domain signal through Wigner transformation and SFFT transformation in sequence;

[0025] Based on the obtained delay-Doppler domain signal, the received signal data is stored to construct the training set and test set of the deep learning model.

[0026] The specific process of step S2 is as follows:

[0027] The coarse channel estimation network is built through the pilot channel extraction module and the denoising sub-network;

[0028] The iterative data detection network is built through the interference elimination module, channel feature extraction sub-network, slice processing module, and data detection sub-network;

[0029] The data-assisted channel estimation network is constructed through the data preprocessing module and the iterative channel estimation subnetwork;

[0030] The coarse channel estimation network, the iterative data detection network and the data-aided channel estimation network are connected in series to obtain a complete joint channel estimation and data detection network.

[0031] The specific method of training the deep learning model in step S3 is as follows:

[0032] First, the received signal is preprocessed, the real part and the imaginary part of the received signal are split and then spliced ​​in the data channel dimension to form tensor data;

[0033] For the coarse channel estimation network, the preprocessed received signal is directly selected as the input training set data of the coarse channel estimation network;

[0034] For the iterative data detection network, the preprocessed received signal and channel state information label data are selected as input training set data of the iterative data detection network;

[0035] For a data-assisted channel estimation network, a preprocessed received signal and data symbol label data are selected as input training set data of the data-assisted channel estimation network;

[0036] At the same time, the training set data of different networks are input into the corresponding network models at the same time, and the three different networks are trained in parallel.

[0037] The minimum mean square error is used as the loss function to train the channel estimation network model. The specific formula is as follows:

[0038]

[0039] in, represents the loss value of the minimum mean square error loss function, Refers to the equivalent channel matrix prediction output of a fully connected deep learning network, refers to the label equivalent channel matrix;

[0040] The cross entropy loss is used as the loss function to train the data detection model. The specific formula is as follows:

[0041]

[0042] in represents the loss value of the cross entropy loss function, Represents the number of all data symbols, is the category of the data symbol, is the symbolic function, Representative observation sample Belongs to category The predicted probability of .

[0043] The calculated value of the loss function is back-propagated to update the deep learning model parameters. The update process is specifically expressed by the following formula:

[0044]

[0045] in, For the The parameters of the deep learning model, is the training learning rate, is the first-order derivative of the deep learning model parameters, is the output of the loss function, including and .

[0046] The OTFS joint channel estimation and data detection model based on deep learning includes a coarse channel estimation network, an iterative data detection network and a data-assisted channel estimation network connected in series, wherein:

[0047] The coarse channel estimation network includes a pilot signal extraction module and a noise reduction subnetwork;

[0048] The iterative data detection network includes an interference elimination module, a channel feature extraction subnetwork, a feature fusion module, a data slicing module and a data detection subnetwork;

[0049] The data-assisted channel estimation network includes a data preprocessing module and an iterative channel estimation subnetwork.

[0050] The denoising subnetwork, the data detection subnetwork and the iterative channel estimation subnetwork are all composed of a U-Net network; the feature extraction subnetwork is composed of multi-layer convolutional units.

[0051] A computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, call all or part of the steps of the method.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. The present invention designs a novel pilot design scheme. Compared with the traditional embedded pilot design scheme, this scheme significantly improves the pilot resource utilization rate. Compared with the traditional superimposed pilot design scheme, this scheme significantly reduces the interference between pilot and data symbols, significantly improves the performance indicators of the system, and the pilot design scheme of this scheme can be used as a module for other channel estimation algorithms.

[0054] 2. The method described in the present invention models the channel estimation task and the data detection task in OTFS into image denoising and pixel segmentation tasks respectively, and jointly processes the two tasks, so that the original channel estimation and data detection tasks complement each other, improve the performance indicators of the method, and make the method more generalizable and robust.

[0055] 3. The pilot design scheme and data slicing operation methods proposed in the present invention significantly improve the system performance compared with traditional algorithms. At the same time, they are universal and can be widely used in other channel estimation and data detection methods to help improve performance indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 Schematic diagram of a joint channel estimation and data detection pilot design solution in an embodiment of the present invention.

[0057] Figure 2 It is a framework flow chart of the deep learning network model for joint channel estimation and data detection in an embodiment of the present invention.

