Signal processing method, communication node, medium, and program product

Abstract

The present application discloses a signal processing method, a communication node, a medium, and a program product. The method comprises: acquiring a signal to be processed; and on the basis of physical layer parameter configuration information of said signal and an adaptive adjustment mechanism, calling a basic model corresponding to said signal to determine a signal processing result.

Description

Signal processing methods, communication nodes, media and software products Technical Field

[0001] This application relates to the field of communication technology, and in particular to signal processing methods, communication nodes, media, and program products. Background Technology

[0002] With the development and maturation of Artificial Intelligence (AI) technology, more and more fields are recognizing its value and gradually applying it to solve problems that are difficult to address using traditional methods. Integrating AI with Radio Access Networks (RANs) can optimize network resource allocation, improve spectrum efficiency, reduce operating costs, and enhance user experience through intelligent means, bringing significant business value to enterprises.

[0003] However, due to the fixed input and output dimensions of AI models, there are limitations in the adaptability and flexibility of physical layer parameter configuration when they are integrated with RAN. This means that whenever the parameters change, the AI ​​model often needs to be retrained, which not only greatly increases the training workload but also poses a severe challenge to hardware storage. Summary of the Invention

[0004] This application provides a signal processing method, communication node, medium, and program product. Through the constructed adaptive adjustment mechanism, the basic model can be invoked under various physical layer parameter configurations, ensuring the generalization ability and reliability of the AI ​​model in the signal processing process.

[0005] This application provides a signal processing method, including: acquiring a signal to be processed; and, based on the physical layer parameter configuration information and adaptive adjustment mechanism of the signal to be processed, calling the basic model corresponding to the signal to be processed to determine the signal processing result.

[0006] This application provides a communication node, including: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for implementing communication between the processor and the memory. When the program is executed by the processor, it implements the signal processing method as described in any of the embodiments of this application.

[0007] This application provides a storage medium for computer-readable storage, which stores one or more programs that can be executed by one or more processors to implement the signal processing method of any of the embodiments of this application.

[0008] This application provides a computer program product, including a computer program that, when executed by a processor, implements any of the signal processing methods described in this application.

[0009] The signal processing method, communication node, medium, and program product provided in this application acquire the signal to be processed; based on the physical layer parameter configuration information and adaptive adjustment mechanism of the signal to be processed, the basic model corresponding to the signal to be processed is invoked to determine the signal processing result. By adopting the above technical solution, based on the model input and output requirements of AI models commonly used in RANs, adaptive adjustment mechanisms are constructed under various physical layer parameter configurations. This allows only a small number of basic models to meet the intelligent network resource configuration under various physical layer parameter configurations, ensuring the generalization ability and reliability of the AI ​​model in the signal processing process. Attached Figure Description

[0010] Figure 1 is a flowchart of a signal processing method provided in an embodiment of this application;

[0011] Figure 2 is an example diagram of a model training mechanism and inference mechanism provided in an embodiment of this application;

[0012] Figure 3 is an example diagram of the data processing flow of a first basic model provided in an embodiment of this application;

[0013] Figure 4A shows the input signals of a first basic model provided in an embodiment of this application. Structural example diagram;

[0014] Figure 4B shows the input signal of the first basic model in Figure 4A provided in the embodiment of this application. Structural example diagram;

[0015] Figure 4C shows the input signal of the first basic model in Figure 4A provided in the embodiment of this application. Structural example diagram;

[0016] Figure 5A shows the input signals of another first basic model provided in the embodiments of this application. Structural example diagram;

[0017] Figure 5B shows the input signal of the first basic model in Figure 5A provided in the embodiment of this application. Structural example diagram;

[0018] Figure 5C shows the input signal of the first basic model in Figure 5A provided in the embodiment of this application. Structural example diagram;

[0019] Figure 6 is an example diagram of the output structure of a first basic model provided in an embodiment of this application;

[0020] Figure 7 is an example diagram of the data processing flow of a second basic model provided in an embodiment of this application;

[0021] Figure 8A shows the input signal of a second basic model provided in an embodiment of this application. Structural example diagram;

[0022] Figure 8B shows the input signal of the second basic model in Figure 8A provided in the embodiment of this application. Structural example diagram;

[0023] Figure 8C shows the input signal of the second basic model in Figure 8A provided in the embodiment of this application. Structural example diagram;

[0024] Figure 9A shows the input signals of another second basic model provided in the embodiments of this application. Structural example diagram;

[0025] Figure 9B shows the input signal of the second basic model in Figure 9A provided in the embodiment of this application. Structural example diagram;

[0026] Figure 9C shows the input signal of the second basic model in Figure 9A provided in the embodiment of this application. Structural example diagram;

[0027] Figure 10 is an example diagram of the output structure of a second basic model provided in an embodiment of this application;

[0028] Figure 11 is a structural example diagram of soft information on two final effective data resource units provided in the embodiments of this application;

[0029] Figure 12 is a flowchart illustrating the segmentation mechanism in an adaptive adjustment mechanism provided in an embodiment of this application;

[0030] Figure 13 is a flowchart illustrating the sliding window mechanism in an adaptive adjustment mechanism provided in an embodiment of this application;

[0031] Figure 14 is a flowchart illustrating the random copying of data to complete subcarriers in an adaptive adjustment mechanism provided in an embodiment of this application;

[0032] Figure 15 is a flowchart illustrating a zero-padding subcarrier mechanism provided in an embodiment of this application.

[0033] Figure 16 is a flowchart illustrating an adaptive adjustment mechanism for linear extrapolation of subcarriers provided in an embodiment of this application.

[0034] Figure 17 is a flowchart illustrating the sequential copying of data to complete subcarriers in an adaptive mechanism provided in an embodiment of this application;

[0035] Figure 18 is a flowchart illustrating a zero-padding subcarrier mechanism provided in an embodiment of this application.

