Frequency offset estimation method, communication node and storage medium

Abstract

Provided in the present application are a frequency offset estimation method, a communication node and a storage medium The method comprises: determining a modulation mode corresponding to received data; and inputting the received data into a frequency offset estimation model based on a neural network that corresponds to the modulation mode, so as to obtain a frequency offset estimation result.

Description

Frequency offset estimation method, communication node and storage medium Technical Field

[0001] The present application relates to the field of wireless communication technology, for example, to a frequency offset estimation method, a communication node and a storage medium. Background Art

[0002] Orthogonal Frequency Division Multiplexing (OFDM) technology uses multiple orthogonal subcarriers to transmit data in parallel, significantly reducing the data rate on each path and adding a time guard interval. This provides strong resistance to multipath interference and frequency selective fading. However, due to the orthogonality of the frequency subcarriers in the OFDM system, it is sensitive to channel frequency offset, making frequency offset one of the important parameters affecting communication quality. The terminal side needs to measure the frequency offset and perform phase compensation on the received signal to improve communication quality. Currently, the main frequency offset estimation algorithms include those based on the primary synchronization signal, the demodulation reference signal, and the cyclic prefix. However, the accuracy of these algorithms is relatively low and may not meet actual service requirements, affecting communication quality.

[0003] Summary of the Invention

[0004] The present application provides a frequency offset estimation method, a communication node, and a storage medium.

[0005] This embodiment of the present application provides a frequency offset estimation method, including:

[0006] Determine the modulation mode corresponding to the received data;

[0007] The received data is input into a frequency offset estimation model based on a neural network corresponding to the modulation mode to obtain a frequency offset estimation result.

[0008] An embodiment of the present application further provides a communication node, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned frequency offset estimation method when executing the program.

[0009] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the above-mentioned frequency offset estimation method is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] FIG1 is a flow chart of a frequency offset estimation method provided by an embodiment;

[0011] FIG2 is a schematic diagram illustrating an implementation of a frequency offset estimation method based on a neural network provided by an embodiment;

[0012] FIG3 is a schematic diagram illustrating an implementation of another frequency offset estimation method based on a neural network provided by an embodiment;

[0013] FIG4 is a schematic structural diagram of a frequency offset estimation device provided by an embodiment;

[0014] FIG5 is a schematic diagram of the hardware structure of a communication node provided by an embodiment. DETAILED DESCRIPTION

[0015] The present application is described below in conjunction with the accompanying drawings and embodiments. It will be understood that the specific embodiments described herein are merely intended to explain the present application and are not intended to limit the present application. It should be noted that, unless there is a conflict, the embodiments and features within the embodiments of the present application may be combined with each other in any manner. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present application, not all structures.

[0016] The present application provides a frequency offset estimation method for improving the performance of the frequency offset estimation module and improving the adaptability and generalization of the network model. On the basis of wireless communication technology, AI machine learning or deep learning technology is used to improve the communication function module in the related technology to improve the communication quality. With the original communication architecture as the bottom layer and retaining the original communication model or algorithm model, the amount of information required for model training or iteration can be reduced. At the same time, compared with the method of using a deep learning module to completely replace one or more communication modules in a black box manner, the method of using an AI module to replace part of the communication module based on the communication model in the related technology has better generalization and adaptability, and also retains the operability and readability of the module, and has better performance.

[0017] FIG1 is a flowchart of a frequency offset estimation method provided by an embodiment. As shown in FIG1 , the method provided by this embodiment includes steps 110 to 130 .

[0018] In step 110, a modulation mode corresponding to the received data is determined.

[0019] In step 120, the received data is input into a frequency offset estimation model based on a neural network corresponding to the modulation mode to obtain a frequency offset estimation result.