[0058] Figure 3Flow chart of data slicing operation in an embodiment of the present invention.

[0059] Figure 4 This is a simulation diagram of the present invention, which compares and tests the normalized minimum mean square error performance indicators of different pilot design schemes under different signal-to-noise ratios and different pilot-data power ratios in the channel estimation task.

[0060] Figure 5 This is a simulation diagram of the present invention, which compares and tests the normalized minimum mean square error performance indicators of different channel estimation algorithms under different signal-to-noise ratios in the channel estimation task.

[0061] Figure 6 This is a simulation diagram of the present invention, which compares and tests the bit error rate performance indicators of different data detection algorithms under different signal-to-noise ratios in the data detection task.

[0062] Figure 7 This is a simulation diagram of the present invention, which compares and tests the bit error rate performance indicators of different deep learning networks using data slicing operations in data detection tasks. DETAILED DESCRIPTION

[0063] To further help understand the technical solution of the present invention, exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. It should be understood that all of these exemplary embodiments described are only partial embodiments and examples of the present invention, rather than all of them. On the contrary, these exemplary embodiments are provided so that those skilled in the art can more thoroughly understand the present invention and can more completely convey the technical content of the present invention to those skilled in the art.

[0064] In the following detailed description, many specific details are set forth to provide a more thorough understanding of the present invention. However, it is obvious to those skilled in the art that well-known algorithms and models are not shown in detail to avoid obscuring the subject matter of the present invention.

[0065] Specific embodiments, such as Figures 1 to 7 As shown,

[0066] Embodiment 1,

[0067] The pilot design scheme provided in this example designs a guard interval at the pilot symbol position, such as Figure 1 As shown, the mathematical representation of the pilot design scheme is as follows:

[0068]

[0069] in, represents the transmitted signal in the delay-Doppler domain, denote Doppler and delay indices respectively, They represent the pilot symbol and the Doppler index respectively. , the delay index is The data symbol, They represent the Doppler and delay indexes of the pilot symbols respectively.

[0070] The pilot design scheme provided in this embodiment can make the output result better than other pilot design schemes while using fewer pilot resources. At the same time, when the proposed pilot design scheme is applied to other channel estimation algorithms, there is also a significant performance improvement. The pilot design scheme involved in the present invention has the ability to help the OTFS channel estimation algorithm improve performance indicators while using lower pilot resources.

[0071] Embodiment 2:

[0072] This embodiment provides an OTFS joint channel estimation and data detection method and model based on deep learning in a satellite scenario, which specifically includes the following steps:

[0073] S1. According to the pilot design scheme described in Example 1, a data set is constructed and divided into a training set and a test set: based on the standard protocol provided by 3GPP, the OTFS modulation communication data obtained by simulation is preprocessed to obtain the training set data and its corresponding validation set label data for model training, and these data are processed and mapped into the training set and test set of the deep neural network;

[0074] The specific process is as follows:

[0075] In this embodiment, firstly, a plurality of M N two-dimensional matrix to simulate the OTFS data frame in the delay-Doppler domain, which represents the data symbols that need to be sent by the transmitting antenna in the actual communication system;

[0076] Next, according to the designed pilot design scheme, pilot symbols, data symbols and guard intervals are inserted, and the constellation mapping is completed according to the modulation method. After the symbol insertion is completed, the two-dimensional matrix is ​​subjected to ISFFT transformation and Heisenberg transformation in turn, and the delay-Doppler domain signal is converted into the time-frequency domain signal and the time domain signal in turn;

[0077] The channel parameters are set according to the NTN-TDL scenario in the 3GPP standard, mainly including the number of transmitting antennas, the number of receiving antennas, the number of channel paths, the channel path gain, the channel path delay, and the channel path Doppler, to complete the construction of the channel model.

[0078] The time domain transmission signal is input into the channel model through the transmission antenna, and is transmitted through the channel to further obtain the time domain reception signal of the receiving antenna;

[0079] After the receiving antenna obtains the output of the channel model, the time domain signal is converted into a time-frequency domain signal and a delay-Doppler domain signal in turn through the Wigner transformation and SFFT transformation. The receiving antenna data is collected, and the channel model parameters, the original data of the transmitting antenna, etc. are stored at the same time to construct the training set and test set of the deep learning model.