[0036] Figure 19 is an example diagram of the output result of extracting channel estimation value based on the actual number of scheduled symbols in an adaptive adjustment mechanism provided in an embodiment of this application;

[0037] Figure 20 is an example diagram of the output result of extracting channel estimation values ​​based on whether it coexists with SRS in an adaptive adjustment mechanism provided in an embodiment of this application;

[0038] Figure 21A shows the adaptive input signal based on the actual number of scheduled symbols in an adaptive adjustment mechanism provided in an embodiment of this application. Structural example diagram;

[0039] Figure 21B shows the adaptive input signal based on the actual number of scheduled symbols in the adaptive adjustment mechanism of Figure 21A provided in the embodiment of this application. Structural example diagram;

[0040] Figure 21C shows the adaptive input signal based on the actual number of scheduled symbols in the adaptive adjustment mechanism of Figure 21A provided in the embodiment of this application. Structural example diagram;

[0041] Figure 22 is an example diagram of the output result of extracting LLR based on the actual number of scheduled symbols in an adaptive adjustment mechanism provided in an embodiment of this application;

[0042] Figure 23A shows the adaptive input signal based on whether it coexists with SRS in an adaptive adjustment mechanism provided in an embodiment of this application. Structural example diagram;

[0043] Figure 23B shows the adaptive input signal based on whether it coexists with SRS in the adaptive adjustment mechanism of Figure 23A provided in the embodiment of this application. Structural example diagram;

[0044] Figure 23C shows the adaptive input signal based on whether it coexists with SRS in the adaptive adjustment mechanism of Figure 23A provided in the embodiment of this application. Structural example diagram;

[0045] Figure 24 is an example diagram of the output result of extracting LLR based on whether it coexists with SRS in an adaptive adjustment mechanism provided in an embodiment of this application;

[0046] Figure 25A shows the adaptive input signal based on the configuration information of the CDM group in an adaptive adjustment mechanism provided in an embodiment of this application. Structural example diagram;

[0047] Figure 25B shows the adaptive input signal based on the configuration information of the CDM group in the adaptive adjustment mechanism of Figure 25A provided in the embodiment of this application. Structural example diagram;

[0048] Figure 25C shows the adaptive input signal based on the configuration information of the CDM group in the adaptive adjustment mechanism of Figure 25A provided in the embodiment of this application. Structural example diagram;

[0049] Figure 26 is an example diagram of the adaptive output result based on the configuration information of the CDM group in an adaptive adjustment mechanism provided in an embodiment of this application;

[0050] Figure 27A shows the input signal of a basic model in a MIMO system provided in an embodiment of this application. Structural example diagram;

[0051] Figure 27B shows the input signals of the basic model of the MIMO system in Figure 27A provided in the embodiment of this application. Structural example diagram;

[0052] Figure 28 is a schematic diagram of a signal processing device provided in an embodiment of this application;

[0053] Figure 29 is a schematic diagram of the structure of a communication node provided in an embodiment of this application. Detailed Implementation

[0054] Unless otherwise specified, the embodiments and features described in this application may be combined arbitrarily with each other.

[0055] The operations illustrated in the flowcharts in the accompanying drawings can be performed on a computer system, such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowcharts, in some cases, the operations shown or described may be performed in a different order than that presented here.

[0056] The signal processing method provided in this application can be applied to wireless communication systems, flexibly applying neural network models to signal processing with different physical layer parameter configurations. With the development and maturation of AI technology, more and more fields are recognizing the value of artificial intelligence and gradually applying it to solve problems that are difficult to address using traditional methods. Taking the integration of AI and RAN as an example, neural networks can replace traditional modules, such as channel estimation, equalization, and demodulation modules, at the physical layer of wireless communication systems. This can optimize network resource allocation through intelligent means, improve spectrum efficiency, reduce operating costs, enhance user experience, and bring significant commercial value to enterprises. This integration not only supports fifth-generation mobile communication technology (5G) networks but also lays the foundation for future research on sixth-generation mobile communication technology (6G) networks. Therefore, the exploration and application of AI and RAN integration has significant practical implications and long-term strategic value. In particular, the integration of AI and the wireless communication physical layer is an important direction for the future development of wireless communication technology.

[0057] When introducing AI to enhance the processing at the wireless communication receiver side, such as in key areas like channel estimation, equalization, demodulation, and interference suppression, AI has demonstrated its powerful potential to solve dynamic communication environments and complex interference problems. However, although AI can bring performance improvements compared to traditional algorithms, the fixed input and output dimensions of AI models contrast sharply with the flexibility of physical layer parameter configuration. Physical layer parameter configuration varies depending on the configuration of user resource blocks (RBs), modulation and coding schemes (MCSs), pilot configuration patterns, and whether service data coexists with the uplink sounding reference signal (SRS) in the same time slot. This limits its adaptability and flexibility in physical layer parameter configuration. This means that whenever the physical layer parameter configuration changes, the AI ​​model often needs to be retrained, which not only significantly increases the training workload but also poses a severe challenge to hardware storage. One approach to this problem is to introduce a neural network model training and inference mechanism. This involves designing the model's input and output to cover as many parameter configurations as possible, allowing the model to support different parameter configurations during both the training and inference phases. Alternatively, an adaptive parameter configuration mechanism can be introduced during neural network model inference. This allows for specific adjustment mechanisms to be designed under different parameter settings, making the model widely applicable to various configurations. This strategy requires the model to have sufficient generality in the time-frequency spatial domain. By combining this with a reasonable adjustment mechanism, the general model can ensure generalization ability and reliability. To address the above problems, this application proposes a signal processing method that can be implemented at a communication node. This communication node can be a terminal or base station used for signal transmission in a wireless communication system. Communication nodes are generally electronic devices with certain computing capabilities. In some possible implementations, the signal processing method can be implemented by a processor calling computer-readable instructions stored in memory.

[0058] In one exemplary embodiment, Figure 1 is a flowchart of a signal processing method provided in an embodiment of this application. This method can be applied to situations in wireless communication systems where AI model invocation processing is performed on signals with different physical layer parameter configurations. This method can be executed by a signal processing device, which can be executed by software and / or hardware and integrated on a communication node. The communication node can be a receiving communication node or a transmitting communication node, and this embodiment of the application does not limit this.

[0059] As shown in Figure 1, the signal processing method provided in this application embodiment specifically includes S101-S102.

[0060] S101, Obtain the signal to be processed.

[0061] In this embodiment, the signal to be processed can be specifically understood as a signal sent from one communication node to the desired transmitting communication node in a wireless communication system, which requires processing such as channel estimation, equalization, and demodulation.

[0062] In a specific example, when a communication node needs to transmit wireless data to its corresponding communication node, it will first acquire the original signal that needs to be processed, such as channel estimation, equalization, and demodulation, which is required for data transmission. At this time, the acquired original signal can be used as the signal to be processed.

