Communication method, electronic device, and computer program product

Abstract

Embodiments of the present disclosure provide a communication method, an electronic device, and a computer program product. The communication method comprises: a sending end acquires mapping relationships between different data streams of a communication system and a power allocation matrix obtained by fusion and dimensionality reduction; the sending end superimposes, on the basis of the mapping relationships, pilot data and original data to obtain aliased data; and the sending end acquires, on the basis of the aliased data, a signal to be transmitted, and sends said signal and said power allocation matrix to a receiving end, so that the receiving end acquires bit information of the original data.

Description

Communication methods, electronic devices and computer program products

[0001] Cross-references to related applications

[0002] This disclosure is based on and claims priority to Chinese Patent Application No. 2024110590314, filed on August 2, 2024, the entire contents of which are incorporated herein by reference. Technical Field

[0003] This disclosure relates to the field of communications, and more specifically, to a communication method, electronic device, and computer program product. Background Technology

[0004] Channel estimation is an indispensable part of traditional wireless mobile communication systems. It obtains Channel State Information (CSI) between the base station and users, and the accuracy of CSI estimation often significantly impacts the system's performance. Currently, most communication systems employ pilot-based channel estimation, but this method consumes time-frequency resources for transmitting pilot data, leading to reduced system efficiency. The superimposed pilot method is a semi-blind channel estimation algorithm. Its basic idea is to directly superimpose pilot data onto the data signal at an appropriate power before data transmission. The base station then estimates the channel by treating the received data signal as noise. Therefore, in the superimposed pilot method, pilots and data are transmitted simultaneously, eliminating the need to allocate time slots specifically for pilot data. This allows for high estimation accuracy and low computational complexity without bandwidth loss.

[0005] However, in channel estimation based on superimposed pilots, there is severe pilot and data interference during channel estimation and data decoding, which degrades the performance of channel estimation and data decoding. Summary of the Invention

[0006] This disclosure provides a communication method, electronic device, and computer program product to at least address the problems of degraded channel estimation and data decoding performance in the related art.

[0007] According to one embodiment of this disclosure, a communication method is provided, comprising: a transmitting end acquiring the mapping relationship between different data streams of a communication system and a fused and dimension-reduced power allocation matrix; the transmitting end superimposing pilot data and original data based on the mapping relationship to obtain aliased data; the transmitting end acquiring a signal to be transmitted based on the aliased data, and sending the signal to be transmitted and the fused and dimension-reduced power allocation matrix to a receiving end, so that the receiving end acquires the bit information of the original data.

[0008] According to another embodiment of this disclosure, a communication method is provided, comprising: a receiving end receiving a signal to be transmitted and a fused and dimensionality-reduced power allocation matrix from a transmitting end, and decoding the signal to be transmitted to obtain aliased data; the receiving end integrating the aliased data to obtain an integrated signal, and obtaining bit information of the original data based on the integrated signal.

[0009] According to another embodiment of this disclosure, an electronic device is also provided, including a memory and a processor, wherein a computer program is stored in the memory and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0010] According to yet another embodiment of this disclosure, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps in any of the above method embodiments. Attached Figure Description

[0011] Figure 1 is a hardware structure block diagram of a computer terminal for a communication method according to an embodiment of the present disclosure;

[0012] Figure 2 is a flowchart of a communication method according to an embodiment of this disclosure;

[0013] Figure 3 is another flowchart of the communication method according to an embodiment of the present disclosure;

[0014] Figure 4 is a schematic diagram of the structure of a wireless communication system according to an embodiment of the present disclosure;

[0015] Figure 5 is a schematic diagram of the power allocation matrix generation module according to an embodiment of this disclosure;

[0016] Figure 6 is an end-to-end neural network structure diagram of an embodiment of this disclosure;

[0017] Figure 7 is a flowchart illustrating the communication method according to an embodiment of this disclosure. Detailed Implementation

[0018] The embodiments of this disclosure will be described in detail below with reference to the accompanying drawings and examples.

[0019] It should be noted that the terms "first," "second," etc., in the specification, claims, and drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0020] The methods and embodiments provided in this disclosure can be executed in a mobile terminal, a computer terminal, or a similar computing device. Taking a computer terminal as an example, FIG1 is a hardware structure block diagram of a computer terminal for the communication method of this disclosure. As shown in FIG1, the computer terminal may include one or more (only one is shown in FIG1) processors 102 (processors 102 may include, but are not limited to, processing devices such as microprocessors MCUs or programmable logic devices FPGAs) and a memory 104 for storing data. The computer terminal may also include a transmission device 106 for communication functions and an input / output device 108. It will be understood by those skilled in the art that the structure shown in FIG1 is only illustrative and does not limit the structure of the computer terminal. For example, the computer terminal may also include more or fewer components than shown in FIG1, or have a different configuration than shown in FIG1.