[0080] Furthermore, the specific steps are:

[0081] OTFS data frame in satellite scenario (i.e., transmitted signal in the time delay-Doppler domain) The construction is as follows:

[0082]

[0083] Assume that during the communication process, the OTFS transmitter first sends the delay-Doppler symbol Mapping to time-frequency domain signal through ISFFT transformation The process is as follows:

[0084]

[0085] in {0, 1, … , -1}, {0, 1, … , -1} denotes time and frequency indexes, respectively.

[0086] Time-frequency domain signal Through Heisenberg transform, sampling is converted into time domain signal , so that the transmitter sends:

[0087]

[0088] in is the transmitting pulse waveform, is the duration of one OTFS data frame, is the subcarrier spacing.

[0089] Time domain signal By characterizing the delay and Doppler Time-varying channel , receiving signal It is expressed as:

[0090]

[0091] in represents additive Gaussian noise, based on which the time-varying channel It can be expressed as:

[0092]

[0093] in, is the number of propagation paths, , and They are The gain, delay and Doppler frequency deviation of each propagation path;

[0094] in, and They can be expressed as:

[0095]

[0096] in , }, representing the The delay taps and Doppler taps of the channel paths.

[0097] After the receiver receives the signal, it calculates the cross-mixing function through matched filtering After processing, the time domain received signal is obtained, which is expressed as follows:

[0098]

[0099]

[0100] in To receive the pulse waveform, and Represent the time and frequency index respectively.

[0101] Finally, the time-frequency domain received signal received by the receiver is Convert the received signal into the delay-Doppler domain through SFFT , expressed as follows:

[0102]

[0103] After obtaining the delay-Doppler domain received signal, the original transmitted signal, channel parameters and received signal are collected and stored to construct the training set and test set data sets.

[0104] S2. Construct a deep learning model for joint channel estimation and data detection, which includes a coarse channel estimation network model, an iterative data detection network model, and a data-assisted channel estimation network model;

[0105] The deep learning model framework process described in this embodiment is as follows Figure 2 As shown, it mainly includes a coarse channel estimation network, an iterative data detection network and a data-assisted channel estimation network;

[0106] Furthermore, the specific details of the deep learning network model of the joint channel estimation and data detection network are as follows:

[0107] The coarse channel estimation network is constructed by the pilot signal extraction module and the denoising subnetwork;

[0108] The iterative data detection network is constructed by the interference elimination module, the channel feature extraction sub-network, the feature fusion module, the data slicing module and the data detection sub-network;

[0109] The data-assisted channel estimation network is constructed by a data preprocessing module and an iterative channel estimation subnetwork;

[0110] The coarse channel estimation network, the iterative data detection network and the data-aided channel estimation network are connected in series to obtain a complete joint channel estimation and data detection network.

[0111] Furthermore, the denoising subnetwork, data detection subnetwork, and iterative channel estimation subnetwork are all composed of U-Net networks. The U-Net network structure is as follows: Figure 2 As shown, a U-shaped structure is formed by a symmetrical encoder and decoder, and its specific structure is as follows:

[0112] The encoder consists of a series of Figure 2 The convolutional units shown in the figure are composed of a convolutional layer, a batch normalization layer, and a rectified linear unit (ReLU). After a series of convolutional units, a stride of 1 and size of 2 A pooling operation of 2 is used for downsampling. In addition, with each downsampling operation, the feature channel dimension of the data is doubled to facilitate the network to learn high-dimensional features. In the decoder part, all operations will be mirrored to the encoder. After a series of convolutional end-units, a stride of 1 and size of 2 A transposed convolution operation of 2 is used for upsampling. At the same time, the feature channel dimension of the data will be halved.

[0113] In addition, the feature extraction subnetwork is also composed of multiple layers of convolutional units, such as Figure 2 shown.

[0114] S3. Input the training data of different networks in the training set into the corresponding network models respectively, train the feature extraction function of the deep learning model, and enable the deep learning model to predict the channel state information and data symbol output from the input data; the specific process is as follows:

[0115] The input data of the coarse channel estimation network is the received signal , receiving signal First, the pilot signal is extracted through the pilot signal extraction module. ; Then, the pilot signal is input into the denoising subnetwork to eliminate the interference of data symbols and noise through the network to obtain a clean predicted pilot signal .