[0063] S102. Based on the physical layer parameter configuration information and adaptive adjustment mechanism of the signal to be processed, call the basic model corresponding to the signal to be processed to determine the signal processing result.

[0064] In this embodiment, physical layer parameter configuration information can be specifically understood as various parameter settings for the physical layer in a wireless communication system. For example, physical layer parameter configuration information may include waveform, modulation method, pilot mode, digitization and frame structure, etc. The above is only an example of physical layer parameter configuration information in this application embodiment, and this application embodiment does not limit the physical layer parameter configuration information actually used.

[0065] In this embodiment, the adaptive adjustment mechanism can be specifically understood as a processing method used in a wireless communication system to optimize network performance and resource utilization. In this embodiment, it can be understood as a method of calling the AI ​​model in the physical layer of the wireless communication system based on different physical layer parameter configuration information to achieve the signal processing required.

[0066] In this embodiment, the basic model can be specifically understood as an AI model in the physical layer of a wireless communication system that can support signal input under different physical layer parameter configurations, and whose output has a universal length that can cover different time delay spread channels. It has sufficient universality in the time-frequency spatial domain and can be used for various signal processing needs.

[0067] In a specific example, based on the physical layer parameter configuration information of the signal to be processed, the adaptive adjustment mechanism determines the basic model calling method applicable to the physical layer parameter configuration information, and calls the basic model corresponding to the signal to be processed based on the basic model calling method. The basic model is then used to perform signal processing on the signal to be processed, and the signal processing result of the signal to be processed is obtained.

[0068] The signal processing method provided in this application acquires the signal to be processed; based on the physical layer parameter configuration information and adaptive adjustment mechanism of the signal to be processed, it calls the basic model corresponding to the signal to be processed to determine the signal processing result. By adopting the above technical solution, based on the model input and output requirements of AI models commonly used in RAN, adaptive adjustment mechanisms are constructed under various physical layer parameter configurations. This allows only a small number of basic models to meet the intelligent network resource configuration under various physical layer parameter configurations, ensuring the generalization ability and reliability of the AI ​​model in the signal processing process.

[0069] In one embodiment, the physical layer parameter configuration information includes at least one of the following: pilot mode; modulation scheme; actual number of scheduled symbols; coexistence status of the signal to be processed and the uplink probe reference signal; and configuration information of the code division multiplexing (CDM) group.

[0070] In this embodiment, the pilot pattern can be specifically understood as the configuration method of the reference signal used for channel estimation and synchronization. The modulation method can be specifically understood as the processing method used to convert digital signals into analog signals suitable for transmission in a wireless channel. The actual number of scheduled symbols can be specifically understood as the number of Orthogonal Frequency Division Multiplexing (OFDM) symbols actually allocated to users or data transmission during the scheduling process in a wireless communication system. The coexistence status of the signal to be processed and the SRS can be specifically understood as information used to characterize whether the signal to be processed and the SRS coexist in the same time slot. The configuration information of the CDM group can be specifically understood as configuration information used to indicate whether the resource units vacated on the pilot symbols can be used for data transmission.

[0071] In one embodiment, the input signal of the basic model exists in the form of a three-dimensional matrix; wherein the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the number of feature layers.

[0072] In this embodiment, the number of resource block symbols can be specifically understood as the number of symbols included in each resource block (RB) in the wireless communication system.

[0073] In this embodiment, the number of feature layers can be specifically understood as a combination of signal feature quantities on the scheduled resource block at the receiving side in the wireless communication system, which is related to the number of antennas, the real and imaginary parts of the signal on each antenna (including received data, local pilots and least squares channel estimation information), and the number of transmission layers of the required transmitted signal.

[0074] In some examples, the number of resource block symbols can be represented by Nsym, the number of subcarriers by NRe, and the number of feature layers by Ndim. Then, the three-dimensional matrix representation of the input signal of the basic model can be: Nre*Nsym*Ndim.

[0075] In one embodiment, the input signal includes at least one of the following: the pilot position frequency domain signal of the demodulated signal to be processed; the demodulated signal of the signal to be processed; the local pilot signal; and the least squares channel estimate.

[0076] In one embodiment, the number of feature layers for the demodulated pilot position frequency domain signal and the demodulated signal of the signal to be processed is 2 × the number of receiving antennas; the number of feature layers for the least squares channel estimate is 2 × the number of receiving antennas × the number of transmission layers; and the number of feature layers for the local pilot signal is 2 × the number of transmission layers.

[0077] In one embodiment, the pilot position frequency domain signal, the local pilot signal, and the least squares channel estimate after demodulation of the signal to be processed have values ​​only at the pilot positions, and are set to zero at the other positions.

[0078] In one embodiment, the output signal of the basic model exists in the form of a three-dimensional matrix; wherein, when the basic model is a first basic model, the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the number of feature layers; wherein, when the basic matrix is ​​a second basic model, the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the modulation scheme.

[0079] In this embodiment, the first basic model can be specifically understood as an AI-based channel estimation network, which serves the functions of denoising and time-frequency domain interpolation. The second basic model can be specifically understood as an AI-based receiver network, which can be used to replace the traditional channel estimation, channel equalization, and demodulation functions in the physical layer.

[0080] In one embodiment, when the base model is a first base model, the output signal of the base model is the channel estimate of each resource unit corresponding to the signal to be processed; when the base model is a second base model, the output signal of the base model is the soft information of each resource unit corresponding to the signal to be processed.

[0081] In this embodiment, soft information can be specifically understood as the log-likelihood ratio (LLR) of each resource unit obtained after the signal to be processed.

[0082] In one embodiment, the adaptive adjustment mechanism includes at least one of the following: when the number of subcarriers of the signal to be processed is greater than the number of subcarriers of the base model, calling the base model through a segmentation or sliding window mechanism; when the number of subcarriers of the signal to be processed is less than the number of subcarriers of the base model, and the waveform of the signal to be processed is a Cyclic Prefix-OFDM (CP-OFDM) waveform, calling the base model by randomly copying data, zero-filling, or linear extrapolation; when the number of subcarriers of the signal to be processed is less than the number of subcarriers of the base model, and the waveform of the signal to be processed is a Discrete Fourier Transform-spread (DFT) waveform. In the case of OFDM (DFT-s-OFDM) waveforms, the basic model is invoked by sequentially copying data or filling with zeros; based on the actual number of scheduled symbols of the signal to be processed, the information belonging to the actual number of scheduled symbols in the output of the basic model is extracted; when the signal to be processed and the uplink probe reference signal coexist, the information excluding the symbols occupied by the uplink probe reference signal is extracted from the output of the basic model; based on the pilot mode of the signal to be processed, the information belonging to the pilot position in the output of the basic model is removed; and the output of the basic model is pruned according to the modulation method of the signal to be processed.