[0021] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the communication method in this embodiment. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, thus implementing the above-described method. The memory 104 may include high-speed random access memory and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0022] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider for the computer terminal. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0023] This disclosure provides a communication method. Figure 2 is a flowchart of the communication method according to this disclosure. As shown in Figure 2, the process includes the following steps:

[0024] Step S202: The transmitting end obtains the mapping relationship between the different data streams of the communication system and the power allocation matrix after fusion and dimensionality reduction.

[0025] In an exemplary embodiment, before the transmitting end obtains the mapping relationship between different data streams of the communication system and the fused and dimensionality-reduced power allocation matrix, the method further includes: the transmitting end generating a power allocation matrix through a first neural network and performing fused and dimensionality-reduced power allocation matrix.

[0026] In this embodiment of the disclosure, the transmitting end first generates a power allocation matrix through a first neural network, then performs fusion and dimensionality reduction processing on the power allocation matrix, and then maps the fused and dimensionality-reduced power allocation matrix to the system's data stream. The mapping rule is a preset mapping rule, and the mapping method can be determined according to the actual situation. The power allocation matrix is ​​obtained through mapping, and each power allocation matrix corresponds to a data stream.

[0027] In an exemplary embodiment, the transmitting end obtains the mapping relationship between different data streams of the communication system and the fused and dimensionality-reduced power allocation matrix, including: the transmitting end maps the fused and dimensionality-reduced power allocation matrix and different data streams according to a preset mapping rule to obtain a power allocation mapping matrix corresponding to different data streams; the transmitting end obtains a first mapping relationship between different data streams and the fused and dimensionality-reduced power allocation matrix, and a second mapping relationship between different data streams and the power allocation mapping matrix according to the preset mapping rule, wherein the mapping relationship includes the first mapping relationship and the second mapping relationship.

[0028] In step S204, the transmitting end superimposes the pilot data and the original data based on the mapping relationship to obtain aliased data.

[0029] In this embodiment of the disclosure, the transmitting end superimposes the pilot data and the original data corresponding to the data stream according to the fused and dimensionality-reduced power allocation matrix corresponding to different data streams to obtain aliased data.

[0030] In an exemplary embodiment, the transmitting end superimposes pilot data and original data based on a mapping relationship, including: the transmitting end superimposes pilot data and original data on the same time-frequency resources based on the mapping relationship.

[0031] In this embodiment of the disclosure, when the power allocation mapping matrix is ​​not equal to 0, pilot data and original data are superimposed on the same time-frequency resources. When the power allocation mapping matrix is ​​equal to 0, the aliased data contains only the original data.

[0032] In step S206, the transmitting end obtains the signal to be transmitted based on the aliased data, and sends the signal to be transmitted and the fused and reduced power allocation matrix to the receiving end so that the receiving end can obtain the bit information of the original data.

[0033] In one exemplary embodiment, before the transmitting end sends the signal to be transmitted and the fused and dimensionality-reduced power allocation matrix to the receiving end, the method further includes: the transmitting end sending a first signaling to the receiving end, the first signaling being used to instruct the receiving end to receive the fused and dimensionality-reduced power allocation matrix.

[0034] In this embodiment of the disclosure, the transmitting end first needs to send a first signaling to the receiving end, and the receiving end will receive the power allocation matrix only after receiving the first signaling.

[0035] This disclosure provides a communication method. Figure 3 is another flowchart of the communication method according to this disclosure. As shown in Figure 3, the process includes the following steps:

[0036] In step S302, the receiving end receives the signal to be transmitted and the fused and dimensionality-reduced power allocation matrix from the transmitting end, and decodes the signal to be transmitted to obtain aliased data.

[0037] In one exemplary embodiment, the power allocation matrix is ​​generated by the transmitter via a first neural network.

[0038] In step S304, the receiving end integrates the aliased data and obtains the integrated signal to obtain the bit information of the original data based on the integrated signal.

[0039] In one exemplary embodiment, after the receiving end obtains the bit information of the original data based on the integrated signal, the method further includes: the receiving end verifies the bit information of the original data based on the log-likelihood ratio (LLR).

[0040] In one exemplary embodiment, the receiving end obtains bit information of the original data based on the integrated signal, including: the receiving end analyzes and processes the integrated signal based on a second neural network to obtain bit information of the original data.

[0041] In an exemplary embodiment, the first neural network and the second neural network are obtained based on joint training. The first neural network is trained with uplink channel data as input, and the second neural network is trained with pilot data, aliasing data and power allocation mapping matrix as input. The power allocation mapping matrix is ​​a matrix obtained by mapping the fused and dimensionality-reduced power allocation matrix and different data streams of the communication system according to a preset mapping rule.

[0042] In this embodiment, the receiver uses a deep neural network to decode the received aliased data. In this embodiment, the transmitter's first neural network and the receiver's second neural network are jointly trained using an end-to-end architecture, achieving the same block error rate as traditional communication systems. Furthermore, the pilot data does not consume additional time-frequency resources, thus further improving data transmission efficiency.