[0020] In this embodiment, the received data can be equalized data in a traditional communication system, or it can be receiving end data sent after preprocessing in a new pilot-free communication system; the received data is used as the input of the neural network. When the modulation mode of the received data is determined, the corresponding frequency offset estimation model is used to perform frequency offset estimation. For example, the received data can be input into a modulation detection model based on a neural network to detect the modulation mode of the received data, or the indicated modulation mode can be determined based on the indication information. The modulation mode can be a modulation level (also known as a modulation mode or modulation order, etc.). The frequency offset estimation result can be represented by a frequency offset estimation angle value or a frequency offset calibration value, etc. The structure of the neural network includes but is not limited to a deep neural network, a convolutional neural network, a residual neural network, or a residual neural network with an attention mechanism. The received data is input into the neural network-based frequency offset estimation model corresponding to the modulation mode, and the residual frequency offset after equalization is estimated and corrected using the data-based neural network frequency offset estimation function module to obtain the frequency offset estimation result. This can achieve the purpose of high-speed transmission with fewer pilot configurations, greatly improving the transmission capacity of traditional communication systems in high-speed scenarios. The data-based frequency offset estimation method can be applied to new pilot-free transmission communication systems, improve the transmission quality of pilot-free transmission communication systems, and can also be flexibly adapted to different communication systems.

[0021] In one embodiment, the modulation mode includes a modulation level.

[0022] In this embodiment, the modulation mode includes the modulation level, and the frequency deviation estimation module needs to select different neural networks for training according to the modulation level. Therefore, the two-dimensional data mapping matrix of different modulation levels is used as the input of the neural network-based frequency deviation estimation model matching different modulation levels. According to the different modulation levels output by the modulation detection module, the corresponding neural network-based frequency deviation estimation model is selected for online detection. The modulation mode is the estimated frequency deviation value, which can be expressed in different forms such as frequency deviation estimation angle value or frequency deviation calibration value. This estimated value can also be used for subsequent frequency deviation compensation processing.

[0023] In one embodiment, determining a modulation mode corresponding to received data includes:

[0024] The received data is input into a modulation detection model based on a neural network to obtain a modulation mode.

[0025] In one embodiment, before inputting the received data into the neural network-based modulation detection model, the method further includes: normalizing the received data;

[0026] The normalized received data is converted into a two-dimensional data mapping matrix, and the two-dimensional data mapping matrix is ​​used as an input of the modulation detection model and the frequency offset estimation model.

[0027] In this embodiment, the received data needs to be preprocessed before being input into the neural network-based modulation detection model. The preprocessing primarily involves normalizing the received data. The normalized data is then fed into the input data two-dimensional format conversion module for data format conversion. The data format conversion module primarily converts the received complex data into an N*N dimensional real number format, resulting in a converted two-dimensional data mapping matrix. This two-dimensional data mapping matrix serves as input to the modulation detection model and the frequency offset estimation model.

[0028] In one embodiment, converting the normalized received data into a two-dimensional data mapping matrix includes:

[0029] Generate a two-dimensional data initial matrix, where each element in the two-dimensional data initial matrix is ​​a first value;

[0030] Mapping the real and imaginary parts of the normalized complex number received data into coordinates of a two-dimensional matrix;

[0031] The values ​​corresponding to the coordinates in the two-dimensional data initial matrix are assigned as second values ​​to obtain the two-dimensional data mapping matrix.

[0032] In this embodiment, the two-dimensional data mapping matrix Form is initialized to an N*N matrix, and each element is initialized to a, where a is the first value, as shown in formula (1): Form i,j =a,i=[-1:2 / N:1],j=[-1:2 / N:1] (1)

[0033] The complex received data Y Data,norm The real and imaginary parts of are mapped to the coordinates (x, y) of the two-dimensional matrix according to formula (2):

[0034] Then assign the value of the corresponding coordinate in the mapping table to b, where b is the second value, as shown in formula (3): x,y =b (3)

[0035] The obtained N*N two-dimensional data mapping matrix Form is the input of the neural network.