[0083] In some examples, when the number of subcarriers of the signal to be processed is greater than the number of subcarriers of the base model, it can be assumed that the signal to be processed cannot be input into a base model as a whole for processing. In this case, the signal to be processed can be segmented according to the number of subcarriers of the base model, or the number of subcarriers of the base model can be used as the window size of the sliding window. The base model can be called through segmentation or sliding window mechanism so that the number of subcarriers of the signal to be processed input to the base model each time is consistent with the number of subcarriers that the base model can support. Then, the output of the base model can be processed accordingly each time to obtain the signal processing result corresponding to the signal to be processed.

[0084] In some examples, when the number of subcarriers of the signal to be processed is less than the number of subcarriers of the base model, and the waveform of the signal to be processed is a CP-OFDM waveform, in order to meet the input requirements of the base model, the signal to be processed can be supplemented at other positions where the number of subcarriers of the base model is less than the number of subcarriers of the base model by randomly copying data, filling with zeros, or linear extrapolation. Then, the base model is called to use the supplemented signal to be processed as the input of the base model for processing, and the signal processing result corresponding to the signal to be processed is obtained.

[0085] In some examples, when the number of subcarriers in the signal to be processed is less than the number of subcarriers in the base model, and the waveform of the signal to be processed is a DFT-s-OFDM waveform, in order to meet the input requirements of the base model, data can be supplemented at other positions where the number of subcarriers in the signal to be processed is less than the number of subcarriers in the base model by copying data or padding with zeros. Based on the waveform characteristics, the signal to be processed cannot be supplemented by linear extrapolation, and the timing of copying data must be consistent with the data storage order in the signal to be processed. That is, data supplementation needs to be achieved by sequentially copying data or padding with zeros. Then, the base model is called to use the supplemented signal to be processed as the input of the base model for processing, and the signal processing result corresponding to the signal to be processed is obtained.

[0086] In some examples, when the actual number of symbols scheduled for the signal to be processed does not match the number of resource block symbols in the input structure of the base model, in order to accurately determine the signal processing result, information at the corresponding symbol can be extracted from the output result of the base model based on the actual number of scheduled symbols.

[0087] In some examples, when the signal to be processed and the SRS coexist, since the SRS occupies one symbol, the output of the basic model can be extracted after applying the basic model to obtain the signal processing result corresponding to the signal to be processed.

[0088] In some examples, the pilot pattern of the signal to be processed can determine the pilot position in the output of the base model, and the information belonging to the pilot position can be removed.

[0089] In some examples, since different modulation methods require different output dimensions of the underlying model, and the underlying model is an AI model applicable to different modulation methods, when the dimension corresponding to the modulation method of the signal to be processed is smaller than the dimension corresponding to the modulation method of the underlying model, the part of the output result of the underlying model that exceeds the modulation method of the signal to be processed can be clipped to obtain the signal processing result corresponding to the signal to be processed.

[0090] The aforementioned adaptive adjustment mechanism applies to all base models, including both the first and second base models. However, when the base model is the second base model, some special adaptive adjustment mechanisms may also exist.

[0091] In one embodiment, when the base model is a second base model, the adaptive adjustment mechanism further includes at least one of the following: setting zeros to positions in the input base model that do not belong to the actual number of scheduled symbols, based on the actual number of scheduled symbols of the signal to be processed; setting zeros to symbols occupied by the uplink probe reference signal in the demodulated signal of the signal to be processed input to the base model when the signal to be processed and the uplink probe reference signal coexist; and adjusting the input and output signals of the base model according to the code division multiplexing group configuration information of the signal to be processed.

[0092] In some examples, when the actual number of symbols scheduled for the signal to be processed does not match the number of resource block symbols in the input structure of the base model, in order to make the signal input to the base model conform to the input structure of the base model, the positions of the input base model that do not belong to the actual number of scheduled symbols can be set to zero so that they do not affect the processing of the base model.

[0093] In some examples, when the signal to be processed and SRS coexist, considering the symbol occupancy of the signal to be processed by SRS, the symbol positions occupied by SRS in the demodulated signal of the signal to be processed input to the basic model can be zeroed so that they do not affect the processing of the basic model.

[0094] In some examples, based on the configuration information of the code division multiplexing group of the signal to be processed, it can be determined whether the resource units vacated on the pilot symbols can be used for data transmission. Different types of second basic models can be trained based on different situations. After determining the configuration information of the code division multiplexing group of the signal to be processed, a corresponding second basic model can be selected. However, if the selected second basic model is trained based on different situations, it is necessary to adjust the input and output to adapt the input and output structure of the second basic model. Therefore, the input and output signals of the basic model can be adjusted based on the configuration information of the code division multiplexing group of the signal to be processed.

[0095] In one embodiment, when the code division multiplexing group configuration information indicates that the resource units vacated on the pilot symbols cannot be used for data transmission, the input and output signals of the basic model are adjusted according to the code division multiplexing group configuration information of the signal to be processed, including:

[0096] Set all resource units except pilots in the demodulated signal of the input base model to be processed to zero; extract information other than pilot symbols from the output of the base model; wherein, the second base model is trained based on the configuration information of the code division multiplexing group, which is the case that the resource units vacated on the pilot symbols can be used for data transmission.

[0097] In one embodiment, if the second basic model is trained based on the configuration information of the code division multiplexing group, where the resource units vacated on the pilot symbols are not available for data transmission, then the second basic model cannot be used to process the signal to be processed when the configuration information of the code division multiplexing group is that the resource units vacated on the pilot symbols are available for data transmission.