[0043] In this embodiment of the disclosure, the AI ​​power allocation matrix generation network at the transmitting end and the AI ​​receiver network at the receiving end constitute an end-to-end neural network architecture. This neural network architecture aims to jointly optimize and train the power allocation matrix generation module and the receiver with the goal of minimizing the error between the decoding result output by the AI ​​receiver and the actual transmitted data, which can further improve the performance of the neural network.

[0044] This disclosure provides a communication method in which a transmitting end obtains the mapping relationship between different data streams of a communication system and a fused, dimension-reduced power allocation matrix; based on the mapping relationship, the transmitting end superimposes pilot data and original data to obtain aliased data; the transmitting end obtains the signal to be transmitted based on the aliased data, and sends the signal to be transmitted and the fused, dimension-reduced power allocation matrix to a receiving end, so that the receiving end can obtain the bit information of the original data. This solves the problem of degraded channel estimation and data decoding performance in related technologies, achieving the effect of improving data decoding performance and increasing system spectral efficiency.

[0045] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this disclosure.

[0046] This embodiment also provides a communication device for implementing the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0047] The communication device provided in this embodiment can be located at a transmitting end and includes: a mapping module, an overlay module, and a transmitting module. The mapping module is configured to acquire the mapping relationship between different data streams of the communication system and the fused and dimension-reduced power allocation matrix. The overlay module is configured to overlay pilot data and original data based on the mapping relationship to obtain aliased data. The transmitting module is configured to acquire the signal to be transmitted based on the aliased data and send the signal to be transmitted and the fused and dimension-reduced power allocation matrix to the receiving end, so that the receiving end can acquire the bit information of the original data.

[0048] The communication device provided in this embodiment can be located at a receiving end, including a receiving module and an integration module. The receiving module is configured to receive the signal to be transmitted and the fused and dimensionality-reduced power allocation matrix from the transmitting end, and to decode the signal to be transmitted to obtain aliased data. The integration module is configured to integrate the aliased data to obtain an integrated signal, and to obtain the bit information of the original data based on the integrated signal.

[0049] In this embodiment of the disclosure, the communication device may further include different modules, and the naming and functional division of the modules may be selected in different ways according to the actual situation, without specific limitations.

[0050] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0051] Embodiments of this disclosure also provide a computer-readable storage medium storing a computer program configured to perform the steps in any of the above method embodiments when executed.

[0052] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0053] Embodiments of this disclosure also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0054] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0055] This disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in any of the above method embodiments.

[0056] In one exemplary embodiment, the computer program product described above includes a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the methods described in various embodiments of this disclosure.

[0057] Specific examples in this embodiment can be found in the examples described in the above embodiments and exemplary implementations, and will not be repeated here.

[0058] It is obvious to those skilled in the art that the modules or steps of this disclosure described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. They can be implemented using computer-executable program code, and thus can be stored in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this disclosure is not limited to any particular combination of hardware and software.

[0059] To enable those skilled in the art to better understand the technical solutions disclosed herein, the following description is provided in conjunction with different embodiments.

[0060] Example 1

[0061] This disclosure provides a wireless communication system to implement the communication method described in the above embodiments. Figure 4 is a schematic diagram of the wireless communication system according to this disclosure. As shown in Figure 4, the wireless communication system mainly includes a transmitter, a receiver (AI receiver), and a power allocation matrix generation module. In this disclosure, the transmitter and receiver correspond to the transmitting end and receiving end in the above embodiments, respectively.

[0062] In this embodiment, the power allocation matrix generation module can be located inside the transmitter. The transmitter uses the power allocation matrix to superimpose the raw data and pilot data on the same time-frequency resources to obtain aliased data. The power allocation matrix generation module uses a deep neural network to generate the power allocation matrix to further improve transmitter performance. In this embodiment, the power allocation matrix can be dynamically adjusted according to changes in channel and usage requirements, making the overall structure more flexible.

[0063] In this embodiment, the receiver uses a deep neural network to decode the received aliased data. In this embodiment, the transmitter and receiver deep neural networks are jointly trained using an end-to-end architecture, achieving the same block error rate as traditional communication systems. Furthermore, the pilot data does not consume additional time-frequency resources, thus further improving data transmission efficiency.

[0064] Figure 5 is a schematic diagram of the power allocation matrix generation module according to an embodiment of the present disclosure. As shown in Figure 5, the power allocation matrix generation module includes three parts: AI power allocation matrix generation network, power allocation matrix fusion and dimensionality reduction, and power allocation matrix mapping.

[0065] In related technologies, the power allocation matrix in the channel estimation algorithm with superimposed pilots is usually calculated according to a certain criterion, such as maximizing the signal-to-interference plus-noise ratio (SINR) of the received signal. However, the power allocation matrix calculated according to the relevant criteria is not necessarily the optimal solution.

[0066] Therefore, in this embodiment of the disclosure, a deep neural network is used at the transmitting end to generate a power allocation matrix, and a deep neural network is used at the receiving end to decode the received signal. The deep neural networks at the transmitting and receiving ends constitute an end-to-end deep neural network architecture. In this embodiment of the disclosure, the deep neural network at the transmitting end is a first neural network, and the deep neural network at the receiving end is a second neural network.