[0116] The input data of the iterative data detection network is The predicted pilot signal and receive signals , receiving signal First, the pilot signal is eliminated by the interference elimination module interference; at the same time, the channel feature extraction sub-network extracts the pilot signal After extracting the channel state information features, it is concatenated with the received signal after interference elimination in the channel dimension. Subsequently, the data is input into the data slicing module, and after loop filling, data slicing and other operations, it is input into the data detection subnetwork. Finally, the data detection subnetwork obtains the first Second prediction emission signal .

[0117] The input data of the data-assisted channel estimation network is Second prediction emission signal And receive signal , and After data preprocessing and feature fusion, it is input into the iterative channel estimation subnetwork and finally the first The predicted pilot signal .

[0118] Furthermore, the specific details of each module in the joint channel estimation and data detection network are as follows:

[0119] The pilot extraction module is from a size of M The received signal of N Extract the size of The pilot signal is and Represent the maximum Doppler and delay offset of the channel path respectively, and the index of this area satisfies .

[0120] The interference elimination module converts the pilot signal matrix From the received signal matrix Eliminate the pure received signal , in order to avoid the interference of pilot symbols on data symbols in the data detection task, which is specifically expressed as follows:

[0121]

[0122] in .

[0123] The data slicing module mainly includes loop filling operations and data slicing operations, such as Figure 3 As shown, the specific process is as follows:

[0124] The loop fill operation will receive the signal Cyclic padding is performed in the Doppler dimension to help the network better learn the features of the data edge, which is specifically expressed as follows:

[0125]

[0126] in Represents the data after loop filling, Represents the characteristic dimension of the data, is the Doppler index after circular filling, is the size of the slicing window.

[0127] The data slicing operation will loop the received signal after filling Split into The data fragments, of which Data fragments It is expressed as follows:

[0128]

[0129] The data preprocessing module remaps the transmission signal predicted by the network according to the modulation constellation diagram. Taking QAM modulation as an example, the specific process is as follows:

[0130]

[0131] in represents the emission signal predicted by the network, Indicates the constellation type in QAM modulation.

[0132] S4, recovering an equivalent channel matrix from the output channel state information, comparing the equivalent channel matrix with its corresponding channel state information label, and evaluating the output of the coarse channel estimation network model and the data-assisted channel estimation network model; specifically,

[0133] After obtaining the predicted output of the channel estimation task, the predicted result is compared with its corresponding channel state information label, and the output of the coarse channel estimation network model and the data-assisted channel estimation network model is evaluated through the loss function:

[0134] In the channel estimation task, the minimum mean square error is used as the loss function, which is defined as follows:

[0135] in, Refers to the channel matrix predicted by the network, refers to the label channel matrix.

[0136] MSE measures the error between the predicted output of the deep learning model and the expected label data. If the network does not output the correct result, its value will be large. Therefore, this measurement indicator can be used to constrain the deep learning network to update the deep learning model parameters.

[0137] S5. Compare the output data symbol with the label of its transmitted signal and evaluate the output of the iterative data detection network model; specifically,

[0138] After obtaining the predicted output of the data detection task, the detection result is compared with the label of the transmitted signal, and the output of the iterative channel estimation network model is evaluated through the loss function:

[0139] In the data detection task, cross entropy loss is used as the loss function, which is defined as follows:

[0140]

[0141] The cross entropy loss measures the error between the predicted output of the deep learning model and the expected label data in the classification task. The smaller the probability that the network predicts the correct data symbol, the larger its value. Therefore, this measurement indicator can be used to update the parameters of the iterative data detection network model.

[0142] S6. After repeating steps S3 to S5 for a set number of times, an optimized joint channel estimation and data detection deep learning model is obtained; specifically,

[0143] The deep learning model performs back propagation through the loss function calculation value and continuously updates the deep learning model parameters. The specific update process is as follows:

[0144] ;

[0145] By observing the channel prediction results output by the deep learning network on the training data set and comparing the prediction results with the actual channel, we can see the training status of the deep learning network;

[0146] Repeatedly update the deep learning model parameters until all data in the training set are input;

[0147] Repeat the above process until the maximum number of iterations is reached.

[0148] S7. After the received signal obtained by the receiving antenna in real time is input into the trained model, the channel state information and data detection results are obtained. Furthermore, the received signal of the receiving antenna is input into the trained joint channel estimation and data detection network to obtain the channel state information and predicted data symbols, thereby completing the channel estimation and data detection tasks.