[0098] In a specific example, when the second basic model is trained based on the configuration information of the code division multiplexing group, where the resource units vacated on the pilot symbols are available for data transmission, its applicability is wider. That is, it is applicable in both cases where the configuration information of the code division multiplexing group of the signal to be processed indicates that the resource units vacated on the pilot symbols are available for data transmission, and cases where the configuration information of the code division multiplexing group indicates that the resource units vacated on the pilot symbols are not available for data transmission. Only when the configuration information of the code division multiplexing group of the signal to be processed indicates that the resource units vacated on the pilot symbols are not available for data transmission, the corresponding processing is performed on the pilot symbols that the signal to be processed needs to be input to the second basic model, and only the soft information of other resource units besides the pilot symbols in the output of the second basic model is extracted. However, when the second basic model is a basic model trained based on the configuration information of the code division multiplexing group, where the resource units vacated on the pilot symbols cannot be used for data transmission, the input of the second basic model requires the resource units on the pilot symbols to be vacated. In other words, the signal to be processed that uses the resource units vacated on the pilot symbols for data transmission cannot be input into the second basic model for processing.

[0099] In one embodiment, the training samples of the base model include at least one of the following: channel estimation training samples containing all pilot modes; receiver network training samples containing all pilot modes; receiver network training samples containing all modulation schemes; and receiver network training samples constructed using the maximum value of the modulation scheme.

[0100] In a specific example, to ensure that the basic model in the physical layer of the wireless communication system covers as much physical layer parameter configuration information as possible, the basic model can be trained using channel estimation training samples containing all pilot modes to obtain a first basic model applicable to channel estimation of signals to be processed for all pilot modes. Alternatively, it can be trained using receiver network training samples containing all pilot modes to obtain a second basic model applicable to soft information determination of signals to be processed for all pilot modes. Furthermore, it can be trained using receiver network training samples containing all modulation schemes to obtain a second basic model applicable to soft information determination of signals to be processed for all modulation schemes. Finally, it can be trained using receiver network training samples constructed with the maximum value of the modulation scheme, enabling the trained second basic model to support inputs with the maximum value of the modulation scheme. When the modulation scheme of the signal to be processed is less than the maximum value, the aforementioned adaptive adjustment mechanism can be used to process the output of the second basic model, resulting in a second basic model applicable to soft information determination of signals to be processed for all modulation schemes.

[0101] The signal processing method of this application is illustrated below through some exemplary solutions. The basic model in the above embodiments can be understood as the neural network model in the AI ​​module below, and will not be repeated in the following description of the embodiments of this application.

[0102] Solution 1: A model training and inference mechanism is provided where an AI module replaces a certain module in the physical layer link, enabling the neural network model of this AI module to support different physical layer parameter configurations. Figure 2 is an example diagram of a model training and inference mechanism provided in an embodiment of this application. As shown in Figure 2, the key features of this embodiment are as follows:

[0103] 1) During the model training phase, the configuration of different physical layer parameters should be considered in the design of the input and output structures of the neural network model of the AI ​​module.

[0104] In a specific example, the input structure in the time domain can be designed as a resource block containing pilot formats. When training the neural network model, its input samples also need to contain samples of all pilot formats required by the wireless communication system as training samples. In the frequency domain, the input structure can be designed with a universal length to cover different delay spread channels. The output structure can be compatible with different modulation methods and designed with a universal length.

[0105] 2) During the model inference stage, the trained neural network model can flexibly support different physical layer parameter configurations under the adaptive adjustment mechanism.

[0106] In a specific example, the adaptive adjustment mechanism in the frequency domain may include segmentation, windowing, zero-padding, or duplication to adapt a neural network model with a fixed number of RBs to a configuration with other numbers of RBs; the adaptive adjustment mechanism in the time domain may include zero-padding or punching based on the specific physical layer parameter configuration in both the input and output dimensions.

[0107] Option 2: Provides a specific example of the input / output and training method for a base model that is the first base model. This corresponds to the case in Option 1 where the AI ​​module is an AI-based channel estimation network, i.e., the case where the base model is the first base model in the above embodiments. This AI-based channel estimation network serves to denoise and perform time-frequency domain interpolation. When designing the input and output of this neural network model, compatibility with different physical layer parameter configurations, such as pilot patterns, can be considered.

[0108] In some examples, Figure 3 is an example of a data processing flow diagram of a first basic model provided in an embodiment of this application. As shown in Figure 3, the signal to be processed is first subjected to a Fast Fourier Transform (FFT) for cyclic prefix removal and time-frequency conversion, and the resulting OFDM demodulated pilot signal is then used to perform frequency domain transformation. Local pilot at pilot location and least squares channel estimate As input to the neural network model, Figure 4A shows the input signal of a first basic model provided in an embodiment of this application. A structural example diagram; Figure 4B shows the input signal of the first basic model in Figure 4A provided in an embodiment of this application. A structural example diagram; Figure 4C shows the input signal of the first basic model in Figure 4A provided in an embodiment of this application. The structural example diagram shows the channel estimates obtained after processing by the first basic model on all resource blocks. Afterwards, The frequency domain signal excluding the pilot position after OFDM demodulation The signal is input to the channel equalization module for channel equalization processing, and then further processed by the subsequent demodulation and decoding modules to complete the processing of the required signal.

[0109] Figures 4A-4C illustrate a Single Input Multiple Output (SIMO) system as an example, using the system with Nsym=14 resource block symbols, NRe subcarriers, and Ndim feature layers to illustrate the three input signals. , and The three input signals are all in the form of three-dimensional NRe*Nsym*Ndim matrices, with values ​​at the pilot positions and zeros at other positions. As shown in Figures 4A-4C, the three matrices can be regarded as a set of training samples that can be used to train the first basic model.

[0110] In some examples, Nrx represents the number of receive antennas in a SIMO system, where Ndim = 2 (real and imaginary parts) * Nrx Ndim = 2 (real and imaginary parts). Ndim = 2 (real and imaginary parts) * Nrx. Figures 4A-4C illustrate pilot patterns with time-domain symbols 3 and 10 and frequency-domain subcarriers occupying an even number of positions. However, during the training phase of the first basic model, samples containing all pilot patterns required by the wireless communication system are needed as training samples. This means samples of pilot patterns different from those shown in Figures 4A-4C are required, allowing the trained first basic model to be applicable to different pilot pattern configurations. For example, Figure 5A shows the input signal of another first basic model provided in this embodiment. A structural example diagram; Figure 5B shows the input signal of the first basic model in Figure 5A provided in the embodiment of this application. A structural example diagram; Figure 5C shows the input signal of the first basic model in Figure 5A provided in the embodiment of this application. The structural example diagrams in Figures 5A-5C are examples of pilot patterns with time-domain symbols 2 and 11 and frequency-domain subcarriers occupying an odd number of bits.