[0036] In one embodiment, the horizontal coordinate of the two-dimensional matrix is ​​such that abs(A(i)-Re(Y Data,norm ))The smallest i;

[0037] The vertical coordinate of the two-dimensional matrix is ​​such that abs(A(j)-Im(Y Data,norm ))The smallest j;

[0038] in, Y data To receive data, Y Data,norm is the normalized received data in complex form, Re(Y Data,norm ) is the real part of the normalized received data in complex form, Im(Y Data,norm ) is the imaginary part of the normalized received data in complex form, and abs() represents the absolute value.

[0039] In one embodiment, for received data of each modulation mode, a corresponding frequency offset estimation model is trained based on sample data.

[0040] In this embodiment, received data of different modulation modes are used, and frequency offset estimation models based on neural networks corresponding to different modulation modes are trained based on sample data, thereby improving the accuracy of frequency offset estimation for various modulation modes.

[0041] In one embodiment, for each modulation mode of received data, sample data corresponding to the modulation mode is obtained according to the following steps:

[0042] generating a modulation signal corresponding to the modulation mode;

[0043] Passing the modulated signal through a channel and superimposing a frequency deviation value to obtain sample received data corresponding to the modulation mode;

[0044] Converting the sample received data into a sample two-dimensional data mapping matrix;

[0045] The superimposed frequency offset values ​​are recorded as training labels.

[0046] In this embodiment, a modulated signal corresponding to the modulation scheme is generated, passed through a channel, and superimposed with the effects of different frequency offset values ​​to obtain sample received data corresponding to the modulation scheme. This sample received data is then converted into a two-dimensional sample data mapping matrix to serve as training samples. The superimposed different frequency offset values ​​are recorded as training labels. Alternatively, index values ​​corresponding to different modulation levels corresponding to the samples can be used as training labels.

[0047] In one embodiment, training a corresponding frequency offset estimation model based on sample data includes:

[0048] Inputting the sample two-dimensional data mapping matrix into a frequency offset estimation model based on a neural network, and calculating a loss function according to a frequency offset estimation value output by the frequency offset estimation model and a training label;

[0049] The frequency offset estimation model is optimized using an Adaptive Moment Estimation (Adam) optimizer until the loss function meets the requirements.

[0050] In this embodiment, multiple sets of training samples (i.e., sample two-dimensional data mapping matrices) and labels corresponding to the modulation scheme are generated under different fading channels and frequency offset values. These sets are then input into a neural network-based frequency offset estimation model for training, resulting in frequency offset estimates (arranged in the same order as the label data). A loss function is then calculated along with the training labels. Based on the loss function, the neural network-based frequency offset estimation model is optimized and trained using the adaptive moment estimation Adam optimizer. Network training is completed when the loss function converges to a certain level and meets the requirements.

[0051] In one embodiment, the loss function includes at least one of the following:

[0052] Mean squared error loss function, Euclidean distance, cosine distance.

[0053] The data transmission method is exemplified below through specific embodiments.

[0054] Example 1

[0055] Assume that in an OFDM multiple input multiple output (MIMO) system, the pilot signal received by the receiver is Y Data In this embodiment, the transmitter preprocessing of the pilotless communication system is combined with the receiving side frequency offset estimation method and device based on AI under the condition that the data adopts Qm=4,16 quadrature amplitude modulation (QAM). How does it obtain the received data signal Y? Data Frequency offset estimation The specific steps are as follows:

[0056] Step 1: Determine the overall architecture of the frequency offset estimation neural network model. FIG2 is a schematic diagram of an implementation of a frequency offset estimation method based on a neural network provided by an embodiment. As shown in FIG2, it includes a data preprocessing module, a data two-dimensional format conversion module, and a frequency offset estimation detection module based on a neural network. Assume that in a communication system, the received signal is represented by Y, Y Data Indicates receiving data. First, the received data Y Data Preprocessing is mainly to normalize the data. Data,normThe data is fed into the two-dimensional matrix mapping module. The data format conversion module primarily converts the received complex data into an N*N dimensional real number format, where N = 51. The converted N*N two-dimensional mapping data (i.e., the two-dimensional data mapping matrix) serves as the input to the neural network. Different neural networks are selected based on different modulation methods. The output is the estimated frequency deviation value, which can be displayed in various forms, such as the frequency deviation estimation angle value or the frequency deviation calibration value. The neural network structure is a convolutional neural network.