[0149] In order to further verify the results of this method and model, simulation experiments were conducted and compared with the existing technology, as follows:

[0150] Continuously adjust variables such as data signal-to-noise ratio and pilot signal-to-noise ratio, and compare them with the channel estimation results of other algorithms. Figure 4 It can be seen that the pilot design scheme provided in this embodiment is better than other pilot design schemes in that the output result is still better than other pilot design schemes when using fewer pilot resources. At the same time, when the proposed pilot design scheme is applied to other channel estimation algorithms, there is also a significant performance improvement. This result shows that the pilot design scheme involved in the present invention has the ability to help the OTFS channel estimation algorithm improve performance indicators while using lower pilot resources.

[0151] Furthermore, from Figure 5 It can be seen that the joint channel estimation and data detection network provided by this embodiment has a performance index that is still on par with other channel estimation algorithms when the pilot signal-to-noise ratio is 3dB lower than that of other channel estimation algorithms. Under the same pilot signal-to-noise ratio, the performance is significantly better than other algorithms, and the number of iterative estimates is less. This result shows that the joint channel estimation and data detection network in the present invention has better performance.

[0152] In the data detection task, Figure 6 It can be seen that the algorithm provided by this embodiment has a significant performance improvement compared with other data detection algorithms under the same pilot signal-to-noise ratio. At the same time, the algorithm provided by this embodiment has a performance index that is more in line with the ideal situation. This result shows that the joint channel estimation and data detection network of the present invention has better performance and robustness.

[0153] Further, from Figure 7 It can be seen that the performance of different networks has been significantly improved after using the proposed data slicing operation. This result shows that the data slicing operation involved in the present invention is also applicable to other algorithms and can help improve the performance of other algorithms, and has universality and generalization.

[0154] If the above functions are implemented in the form of 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 technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program codes.

[0155] Therefore, a computer-readable storage medium is disclosed, on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor, all or part of the steps of the method are invoked.

[0156] The above describes the preferred embodiments of the present invention. It should be understood that the present invention is not limited to the above-mentioned specific embodiments, and the devices and structures that are not described in detail should be understood to be implemented in a common manner in the art; any technician familiar with the art can use the above-disclosed methods and technical contents to make many possible changes and modifications to the technical solutions of the present invention without departing from the scope of the technical solutions of the present invention, or modify them into equivalent embodiments of equivalent changes, which does not affect the essential content of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention are still within the scope of protection of the technical solutions of the present invention.

Claims

1. A pilot design scheme in a satellite scenario, characterized in that: A guard interval is added at the pilot symbol position, which is specifically expressed by the following formula: ， in, represents the transmitted signal in the delay-Doppler domain, denote Doppler and delay indices respectively, They represent the pilot symbol and the Doppler index respectively. , the delay index is The data symbol, They represent the Doppler and delay indexes of the pilot symbols respectively.

2. The OTFS joint channel estimation and data detection method based on deep learning in satellite scenarios is characterized by comprising the following steps: S1. According to the pilot design scheme of claim 1, constructing a data set including a training set and a test set; S2. Construct a deep learning model for joint channel estimation and data detection, which includes a coarse channel estimation network model, an iterative data detection network model, and a data-assisted channel estimation network model; S3, inputting the training data of different networks in the training set into the corresponding network models respectively, training the feature extraction function of the deep learning model, and enabling the deep learning model to predict the channel state information and data symbol output from the input data; S4, recovering an equivalent channel matrix from the output channel state information, comparing the equivalent channel matrix with its corresponding channel state information label, and evaluating the output of the coarse channel estimation network model and the data-assisted channel estimation network model; S5, comparing the output data symbol with the label of its transmitted signal, and evaluating the output of the iterative data detection network model; S6. After repeating steps S3 to S5 for a set number of times, an optimized joint channel estimation and data detection deep learning model is obtained; S7. After the received signal obtained by the receiving antenna in real time is input into the trained model, the channel state information and data detection results are obtained.