[0111] The output of the first basic model described above is the channel estimate for all resource blocks. Figure 6 is an example diagram of the output structure of a first basic model provided in an embodiment of this application. As shown in Figure 6, the output signal of the first basic model also exists in the form of a three-dimensional NRe*Nsym*Ndim matrix, where Ndim=2 (real and imaginary parts)*Nrx.

[0112] Option 3: Provides a specific example of an input / output and training method where the base model is the second base model. This corresponds to the case in Option 1 where the AI ​​module is an AI-based receiver network, i.e., the case where the base model in the above embodiments is the second base model. This AI-based receiver network replaces the traditional functions of channel estimation, channel equalization, and demodulation. When designing the input and output of this neural network model, compatibility with different physical layer parameter configurations can be considered, such as pilot patterns and modulation methods.

[0113] In some examples, Figure 7 is an example diagram of the data processing flow of a second basic model provided in an embodiment of this application. As shown in Figure 7, the signal to be processed is first subjected to deCP and FFT processing, and the resulting OFDM demodulated signal is then processed. Local pilot at pilot location and least squares channel estimate As input to the neural network model, Figure 8A shows the input signal of a second basic model provided in an embodiment of this application. A structural example diagram; Figure 8B shows the input signal of the second basic model in Figure 8A provided in an embodiment of this application. A structural example diagram; Figure 8C shows the input signal of the second basic model in Figure 8A provided in an embodiment of this application. The structural example diagram is shown below. After obtaining the soft information LLR on all resource blocks through the second basic model, the LLR is further processed by the demapping module and the decoding module to complete the processing of the required signal.

[0114] Figures 8A-8C use a SIMO system as an example, with Nsym=14 resource block symbols, NRe subcarriers, and Ndim feature layers, to illustrate the three input signals. , and The three input signals are all represented as three-dimensional NRe*Nsym*Ndim matrices, where... and Values ​​are only present in the pilot positions, while zeros are filled in other positions. The three matrices shown in Figures 8A-8C, which have values ​​at all positions, can be regarded as a set of training samples that can be used to train the second base model.

[0115] In some examples, Nrx represents the number of receive antennas in a SIMO system, where Ndim = 2 (real and imaginary parts) * Nrx Ndim = 2 (real and imaginary parts). Ndim = 2 (real and imaginary parts) * Nrx. Figures 8A-8C illustrate pilot patterns with time-domain symbols 2 and 11 and frequency-domain subcarriers occupying an even number of bits. However, during the training phase of the second basic model, samples containing all pilot patterns required by the wireless communication system are needed as training samples. That is, samples of other pilot patterns different from those shown in Figures 8A-8C are required, so that the trained second basic model can be applied to different pilot pattern configurations. For example, Figure 9A shows the input signal of another second basic model provided in an embodiment of this application. A structural example diagram; Figure 9B shows the input signal of the second basic model in Figure 9A provided in an embodiment of this application. A structural example diagram; Figure 9C shows the input signal of the second basic model in Figure 9A provided in the embodiment of this application. The structural example diagrams in Figures 9A-9C are examples of pilot patterns with time-domain symbols 3 and 11 and frequency-domain subcarriers occupying an even number of positions.

[0116] The output of the second basic model is the LLR value on all resource blocks. Figure 10 is an example diagram of the output structure of the second basic model provided in this application embodiment. As shown in Figure 10, the output signal of the second basic model also exists in the form of a three-dimensional matrix. However, unlike the output signal dimension of the first basic model in the above scheme 2, the feature layer number Ndim is replaced by the modulation scheme Qm, that is, the output signal dimension of the second basic model is NRe*Nsym*Qm. If the maximum modulation scheme Qm=8 of the SIMO system, then the output dimension of the second basic model is NRe*14*8. This design allows the second basic model to be applicable to different modulation schemes Qm=2, Qm=4, Qm=6 and Qm=8. As shown in Figure 10, when Qm=6, the last 2 bits are truncated; when Qm=4, the last 4 bits are truncated; when Qm=2, the last 6 bits are truncated; and when Qm=1, the last 7 bits are truncated.

[0117] Following the AI-based receiver shown in Figure 7, there is a demapping module. This module extracts the LLR on the valid data resource unit based on the pilot pattern of the current sample, i.e., removes the LLR at the pilot position. Figure 11 shows two examples of the soft information structure on the final valid data resource unit provided in the embodiments of this application. The two examples of soft information structure shown in Figure 11 correspond to the final valid data resource units under the two pilot patterns shown in Figures 8A-8C and 9A-9C, respectively.

[0118] Based on the above explanation, when training the second basic model in terms of modulation scheme, it can be done in two ways: The first way is to use the input structure of the second basic model shown in Figures 8A-8C and 9A-9C as samples, and the output structure constructed with the maximum value of the modulation scheme as labels to construct receiver network training samples. During the training process, the output results obtained after inputting the samples into the second basic model are combined with the corresponding labels to construct a loss function and complete the training of the second basic model. The second way is to use the input structure of the second basic model shown in Figures 8A-8C and 9A-9C as samples, and the output format after demapping of each modulation scheme as shown in Figure 11 as labels to construct receiver network training samples. During the training process, the output results obtained after inputting the demapping samples into the second basic model are combined with the corresponding labels to construct a loss function and complete the training of the second basic model.

[0119] Solution 4: Provides a specific example of the number of subcarriers in the basic model. Based on Solutions 2 and 3, it is known that the value of NRe in the input dimension of the neural network model depends on the channel state. Because delay spread causes frequency-selective fading, which is described by coherent bandwidth, the smaller the coherent bandwidth, the larger the delay spread; conversely, the larger the coherent bandwidth, the smaller the delay spread. Therefore, channels with different delay spreads can be selected to choose an appropriate number of subcarriers. To cover different latency spreads, such as =72. It's understandable that this represents the NRe in the input dimensions of the base model when it's put into use.