[0057] Step 2: Construct the input of the frequency offset estimation neural network model.

[0058] 1) First, you need to receive data Y Data Perform normalization processing to obtain the normalized received data Y Data,norm .

[0059] 2) Receive the normalized data Y Data,norm The data is sent to the two-dimensional format conversion module for data format conversion processing to obtain the converted two-dimensional data mapping matrix Form.

[0060] Initialize the two-dimensional data mapping matrix Form to an N*N matrix, where N=51, and initialize each element to 0.1: Form i,j =0.1,i=[-1:2 / N:1],j=[-1:2 / N:1]

[0061] The complex received data Y Data,norm The real and imaginary parts of are mapped to the coordinates (x, y) of the two-dimensional matrix according to the formula:

[0062] Then assign the value of the corresponding coordinate in the mapping table to 1, as shown in formula (3): x,y =1

[0063] The obtained N*N two-dimensional data mapping table Form is the input of the neural network.

[0064] Step 3: Generate training samples and labels for the frequency offset estimation neural network and perform model training. Use received signals with different modulation modes to train different neural networks (i.e., a modulation detection model based on a neural network and a frequency offset estimation model based on a neural network). Generate a Qm (modulation order or modulation level) modulated signal X Data , passing through the channel and superimposing the influence of different frequency offset values Get received data Y Data , follow the second step to get the two-dimensional data mapping table Form as a training sample. The superimposed different frequency deviation values ​​are recorded as training labels Under different fading channels and different frequency offset values, multiple sets of training samples and labels corresponding to Qm modulation modes are generated and input into the neural network for training to obtain the frequency offset estimation value. (The order is the same as the label data), and then Calculate the mean loss function together. Based on the mean square error (MSE) loss function, use the Adam optimizer to train the neural network. Network training is complete when the loss function converges to a certain level.

[0065] Step 4: Implement online frequency offset estimation and detection. After the neural network training is completed, it can be deployed in the communication system for online detection. During online detection, it is also necessary to follow the processing flow from the first step to the second step to obtain the input of the frequency offset estimation neural network. According to different modulation modes, the corresponding neural network model is selected for online detection. The output of the obtained neural network (frequency offset estimation model based on the neural network) is the frequency offset estimation value. This estimate It can also be used for subsequent frequency offset compensation processing.

[0066] Example 2

[0067] Assume that in an OFDM MIMO system, the pilot signal received by the receiver is Y Data This embodiment provides a method and device for estimating frequency offset based on AI, which is based on the equalized Y Data Frequency offset estimation The specific steps are as follows:

[0068] Step 1: Determine the overall architecture of the frequency offset estimation neural network model. FIG3 is a schematic diagram of another implementation of a frequency offset estimation method based on a neural network provided by an embodiment. As shown in FIG3 , it includes a data preprocessing module, a data two-dimensional format conversion module, a modulation detection module based on a neural network, and a frequency offset estimation detection module based on a neural network. In this embodiment, the received data Y Data Represents the data after receiver equalization. First, the received data Y Data Preprocessing is mainly to normalize the data and then convert the data Y Data,normThe data is fed into the two-dimensional matrix mapping module. The data format conversion module primarily converts the received complex data into an N*N dimensional real number format, where N=101. The converted N*N two-dimensional mapping data serves as the input to the modulation detection module's neural network, and the output is the modulation mode Qm. Different neural networks are selected based on the different modulation methods. The input is also the N*N two-dimensional mapping data, and the output is the estimated frequency deviation value, which can be displayed in various forms such as the frequency deviation estimation angle value or the frequency deviation calibration value. The neural network structure is a convolutional neural network.