[0067] Figure 6 is an end-to-end neural network architecture diagram of an embodiment of this disclosure. As shown in Figure 6, the mixed data is aliased data. The AI ​​power allocation matrix generation network at the transmitting end and the AI ​​receiver network at the receiving end constitute an end-to-end neural network architecture. This neural network architecture aims to minimize the error between the decoding result output by the AI ​​receiver and the actual transmitted data by jointly optimizing and training the power allocation matrix generation module and the receiver, which can further improve the performance of the neural network.

[0068] The system architecture of this disclosure embodiment can transmit pilot data and raw data superimposed on the same time-frequency resources and can be correctly decoded at the receiving end. At the same time, since the pilot data does not occupy additional time-frequency resources, the efficiency of the communication system can be further improved.

[0069] The execution flow of the sending end and the receiving end will be described below. Figure 7 is a schematic flowchart of the communication method according to an embodiment of this disclosure. As shown in Figure 7, the data transmission flow of the sending end is as follows:

[0070] The AI ​​power allocation matrix generation network of the power allocation matrix generation module at the transmitting end generates power allocation matrix A:

[0071] In this embodiment of the disclosure, the total N of the communication system scheduling re Each resource element (RE) is divided into several sub-bands, and each sub-band contains M. re There are REs, of which M re Able to be N re Divisible by S, assuming the number of symbols scheduled is S, the dimension of the generated power allocation matrix A is... Where, N re M re Both S and S are positive integers. It is represented as a complex number.

[0072] In this embodiment, the power allocation matrix generation module uses external data as input, taking into account uplink channel data from the previous time step. When the input data to the neural network changes, the corresponding network output value also changes, making it suitable for scenarios where the channel environment changes rapidly.

[0073] The power allocation matrix generation module at the transmitting end performs fusion and dimensionality reduction on the power allocation matrix A:

[0074] In this embodiment of the disclosure, the power allocation matrix A, after fusion, dimensionality reduction, and mapping, finally yields matrices A1…A1 corresponding to each of the L-stream data signals in the communication system. L In this embodiment of the disclosure, matrix A1…A L This is the power allocation mapping matrix in the above embodiments.

[0075] In this embodiment of the disclosure, the fusion and dimensionality reduction of the power allocation matrix A is achieved by processing matrix A according to the actual channel environment and requirements. In this case, the processing method can be that the same power allocation matrix is ​​shared among different data streams within the same RB, and the dimension of matrix A is... Considering there are a total of L stream signals, the integrated dimension A use Should be Adjacent to matrix A The power allocation matrix on each RE is weighted and averaged.

[0076] The dimensionality reduction of the power allocation matrix A can be expressed by the following formula:

[0077] Among them, A use (m,s) represents the power allocation mapping matrix in the above embodiment. m represents the number of REs in the sub-band, and the value of md ranges from 1 to m to M. res represents the symbolic number, and the range of values ​​for s is 1 ≤ s ≤ S.

[0078] The power allocation matrix mapping in the power allocation matrix generation module at the transmitting end is performed as follows:

[0079] In this embodiment of the disclosure, the fused and dimensionality-reduced matrix A use The mapping process follows the basic principle that pilot data from different streams interleave and occupy different time-frequency resources, and matrix A is mapped accordingly. use After mapping, the power allocation matrices A1…A1 corresponding to each of the L-current signals are obtained. L , The mapping process is expressed by Formula 2 as follows:

[0080] In this embodiment of the disclosure, the above mapping rule is a preset mapping rule of the communication system.

[0081] In this embodiment of the disclosure, the power allocation matrix A does not contain time-frequency constraint information. The integrated matrix A use According to specific mapping rules, the power allocation matrix on different time-frequency resources is mapped to the time-frequency position of different flows.

[0082] In this embodiment of the disclosure, the fusion and dimensionality reduction of the power allocation matrix A can also allow different symbols of each data stream to share the same power allocation matrix, and the pilot data of different streams under the same symbol to be interleaved. At this time, the integrated dimension A use Should be At this point, it is necessary to calculate the weighted average of the power allocation matrices at the RE positions corresponding to the S symbols. The integration process can be represented by the following formula:

[0083] Then, the integrated matrix A use Mapping also follows the basic principle that pilot data from different streams are interleaved and occupy different time-frequency resources. The mapping process can be represented by the following formula:

[0084] In this embodiment, the mapping method is not limited to the two mapping methods mentioned above. Different mapping methods can be selected according to the actual situation, which will not be elaborated here.

[0085] In this embodiment of the disclosure, after the mapping is completed, the transmitting end superimposes the pilot data and the original data to obtain aliased data.

[0086] In this embodiment of the disclosure, after integration and mapping, the power allocation mapping matrices A1…A1 corresponding to each of the L-stream data signals in the MIMO system are obtained. LThese power allocation matrices are used to superimpose the pilot data and raw data for each first-order pilot onto the same time-frequency resources, as shown in the superposition formula below:

[0087] in, Represents pilot data, Represents the original data. This represents the aliased data after overlay.