[0120] Solution 5: Provides a concrete example of an adaptive adjustment mechanism applicable to all base models. In the input dimensions of the base models in Solutions 2 and 3, NRe is a fixed value, assumed to be... That is, both the first and second fundamental models can only handle the frequency domain. Data from one subcarrier and 14 symbol resource blocks; for simplicity, the third dimension Ndim in the "basic model" is temporarily ignored here. This is to ensure that users understand that NRe is not equal to... When Nsym=14, it is also possible to flexibly call the previously trained base model, which requires an adaptive adjustment mechanism to support it. This adaptive adjustment mechanism can be divided into the following cases:

[0121] 1) When NRe is greater than At that time, a segmented or sliding window mechanism is used to call the base model.

[0122] In some examples, the segmentation mechanism refers to when NRe is greater than And it is When the value is a multiple of N, the scheduling resource block can be split into N smaller blocks. The resource blocks of *14 are used to call the basic model separately, and then the N outputs are concatenated into NRe*14. Figure 12 is a flowchart of the segmentation mechanism in an adaptive adjustment mechanism provided in an embodiment of this application. As shown in Figure 12, in this example, NRe=288, Nsym=14, Taking 72 as an example, the processing flow of the segmentation mechanism is shown.

[0123] In some examples, when NRe is greater than Regardless of whether NRe is When it is an integer multiple of, it can be followed according to The window length is set, and the basic model is called separately using a sliding window with step=M. Then, the outputs on the overlapping subcarriers are merged, and finally the output on NRe*14 is obtained. Figure 13 is a flowchart of the sliding window mechanism in an adaptive adjustment mechanism provided in this application embodiment. As shown in Figure 13, in this example, NRe=96, Nsym=14, Taking =72 as an example, the processing flow of the sliding window mechanism is shown in Figure 13. =72, step=12, where the merging method on overlapping subcarriers can be either weighted merging or averaging.

[0124] 2) When NRe is less than When calling the base model, the method of copying data, filling in zeros, or linear extrapolation is used.

[0125] In some examples, when the waveform is CP-OFDM and NRe is less than At that time, it is possible to have insufficient At other positions of each subcarrier, data is supplemented by copying data, zero-filling, or linear extrapolation. Each subcarrier is used, and then the base model is called to obtain... After outputting the *14-dimensional data, the output of the first NRe*14 dimensions is taken as the final output to achieve flexible calling. There are no restrictions on the replication method for CP-OFDM waveforms; any method of replicating Nre*14 data from the source can be used. Figure 14 is a flowchart illustrating the random data replication for subcarrier completion in an adaptive adjustment mechanism provided by an embodiment of this application. Figure 14 shows NRe=48, Nsym=14. When the value is 72, data completion is achieved by copying the source NRe*14 data. The data processing flow for each subcarrier. Figure 15 is a flowchart illustrating the zero-padding subcarrier process in an adaptive adjustment mechanism provided by an embodiment of this application. Figure 15 shows NRe=48, Nsym=14. When the value is 72, fill in the blanks with zeros. The data processing flow for each subcarrier. Figure 16 is a flowchart illustrating the linear extrapolation of subcarriers in an adaptive adjustment mechanism provided in an embodiment of this application. Figure 16 shows NRe=48, Nsym=14. When the value is 72, it is completed by linear extrapolation. Data processing flow for each subcarrier.

[0126] In some examples, when the waveform is a DFT-s-OFDM waveform, NRe is less than and When it is an integer multiple of NRe, it can be less than NRe. At other positions on each subcarrier, data is padded with zeros by copying data. Each subcarrier is used, and then the base model is called to obtain... After outputting the *14-dimensional data, the output of NRe*14 dimensions is taken at N intervals along the subcarrier dimension as the final output to achieve flexible calling. For the replication method when the waveform is a DFT-s-OFDM waveform, restrictions need to be imposed. Figure 17 is a flowchart illustrating the sequential replication of data to complete the subcarriers in an adaptive mechanism provided by an embodiment of this application. Figure 17 shows NRe=36, Nsym=14. When =72, at this time It is twice that of NRe, and is completed by copying source NRe*14 data. This can only be achieved through this copying method for each subcarrier. Figure 18 is a flowchart illustrating a zero-padding subcarrier mechanism provided in an embodiment of this application. Figure 18 shows NRe=36, Nsym=14. When =72, at this time It is twice that of NRe, and is padded with zeros. Example of a method with multiple subcarriers.

[0127] Solution 6: A specific example of an adaptive adjustment mechanism for the first fundamental model is given. In Solution 2, the first fundamental model can handle frequency domain... The data of resource blocks of N subcarriers and Nsym symbols, the output of this first basic model is Channel estimation on Nsym resource blocks The dimensions are NRe*Nsym*Ndim, where NRe and Ndim are fixed. An adaptive adjustment mechanism is implemented on the Nsym dimension to support different physical layer parameter configurations. This adaptive adjustment mechanism can be divided into the following cases:

[0128] 1) Extract channel estimates based on the actual number of scheduled symbols actually used by the user for subsequent channel equalization processing. Currently, it is also necessary to remove channel estimates from pilot resource units based on pilot patterns.

[0129] In some examples, the actual number of scheduled symbols is 3-13. In this case, channel equalization can be performed simply by extracting the channel estimates from the output symbols 3-13 of the first basic model. Figure 19 is an example of the output result of extracting channel estimates based on the actual number of scheduled symbols in an adaptive adjustment mechanism provided by an embodiment of this application. The channel estimates on the pilot resource units are omitted in Figure 19; specific cases can be adaptively processed according to the pilot pattern.

[0130] 2) Based on the parameter of whether the current service data coexists with SRS, extract the channel estimate value on the actual scheduled service data symbol. At present, it is also necessary to remove the channel estimate value on the pilot resource unit according to the pilot mode.

[0131] In some examples, assuming that the current service data coexists with the SRS, and the SRS occupies the last symbol 13, the channel estimates on symbols 0-12 are extracted for channel equalization. Figure 20 is an example of the output result of extracting channel estimates based on whether the data coexists with the SRS in an adaptive adjustment mechanism provided by an embodiment of this application. It can be understood that the channel estimates on the pilot resource units are omitted in Figure 20, and the specific situation can be adaptively processed according to the pilot pattern.

[0132] Solution 7: A specific example of an adaptive adjustment mechanism for the second fundamental model is given. In Solution 3, the second fundamental model can handle frequency domain... The data of resource blocks of N subcarriers and Nsym symbols, the output of this second basic model is The soft information LLR value on the Nsym resource block has the dimension NRe*Nsym*Qm. Here, the dimensions NRe and Qm are fixed, and an adaptive adjustment mechanism is implemented for the Nsym dimension to support different physical layer parameter configurations. This adaptive adjustment mechanism can be divided into the following cases:

[0133] 1) Adjust the input and output of the second basic model according to the actual number of scheduling symbols actually scheduled by the user to adapt to different parameter configurations.