[0069] Step 2: Construct the input of the modulation detection subnetwork and the frequency offset estimation subnetwork.

[0070] 1) Need to receive data Y Data Perform normalization processing to obtain the normalized received data Y Data,norm .

[0071] 2) Receive the normalized data Y Data,norm The data is sent to the two-dimensional mapping module for data format conversion to obtain the converted two-dimensional data mapping matrix Form.

[0072] Initialize the two-dimensional data mapping matrix Form to an N*N matrix, where N=101, and initialize each element to 0.1: Form i,j =0.1,i=[-1:2 / N:1],j=[-1:2 / N:1]

[0073] The complex received data Y Data,norm The real and imaginary parts of are mapped to the coordinates (x, y) of the two-dimensional matrix according to the formula:

[0074] Then assign the value of the corresponding coordinate in the mapping table to 1: x,y =1

[0075] The resulting N*N two-dimensional data mapping table (Form) serves as the input to the modulation detection subnetwork and the frequency offset estimation subnetwork. However, the modulation detection subnetwork outputs the modulation level Qm, while the frequency offset estimation module selects different neural networks for training based on the modulation level Qm. Therefore, the received mapping data (Form) for different modulation levels serves as the input to the neural networks matching the modulation levels.

[0076] Step 3: Generate training samples and labels for the modulation detection and frequency offset estimation neural network, and perform model training.

[0077] Generate training samples and labels for the modulation detection neural network and perform model training. Generate and send Qm (modulation order or modulation level) modulation signal XData , passing through the channel and superimposing the influence of different frequency offset values Get received data Y Data . The received data is normalized to obtain complex received data, and then the data mapping table Form is obtained through the two-dimensional mapping module as a training sample. The different modulation levels Qm corresponding to the samples are used as training labels according to the index values ​​in Table 1. Under different Qm modulation modes, multiple groups of training samples and labels are generated and input into the neural network for training to obtain the modulation level, and then the average loss function is calculated together with the label Qm. Based on the Softmax loss function, the Adam optimizer is used to train the neural network. The network training is completed after the loss function converges to a certain extent.

[0078] Table 1 Qm index values ​​for different modulation levels

[0079] Generate training samples and labels for the frequency deviation estimation neural network and perform model training. Use received signals with different modulation methods to train different neural networks. Generate and send Qm (modulation order or modulation level) modulation signal X Data , passing through the channel and superimposing the influence of different frequency offset values Get received data Y Data , follow the second step to get the two-dimensional data mapping table Form as a training sample. The superimposed different frequency deviation values ​​are recorded as training labels Under different fading channels and different frequency offset values, multiple sets of training samples and labels corresponding to Qm modulation modes are generated and input into the neural network for training to obtain the frequency offset estimation value. (The order is the same as the label data), and then Calculate the mean loss function together. Based on the mean squared error (MSE) loss function, use the Adam optimizer to train the neural network. Network training is complete when the loss function converges to a certain level.

[0080] Step 4: Implement online frequency offset estimation and detection. After the neural network training is completed, it can be deployed in the communication system for online detection. During online detection, the received data also needs to be preprocessed according to the first step. Then, according to the two-dimensional mapping in the second step, the input of the modulation detection neural network and the input of the frequency offset estimation neural network are obtained. According to the different modulation levels Qm output by the modulation detection module, the corresponding neural network model is selected for online detection. The output of the obtained neural network is the frequency offset estimation value. This estimate It can also be used for subsequent frequency offset compensation processing of traditional receivers.