[0088] In this embodiment, the first neural network at the transmitting end uses external data as its input matrix, and its output changes continuously with the input data. Therefore, the AI ​​receiver in the end-to-end architecture requires a known power allocation matrix as its input. The transmitting end needs to input the integrated power allocation matrix A. use To send to the receiving end, the sending end first needs to send the first signaling to the receiving end, and the receiving end will receive the power allocation matrix only after receiving the first signaling.

[0089] Afterwards, the transmitting end integrates the aliased data, performs Orthogonal Frequency Division Multiplexing (OFDM) symbol shaping, processes and loads a cyclic prefix (CP), and then transmits the total signal to be transmitted.

[0090] In this embodiment of the disclosure, the transmitter in a wireless communication system superimposes pilot data and raw data using a power allocation matrix. At the receiving end, an AI receiver corresponding to the transmitter receives and processes the signal transmitted by the transmitter. The receiving end employs a deep neural network to decode the received aliased data. The input design of the AI ​​receiver network includes pilot data, the power allocation mapping matrix corresponding to each data stream, and the aliased data. The specific process at the receiving end is as follows:

[0091] (1) Receive signaling and power allocation matrix:

[0092] The receiving end first receives the first signaling and determines that it needs to receive the fused and reduced power allocation matrix. After a mapping process, the fused and reduced power allocation matrix can be used to obtain the power allocation mapping matrices A1…A1 corresponding to the L-stream signals actually used by the transmitting end. L The mapping process is known during the model training phase and remains stable during the model usage phase. At this point, the receiving end knows the mapping rules and maps the fused and dimensionality-reduced power allocation matrix to obtain the power allocation mapping matrix actually used by each data stream.

[0093] (2) Receive and process mixed data:

[0094] For the received time-domain OFDM symbols, first perform CP removal and Fast Fourier Transform (FFT) operations to obtain the total frequency domain signal. Assuming the number of receiver antennas is also L, the dimension of the total frequency domain signal is N. re *S*L. The receiver processes the received signal at a sub-band granularity. First, the signal is divided into sub-bands, each containing M... re One RE. For the signal of each sub-band. First, energy normalization is performed, normalizing the maximum value of the sub-band signal energy to 1. The formula is expressed as follows:

[0095] in, Normal represents the value of the subband signal energy normalized to 1. max This indicates that the maximum energy value has been normalized.

[0096] The known pilot data is also divided into subbands. At this point, the AI ​​receiver network also needs to use the mapped power allocation mapping matrix as input, and the data integration method is signal... Real part, imaginary part, pilot data The real and imaginary parts of the signal and the power allocation matrix are superimposed on the third dimension of the three-dimensional matrix to obtain the integrated signal, which has dimension N. re* S*(5L).

[0097] (3) Signal decoding and integration:

[0098] The integrated signal is used as the input to the AI ​​receiver network, which employs a second neural network. This network uses an end-to-end architecture and an AI power allocation matrix generation module for joint training to obtain model parameters, i.e., neural network parameters. Finally, the AI ​​receiver outputs the bit values ​​of the raw data. The bit decision results of each sub-band and data stream output by the AI ​​receiver network are integrated to output the final raw data bit information. Simultaneously, the LLR value is calculated to verify the correctness of the receiver's decoding.

[0099] Example 2

[0100] In this embodiment, we assume an OFDM Single Input Single Output (SISO) communication system with a single antenna configured for both the base station and users. The system has 12 scheduled Resource Blocks (RBs), 144 Resource Exchange Units (REs), and 14 symbols. The data signal uses an MCS of 28, where MCS is the Modulation and Coding Scheme (MCS). Pilot data is modulated using Binary Phase Shift Keying (BPSK), divided into three sub-bands, each containing 48 REs. The AI ​​power allocation matrix generation module receives input data from the uplink Sounding Reference Signal (SRS) channel and transmits downlink Physical Downlink Shared Channel (PDSCH) signals. The fusion and dimensionality reduction module does not process the power allocation matrix. The transmit / receive process is as follows:

[0101] transmitter:

[0102] 1. The AI ​​power allocation matrix generation module outputs power allocation matrix A.

[0103] The AI ​​power allocation matrix generation module outputs a power allocation matrix A. At this point, the system has 144 scheduled REs, which are divided into 3 sub-bands, each containing 48 REs. The estimated channel H is used as input to the AI ​​power allocation matrix generation module to obtain the power allocation matrix A. The dimension of matrix A is [insert dimension here].

[0104] 2. Fusion, dimensionality reduction, and mapping of the power allocation matrix.

[0105] At this point, the fusion dimensionality reduction module does not process the power allocation matrix A, so A use =A, the dimension of the matrix after fusion and dimensionality reduction is Then we further analyzed the integrated matrix A use Mapping is performed; since this is single-stream data mapping and there is no inter-stream pilot overlap, A1 = A. use .