3. The OTFS joint channel estimation and data detection method based on deep learning in satellite scenarios according to claim 2, characterized in that: The dataset is constructed as follows: Firstly, a two-dimensional matrix of size M×N is used to simulate the OTFS data frame in the delay-Doppler domain, which represents the data symbols sent from the transmitting antenna to the receiving antenna in the actual communication system. Next, data symbols and pilot symbols are inserted according to the pilot design scheme; the inserted matrix is ​​subjected to ISFFT transformation and Heisenberg transformation in turn, and the delay-Doppler domain signal is converted into a time-frequency domain signal and a time domain signal in turn; After the transmitting antenna data is constructed, the channel parameters are set and initialized according to the NTN-TDL scenario in the 3GPP standard to complete the channel model construction; The transmission signal of the transmitting antenna is input into the channel model to further obtain the receiving signal at the receiving antenna affected by the channel and environmental noise; After the receiving antenna obtains the output result of the channel model, it transforms the time domain signal into the time-frequency domain signal and the delay-Doppler domain signal through Wigner transformation and SFFT transformation in sequence; Based on the obtained delay-Doppler domain signal, the received signal data is stored to construct the training set and test set of the deep learning model.

4. The OTFS joint channel estimation and data detection method based on deep learning in satellite scenarios according to claim 2, characterized in that: The specific process of step S2 is as follows: The coarse channel estimation network is built through the pilot channel extraction module and the noise reduction sub-network; The iterative data detection network is built through the interference elimination module, channel feature extraction sub-network, slice processing module, and data detection sub-network; The data-assisted channel estimation network is constructed through the data preprocessing module and the iterative channel estimation subnetwork; The coarse channel estimation network, the iterative data detection network and the data-aided channel estimation network are connected in series to obtain a complete joint channel estimation and data detection network.

5. The OTFS joint channel estimation and data detection method based on deep learning in satellite scenarios according to claim 2, characterized in that: The specific method of training the deep learning model in step S3 is as follows: First, the received signal is preprocessed, the real part and the imaginary part of the received signal are split and then spliced ​​in the data channel dimension to form tensor data; For the coarse channel estimation network, the preprocessed received signal is directly selected as the input training set data of the coarse channel estimation network; For the iterative data detection network, the preprocessed received signal and channel state information label data are selected as input training set data of the iterative data detection network; For a data-assisted channel estimation network, a preprocessed received signal and data symbol label data are selected as input training set data of the data-assisted channel estimation network; At the same time, the training set data of different networks are input into the corresponding network models at the same time, and the three different networks are trained in parallel.

6. The OTFS joint channel estimation and data detection method based on deep learning in satellite scenarios according to claim 2, characterized in that: The minimum mean square error is used as the loss function to train the channel estimation network model. The specific formula is as follows: ， in, represents the loss value of the minimum mean square error loss function, Refers to the equivalent channel matrix prediction output of a fully connected deep learning network, refers to the label equivalent channel matrix; The cross entropy loss is used as the loss function to train the data detection model. The specific formula is as follows: ， in represents the loss value of the cross entropy loss function, Represents the number of all data symbols, is the category of the data symbol, is the symbolic function, Representative observation sample Belongs to category The predicted probability of .

7. The OTFS joint channel estimation and data detection method based on deep learning in satellite scenarios according to claim 6, characterized in that: The calculated value of the loss function is back-propagated to update the deep learning model parameters. The update process is specifically expressed by the following formula: ， in, For the The parameters of the deep learning model, is the training learning rate, is the first-order derivative of the deep learning model parameters, is the output of the loss function, including and .

8. The OTFS joint channel estimation and data detection model based on deep learning is characterized by: It includes a coarse channel estimation network, an iterative data detection network and a data-aided channel estimation network connected in series, wherein: The coarse channel estimation network includes a pilot signal extraction module and a noise reduction subnetwork; The iterative data detection network includes an interference elimination module, a channel feature extraction subnetwork, a feature fusion module, a data slicing module and a data detection subnetwork; The data-assisted channel estimation network includes a data preprocessing module and an iterative channel estimation subnetwork.

9. The OTFS joint channel estimation and data detection model based on deep learning according to claim 8, characterized in that: The denoising subnetwork, the data detection subnetwork and the iterative channel estimation subnetwork are all composed of a U-Net network; the feature extraction subnetwork is composed of multi-layer convolutional units.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by a processor, all or part of the steps of the method described in any one of claims 2 to 7 are called.

Images