[0081] The present application also provides a frequency offset estimation device. FIG4 is a schematic diagram of the structure of a frequency offset estimation device provided by an embodiment. As shown in FIG4, the frequency offset estimation device includes:

[0082] The modulation mode determination module 210 is configured to determine a modulation mode corresponding to the received data.

[0083] The frequency offset estimation result determination module 220 is configured to input the received data into a frequency offset estimation model based on a neural network corresponding to the modulation mode to obtain a frequency offset estimation result.

[0084] The frequency offset estimation device of this embodiment inputs the received data into a neural network-based modulation detection model to obtain a modulation mode; inputs the received data into a neural network-based frequency offset estimation model corresponding to the modulation mode to obtain a frequency offset estimation result, which can be flexibly adapted to different communication systems; combined with a traditional communication system receiver, it uses a data-based neural network frequency offset estimation function module to perform residual frequency offset estimation and correction after equalization, which can achieve the purpose of high-speed transmission with fewer pilot configurations, greatly improving the transmission capacity of traditional communication systems in high-speed scenarios; the data-based frequency offset estimation method can be applied to new pilot-free transmission communication systems to improve the transmission quality of pilot-free transmission communication systems.

[0085] In one embodiment, the modulation mode includes a modulation level.

[0086] In one embodiment, the modulation mode determination module 210 is configured to input the received data into a modulation detection model based on a neural network to obtain a modulation mode.

[0087] In one embodiment, before inputting the received data into the neural network-based modulation detection model, the apparatus further includes a pre-processing module, the pre-processing module including:

[0088] A normalization submodule, configured to normalize the received data;

[0089] The conversion submodule is configured to convert the normalized received data into a two-dimensional data mapping matrix, and the two-dimensional data mapping matrix serves as an input of the modulation detection model and the frequency offset estimation model.

[0090] In one embodiment, the conversion submodule includes:

[0091] A generating unit configured to generate an initial two-dimensional data matrix, wherein each element in the initial two-dimensional data matrix is ​​a first value;

[0092] a mapping unit configured to map the real and imaginary parts of the normalized received data in complex form into coordinates of a two-dimensional matrix;

[0093] The assignment unit is configured to assign the value corresponding to the coordinate in the two-dimensional data initial matrix to a second value to obtain the two-dimensional data mapping matrix.

[0094] In one embodiment, the horizontal coordinate of the two-dimensional matrix is ​​such that abs(A(i)-Re(Y Data,norm ))The smallest i;

[0095] The vertical coordinate of the two-dimensional matrix is ​​such that abs(A(j)-Im(Y Data,norm ))The smallest j;

[0096] in, Y data To receive data, Y Data,norm is the normalized received data in complex form, Re(Y Data,norm ) is the real part of the normalized complex received data, Im(Y Data,norm ) is the imaginary part of the normalized received data in complex form, and abs() represents the absolute value.

[0097] In one embodiment, the frequency offset estimation device further includes:

[0098] The training module is configured to train a corresponding frequency offset estimation model based on sample data for received data of each modulation mode.

[0099] In one embodiment, the training module is configured as follows:

[0100] For each modulation mode of received data, the sample data corresponding to the modulation mode is obtained by following the steps below:

[0101] generating a modulation signal corresponding to the modulation mode;

[0102] Passing the modulated signal through a channel and superimposing a frequency deviation value to obtain sample received data corresponding to the modulation mode;

[0103] Converting the sample received data into a sample two-dimensional data mapping matrix;

[0104] The superimposed frequency offset values ​​are recorded as training labels.

[0105] In one embodiment, the training module is configured to:

[0106] Inputting the sample two-dimensional data mapping matrix into a frequency offset estimation model based on a neural network, and calculating a loss function according to a frequency offset estimation value output by the frequency offset estimation model and a training label;

[0107] The frequency offset estimation model is optimized using an adaptive moment estimation Adam optimizer until the loss function meets the requirements.

[0108] In one embodiment, the loss function includes at least one of the following:

[0109] Mean squared error loss function, Euclidean distance, cosine distance.