[0106] 3. Overlay the original data, pilot data, transmission signaling, and power allocation matrix.

[0107] After fusion and dimensionality reduction, a power allocation matrix A1 is obtained. The pilot data and the original data are then superimposed on the same time-frequency resources using the power allocation matrix.

[0108] in Represents pilot data, Represents the original data. This represents the aliased data after overlay.

[0109] Then, the transmitting end first sends signaling and then sends the integrated power allocation matrix A. use To the receiving end.

[0110] 4. Transmitting aliased data.

[0111] After integrating the aliased data, OFDM symbol shaping is performed, the CP is processed and loaded to obtain the total transmit signal and then it is sent.

[0112] The specific process of the receiver is as follows:

[0113] 1. Receive signaling and power allocation matrix.

[0114] The receiver first receives the first signaling and determines that it needs to receive the fused and dimension-reduced power allocation matrix A. use At this point, the integrated matrix A use After the mapping process, the power allocation matrix A1 actually used by the transmitter is obtained.

[0115] 2. Receive and process mixed data.

[0116] The received time-domain OFDM symbol is first processed by de-CP and FFT to obtain the total frequency-domain signal. At this point, the receiver has only one antenna, and the total frequency-domain signal has a dimension of 144*14*1. The receiver processes the received signal at the sub-band granularity, processing the signal of each sub-band... First, energy normalization is performed, normalizing the maximum value of the sub-band signal energy to 1.

[0117] The known pilot data is also divided into subbands. The data integration method is signal Real part, imaginary part, pilot data The real and imaginary parts of the signal and the power distribution matrix are superimposed on the third dimension of the three-dimensional matrix to obtain the integrated signal, which has a dimension of 48*14*5.

[0118] 3. Signal decoding and integration.

[0119] The integrated signal at the receiving end is used as the input to the AI ​​receiver network. The AI ​​receiver network employs a deep neural network and uses an end-to-end architecture and an AI power allocation matrix generation module for joint training to obtain model parameters. Finally, the AI ​​receiver network outputs the bit values ​​of the raw data. The bit decision results of each subband and stream output by the AI ​​receiver network are integrated to output the final raw data bit information. At the same time, the LLR value is calculated to verify the correctness of the receiver decoding.

[0120] Example 3

[0121] In this embodiment, it is assumed that in an OFDM (Multiple Input Multiple Output) communication system, the data consists of two streams, with two antennas configured for the base station and the user. The number of scheduled RBs is 24, the number of REs is 288, and the number of symbols is 14. The data signal uses MCS=16, and the pilot data uses BPSK modulation, which divides the system into four sub-bands, each containing 72 REs. The input data for the AI ​​power allocation matrix generation module is the uplink SRS channel, and the downlink PDSCH signal is transmitted. The fusion and dimensionality reduction module processes the data so that the same RB in each stream shares the same power allocation matrix. The transmit and receive process is as follows:

[0122] transmitter:

[0123] 1. The AI ​​power allocation matrix generation module outputs matrix A.

[0124] First, the AI ​​power allocation matrix generation module outputs a power allocation matrix A. At this point, the system has 288 scheduled REs, which are divided into 4 sub-bands, each containing 72 REs. The estimated channel H is then used as input to the AI ​​power allocation matrix generation module to obtain the power allocation matrix A. The dimension of matrix A is [insert dimension here].

[0125] 2. Fusion, dimensionality reduction, and mapping of the power allocation matrix.

[0126] The fusion and dimensionality reduction process ensures that the same power allocation matrix is ​​shared across all data streams within the same RB. Therefore, the integrated power allocation matrix A use It can be expressed as:

[0127] At this point, the dimension of the integrated matrix is

[0128] Then, the fused and dimensionality-reduced matrix A use Mapping yields power allocation matrices A1 and A2 for each of the two streams. The mapping process follows the basic principle that pilot data from different streams are interleaved and occupy different time-frequency resources. Therefore, the matrix mapping process can be represented as follows:

[0129] Next, the original data, pilot data, first signaling, and power allocation mapping matrix are superimposed. After fusion, dimensionality reduction, and mapping, the power allocation matrices A1 and A2 corresponding to each of the two streams are obtained. These power allocation matrices are used to superimpose the pilot data and original data of each stream on the same time-frequency resources to obtain aliased signals. Then, the transmitter first sends the signaling and then sends the integrated power allocation matrix A1. use To the receiving end.

[0130] 3. Transmitting aliased data.

[0131] After integrating the aliased data, OFDM symbol shaping is performed, the CP is processed and loaded to obtain the total transmit signal and then it is sent.

[0132] The specific process of the receiver is as follows:

[0133] 1. Receive signaling and power allocation matrix.

[0134] The receiver first receives the signaling from the transmitter and determines that it needs to receive the integrated power allocation matrix A sent by the transmitter. use At this point, the integrated matrix A use After the mapping process, the power allocation matrices A1 and A2 corresponding to the two streams actually used by the transmitter are obtained. The mapping formula is shown in Formula 6.

[0135] 2. Receive and process mixed data.

[0136] The received time-domain OFDM symbols are first de-CP and FFT operations to obtain the total frequency-domain signal. At this point, the receiver has two antennas, and the total frequency-domain signal has a dimension of 288*14*2. The receiver processes the received signal at the sub-band granularity, processing the signal of each sub-band... First, energy normalization is performed, normalizing the maximum value of the sub-band signal energy to 1.

[0137] The known pilot data is also divided into subbands. The AI ​​receiver network also needs to use the mapped power allocation matrix as input. The data integration method is signal... Real part, imaginary part, pilot data The real and imaginary parts of the signal and the power distribution matrix are superimposed on the third dimension of the three-dimensional matrix to obtain the integrated signal, which has a dimension of 72*14*10.

[0138] 3. Signal decoding and integration.

[0139] The integrated signal at the receiving end is used as the input to the AI ​​receiver network. The AI ​​receiver network employs a deep neural network and uses an end-to-end architecture and an AI power allocation matrix generation module for joint training to obtain model parameters. Finally, the AI ​​receiver network outputs the bit values ​​of the raw data. The bit decision results of each subband and stream output by the AI ​​receiver network are integrated to output the final raw data bit information. At the same time, the LLR value is calculated to verify the correctness of the receiver decoding.

[0140] Example 4

[0141] In this embodiment, it is assumed that in an OFDM MIMO communication system, data flows in four streams, the base station and users are configured with four antennas, the number of scheduled RBs is 24, the number of REs is 288, and the number of symbols is 14. The data signal uses MCS=17, and the pilot data uses BPSK modulation. It is divided into four sub-bands, each containing 72 REs. The input data for the AI ​​power allocation matrix generation module is the uplink SRS channel, and the downlink PDSCH signal is transmitted. The fusion and dimensionality reduction module processes the data so that different symbols in each stream share the same power allocation matrix. The transmit / receive process is as follows:

[0142] transmitter:

[0143] 1. The AI ​​power allocation matrix generation module outputs matrix A.

[0144] First, the AI ​​power allocation matrix generation module outputs power allocation matrix A. At this point, the system has 288 scheduled REs, which are divided into 4 sub-bands, each containing 72 REs. The uplink SRS channel from the previous time step is used as input to the AI ​​power allocation matrix generation module to obtain power allocation matrix A. The dimension of matrix A is now [dimension not specified].

[0145] 2. Fusion, dimensionality reduction, and mapping of the power allocation matrix.

[0146] At this point, the fusion and dimensionality reduction process ensures that different symbols in each data stream share the same power allocation matrix. Therefore, the integrated power allocation matrix A use It can be expressed as:

[0147] At this point, the dimension of the integrated matrix is...

[0148] Subsequently, the fused and dimensionality-reduced matrix A was further processed. use Mapping yields power allocation matrices A1, A2, A3, and A4 for each of the four streams. The mapping process follows the basic principle that pilot data from different streams are interleaved, occupying different time-frequency resources. The matrix mapping process can then be represented as:

[0149] Next, the original data, pilot data, transmitted signaling, and power allocation matrices are superimposed. After fusion, dimensionality reduction, and mapping, the power allocation matrices A1, A2, A3, and A4 corresponding to each of the four streams are obtained. These power allocation matrices are then used to superimpose the pilot data and original data of each stream on the same time-frequency resources. The transmitting end then first sends the signaling and then the integrated power allocation matrix A. use To the receiving end.

[0150] 3. Transmitting aliased data.

[0151] After integrating the aliased data, OFDM symbol shaping is performed, the CP is processed and loaded to obtain the total transmit signal and then it is sent.

[0152] The specific process of the receiver is as follows:

[0153] 1. Receive signaling and power allocation matrix.

[0154] The receiver first receives the signaling from the transmitter and determines that it needs to receive the integrated power allocation matrix A sent by the transmitter. use At this point, the integrated matrix A use After the mapping process, the power allocation matrices A1, A2, A3 and A4 corresponding to the four streams actually used by the transmitter are obtained.

[0155] 2. Receive and process mixed data.

[0156] The received time-domain OFDM symbols are first de-CP and FFT operations to obtain the total frequency-domain signal. At this point, the receiver has 4 antennas, and the total frequency-domain signal has a dimension of 288*14*4. The receiver processes the received signal at the sub-band granularity, processing the signal of each sub-band... First, energy normalization is performed, normalizing the maximum value of the sub-band signal energy to 1.

[0157] The known pilot data is also divided into subbands. The AI ​​receiver also needs to take the mapped power allocation matrix as input. The data integration method is signal... Real part, imaginary part, pilot data The real and imaginary parts of the signal and the power distribution matrix are superimposed on the third dimension of the three-dimensional matrix to obtain the integrated signal, which has a dimension of 72*14*20.

[0158] 3. Signal decoding and integration.