[0110] The frequency offset estimation device proposed in this embodiment and the frequency offset estimation method proposed in the above embodiment belong to the same inventive concept. For technical details not fully described in this embodiment, please refer to any of the above embodiments. This embodiment has the same beneficial effects as the frequency offset estimation method.

[0111] The present application also provides a communication node. Figure 5 is a schematic diagram of the hardware structure of a communication node provided in one embodiment. As shown in Figure 5, the communication node provided in the present application includes a processor 310 and a memory 320. The processor 310 in the communication node can be one or more, and Figure 5 uses one processor 310 as an example. The memory 320 is configured to store one or more programs. The one or more programs are executed by the one or more processors 310, so that the one or more processors 310 implement the frequency offset estimation method described in the embodiment of the present application.

[0112] The communication source node further includes: a communication device 330 , an input device 340 and an output device 350 .

[0113] The processor 310 , memory 320 , communication device 330 , input device 340 and output device 350 in the communication node may be connected via a bus or other means. FIG5 takes the bus connection as an example.

[0114] The input device 340 may be used to receive input digital or character information and generate key signal input related to user settings and function control of the electronic device. The output device 350 may include a display device such as a display screen.

[0115] The communication device 830 may include a receiver and a transmitter. The communication device 830 is configured to perform information transmission and reception communication according to the control of the processor 310.

[0116] Memory 320, as a computer-readable storage medium, can be configured to store software programs, computer-executable programs, and modules, such as program instructions / modules corresponding to the frequency offset estimation method described in the embodiments of the present application (e.g., the modulation mode determination module 210 and the frequency offset estimation result determination module 220 in the frequency offset estimation device). Memory 320 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; and the data storage area may store data created based on the use of the communication node. Furthermore, memory 320 may include high-speed random access memory and non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state memory device. In some instances, memory 320 may further include memory remotely located relative to processor 310, and these remote memories may be connected to the communication node via a network. Examples of the aforementioned networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0117] The present application also provides a storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the frequency offset estimation method described in any one of the embodiments of the present application. The frequency offset estimation method, applied to a communication node, comprises:

[0118] Determine the modulation mode corresponding to the received data;

[0119] The received data is input into a frequency offset estimation model based on a neural network corresponding to the modulation mode to obtain a frequency offset estimation result.

[0120] The computer storage medium of the embodiment of the present application can adopt any combination of one or more computer-readable media.Computer-readable media can be computer-readable signal media or computer-readable storage media.Computer-readable storage media can be, for example, but not limited to: electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above.More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM), flash memories, optical fibers, portable CD-ROMs, optical storage devices, magnetic storage devices, or any suitable combination of the above.Computer-readable storage media can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0121] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0122] The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wire, optical cable, radio frequency (RF), etc., or any suitable combination of the foregoing.

[0123] The computer program code for performing the operations of the present application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and also conventional procedural programming languages ​​such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect via the Internet).

[0124] The above description is merely an exemplary embodiment of the present application and is not intended to limit the scope of protection of the present application.

[0125] It will be appreciated by those skilled in the art that the term user terminal covers any suitable type of wireless user equipment, such as a mobile phone, a portable data processing device, a portable web browser or a vehicle-mounted mobile station.

[0126] In general, various embodiments of the present application may be implemented in hardware or dedicated circuits, software, logic, or any combination thereof. For example, some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, although the present application is not limited thereto.

[0127] Embodiments of the present application may be implemented by executing computer program instructions by a data processor of a mobile device, for example, in a processor entity, or by hardware, or by a combination of software and hardware. The computer program instructions may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages.