[0159] The integrated signal at the receiving end is used as the input to the AI ​​receiver network. The AI ​​receiver network employs a deep neural network and uses an end-to-end architecture and an AI power allocation matrix generation module for joint training to obtain model parameters. Finally, the AI ​​receiver network outputs the bit values ​​of the raw data. The bit decision results of each subband and stream output by the AI ​​receiver network are integrated to output the final raw data bit information. At the same time, the LLR value is calculated to verify the correctness of the receiver decoding.

[0160] In summary, the communication method provided in this disclosure superimposes and transmits data signals and pilot data on the same time-frequency resources. Deep neural networks are used at both the transmitting and receiving ends to generate the power allocation matrix and decode the received signal, respectively. An end-to-end architecture is designed for joint training of the deep neural networks. This architecture achieves a block error rate similar to traditional communication systems while avoiding the degradation of system spectral efficiency caused by independent pilot transmission, thus improving the throughput of the communication system.

[0161] The communication method provided in this embodiment achieves matrix dimensionality reduction by customizing the power allocation matrix generated by the deep neural network according to the channel environment and actual needs. The fusion dimensionality reduction can compress the power allocation matrix with minimal loss of system block error rate, thus achieving the effect of transmitting the power allocation matrix with less time-frequency resources.

[0162] The communication method provided in this disclosure uses a matrix mapping process based on the principle that inter-flow pilots interleave and occupy different time-frequency resources to map the fused and dimensionality-reduced power allocation matrix into an independent inter-flow power allocation matrix. The matrix mapping module can effectively reduce inter-flow pilot interference in the communication system, and simultaneously recover the inter-flow power allocation matrix losslessly at the receiving end.

[0163] The above description is merely a preferred embodiment of this disclosure and is not intended to limit this disclosure. Various modifications and variations can be made to this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A communication method, comprising: obtaining, by a sending end, a mapping relationship between different data streams of a communication system and a power allocation matrix after fusion and dimension reduction; superimposing, by the sending end, pilot data and original data based on the mapping relationship to obtain mixed data; obtaining, by the sending end, a to-be-transmitted signal based on the mixed data, and sending the to-be-transmitted signal and the power allocation matrix after fusion and dimension reduction to a receiving end, so that the receiving end obtains bit information of the original data.

2. The method of claim 1, wherein, Before the sending end obtains the mapping relationship between the different data streams of the communication system and the power allocation matrix after fusion and dimension reduction, the method further comprises: generating, by the sending end, the power allocation matrix through a first neural network, and performing fusion and dimension reduction on the power allocation matrix.

3. The method of claim 1, wherein, The sending end obtains the mapping relationship between the different data streams of the communication system and the power allocation matrix after fusion and dimension reduction, comprising: mapping, by the sending end, the power allocation matrix after fusion and dimension reduction and the different data streams according to a preset mapping rule to obtain a power allocation mapping matrix corresponding to the different data streams; obtaining, by the sending end, a first mapping relationship between the different data streams and the power allocation matrix after fusion and dimension reduction, and a second mapping relationship between the different data streams and the power allocation mapping matrix according to the preset mapping rule, wherein the mapping relationship comprises the first mapping relationship and the second mapping relationship.

4. The method of claim 1, wherein, The sending end superimposes the pilot data and the original data based on the mapping relationship, comprising: The sending end superimposes the pilot data and the original data on the same time-frequency resource based on the mapping relationship.

5. The method of claim 1, wherein, Before the sending end sends the to-be-transmitted signal and the power allocation matrix after fusion and dimension reduction to the receiving end, the method further comprises: sending, by the sending end, a first signaling to the receiving end, the first signaling being used to instruct the receiving end to receive the power allocation matrix after fusion and dimension reduction.

6. A communication method, comprising: receiving, by a receiving end, a to-be-transmitted signal and a power allocation matrix after fusion and dimension reduction from a sending end, and decoding and processing the to-be-transmitted signal to obtain mixed data; integrating, by the receiving end, the mixed data to obtain an integrated signal, so as to obtain bit information of original data according to the integrated signal.

7. The method of claim 6, wherein, After the receiving end obtains the bit information of the original data according to the integrated signal, the method further comprises: verifying, by the receiving end, the bit information of the original data based on a log-likelihood ratio (LLR).

8. The method of claim 6, wherein, The receiving end obtains the bit information of the original data according to the integrated signal, comprising: analyzing and processing, by the receiving end, the integrated signal based on a second neural network to obtain the bit information of the original data.

9. The method of claim 8, wherein, The power allocation matrix is generated by the sending end through a first neural network.

10. The method of claim 9, wherein, The first neural network and the second neural network are obtained based on joint training, wherein the first neural network is trained as input of uplink channel data, the second neural network is trained as input of pilot data, the aliasing data and a power allocation mapping matrix, and the power allocation mapping matrix is a matrix obtained by mapping different data streams of a communication system and the power allocation matrix after fusion and dimension reduction according to a preset mapping rule. 11.An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 10 when executing the computer program. 12.A computer program product comprising a computer program, wherein the computer program implements the method of any one of claims 1 to 10 when executed by a processor.

Images