[0128] The block diagram of any logic flow in the drawings of this application may represent program steps, or may represent interconnected logic circuits, modules and functions, or may represent a combination of program steps and logic circuits, modules and functions. A computer program may be stored on a memory. The memory may be of any type suitable for the local technical environment and may be implemented using any suitable data storage technology, such as but not limited to read-only memory (ROM), random access memory (RAM), optical storage devices and systems (digital versatile discs (DVD) or compact disks (CD), etc.). Computer-readable media may include non-transitory storage media. The data processor may be of any type suitable for the local technical environment, such as but not limited to a general-purpose computer, a special-purpose computer, a microprocessor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), and a processor based on a multi-core processor architecture.

[0129] The above description of exemplary embodiments of the present application has been provided by way of exemplary and non-limiting examples. However, various modifications and adaptations of the above embodiments will be apparent to those skilled in the art, when considered in conjunction with the accompanying drawings and the appended claims, without departing from the scope of the present application. Therefore, the proper scope of the present application will be determined by reference to the appended claims.

Claims

1. A frequency offset estimation method, comprising: Determine the modulation mode corresponding to the received data; The received data is input into a frequency offset estimation model based on a neural network corresponding to the modulation mode to obtain a frequency offset estimation result.

2. The method according to claim 1, wherein: The modulation mode includes a modulation level.

3. The method according to claim 1, wherein: Determine the modulation mode corresponding to the received data, including: The received data is input into a modulation detection model based on a neural network to obtain a modulation mode.

4. The method according to claim 3, before inputting the received data into the modulation detection model based on the neural network, further comprising: Normalize the received data; The normalized received data is converted into a two-dimensional data mapping matrix, and the two-dimensional data mapping matrix is ​​used as an input of the modulation detection model and the frequency offset estimation model.

5. The method according to claim 4, wherein: The normalized received data is converted into a two-dimensional data mapping matrix, including: Generate a two-dimensional data initial matrix, each element in the two-dimensional data initial matrix is ​​a first value; Mapping the real and imaginary parts of the normalized received data in complex form into coordinates of a two-dimensional matrix; The values ​​corresponding to the coordinates in the two-dimensional data initial matrix are assigned as second values ​​to obtain the two-dimensional data mapping matrix.

6. The method according to claim 5, wherein: The horizontal coordinates of the two-dimensional matrix are such that abs(A(i)-Re(Y Data,norm ))The smallest i; The ordinate of the two-dimensional matrix is ​​such that abs(A(j)-Im(Y Data,norm ))The smallest j; in, Y data To receive data, Y Data,norm is the normalized received data in complex form, Re(Y Data,norm ) is the normalized reception in complex form The real part of the data, Im(Y Data,norm ) is the imaginary part of the normalized received data in complex form, and abs() represents the absolute value.

7. The method according to claim 1, further comprising: For the received data of each modulation mode, the corresponding frequency offset estimation model is trained based on the sample data.

8. The method according to claim 7, further comprising: For each modulation mode of received data, the sample data corresponding to the modulation mode is obtained according to the following steps: Generating a modulation signal corresponding to the modulation mode; Passing the modulated signal through a channel and superimposing a frequency deviation value to obtain sample received data corresponding to the modulation mode; Converting the sample received data into a sample two-dimensional data mapping matrix; The superimposed frequency offset values ​​are recorded as training labels.

9. The method according to claim 8, wherein: The corresponding frequency deviation estimation model is trained based on sample data, including: Inputting the sample two-dimensional data mapping matrix into a frequency offset estimation model based on a neural network, and calculating a loss function according to a frequency offset estimation value output by the frequency offset estimation model and a training label; The frequency offset estimation model is optimized using an adaptive moment estimation Adam optimizer until the loss function meets the requirements.

10. The method according to claim 9, wherein: The loss function includes at least one of the following: Mean squared error loss function, Euclidean distance, cosine distance.

11. A communication node, comprising: memory, and at least one processor; The memory is configured to store at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the frequency offset estimation method according to any one of claims 1 to 10.

12. A computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the frequency offset estimation method according to any one of claims 1 to 10.

Images