[0134] In some examples, assuming the actual number of scheduled symbols is 3-13, then the input of the second basic model... We need to set the values ​​of symbols 0-2 to zero. and The symbols remain unchanged (because their symbol values ​​are zero for 0-2), and then are fed into the second basic model. The specific input format is shown in Figures 21A-21C. Figure 21A shows the adaptive input signal based on the actual number of scheduled symbols in an adaptive adjustment mechanism provided in an embodiment of this application. A structural example diagram; Figure 21B is an adaptive input signal based on the actual number of scheduled symbols in the adaptive adjustment mechanism of Figure 21A provided in an embodiment of this application. A structural example diagram; Figure 21C shows the adaptive input signal based on the actual number of scheduled symbols in the adaptive adjustment mechanism of Figure 21A provided in an embodiment of this application. The structural example diagram is shown below. For the output of the second basic model, the soft information LLR on symbols 3-13 needs to be extracted, and the specific output structure is shown in Figure 22. Figure 22 is an example diagram of the output result of extracting LLR based on the actual number of scheduled symbols in an adaptive adjustment mechanism provided by an embodiment of this application. The output result is then sent to the demapping module, which removes the soft information LLR at the pilot positions according to the pilot pattern to obtain the final effective data resource unit's LLR.

[0135] 2) Adjust the input and output of the second basic model to adaptively configure different parameters based on whether the current business data coexists with SRS.

[0136] In some examples, assuming that business data and SRS coexist at the current moment, with SRS occupying the last symbol 13, then the input of the second basic model... The value of symbol 13 needs to be set to zero. and The input signal remains unchanged (because its symbol 13 is zero) and is then fed into the second basic model. The specific input format is shown in Figures 23A-23C. Figure 23A shows the adaptive input signal based on whether it coexists with SRS in an adaptive adjustment mechanism provided by an embodiment of this application. A structural example diagram; Figure 23B is an adaptive input signal based on whether or not it coexists with SRS in the adaptive adjustment mechanism of Figure 23A provided in an embodiment of this application. A structural example diagram; Figure 23C is an adaptive input signal based on whether or not it coexists with SRS in the adaptive adjustment mechanism of Figure 23A provided in an embodiment of this application. The structural example diagram is shown below. For the output of the second basic model, it is necessary to extract the soft information LLR on symbols 0-12, and the specific output structure is shown in Figure 24. Figure 24 is an example diagram of the output result of extracting LLR based on whether it coexists with SRS in an adaptive adjustment mechanism provided by an embodiment of this application. The output result is then sent to the demapping module, which removes the soft information LLR at the pilot positions according to the pilot pattern to obtain the LLR on the final effective data resource unit.

[0137] 3) The input and output of the second basic model are adaptively adjusted based on the configuration information of the Code Division Multiplexing Group (CDM group) in the physical layer parameters.

[0138] In some examples, when CDM group=1, the empty resource units on the pilot symbols can be used for data transmission; when CDM group=2, the empty resource units on the pilot symbols cannot be used for data transmission. Assuming the second base model is trained under the CDM group=1 configuration, when CDM group=2, the input and output need to be adjusted to adapt the second base model. In this case, the input of the second base model... The values ​​of the resource units on the pilot symbols, excluding the pilots themselves, need to be set to zero. and The input signals remain unchanged and are then fed into the second basic model. The specific input format is shown in Figures 25A-25C. Figure 25A shows the adaptive input signal based on the configuration information of the CDM group in an adaptive adjustment mechanism provided by an embodiment of this application. A structural example diagram; Figure 25B shows the adaptive input signal based on the configuration information of the CDM group in the adaptive adjustment mechanism of Figure 25A provided in the embodiment of this application. A structural example diagram; Figure 25C shows the adaptive input signal based on the configuration information of the CDM group in the adaptive adjustment mechanism of Figure 25A provided in the embodiment of this application. The structural example diagram is shown below. For the second output of the second basic model, the soft information LLR other than the pilot symbols is extracted by the demapping module, and the final output result is shown in Figure 26. Figure 26 is an example diagram of the adaptive output result based on the configuration information of the CDM group in an adaptive adjustment mechanism provided by an embodiment of this application.

[0139] If the second base model is trained under CDM group=2, it cannot be used to process signals under CDM group=1.

[0140] Solution 8: Provides a specific example of the basic model input and output method for Multiple Input Multiple Output (MIMO) systems.

[0141] The input and output structure design of the basic model mentioned in Scheme 2-7 above should consider different physical layer parameter configurations, as well as some adaptive adjustment mechanisms for the input and output of the model when different physical layer parameter configurations are used during basic model inference. These are also applicable to MIMO systems.

[0142] The difference between MIMO and SIMO systems lies in the number of transmission layers considered in the input and output dimensions of the basic model. Figure 27A shows the input signal of the basic model of a MIMO system provided in an embodiment of this application. A structural example diagram; Figure 27B shows the input signal of the basic model of the MIMO system in Figure 27A provided in the embodiment of this application. Structural example diagrams. Figures 27A-27B show the inputs of the first and second basic models. and The structure in a MIMO system also exists in the form of a three-dimensional matrix NRe*Nsym*Ndim. Compared with schemes 2 and 3, the Ndim dimension value is different here. Taking LayerN as the number of transmission layers as an example, where... Ndim = 2 (real and imaginary parts) * LayerN, Ndim = 2 (real and imaginary parts) * Nrx * LayerN. In the output dimension NRe * Nsym * Ndim based on the first fundamental model, Ndim = 2 (real and imaginary parts) * Nrx * LayerN, with the specific structure shown in Figure 27B. The structure is similar, except that the zero-filled part has an estimated H value. In the output dimension NRe*Nsym*Ndim based on the second basic model, Ndim=Qm*LayerN, and the structure is similar to Figure 10 in Scheme 3. The embodiments of this application will not be described in detail here.

[0143] In one exemplary embodiment, FIG28 is a schematic diagram of a signal processing device provided in an embodiment of this application, which is applied to a communication node. As shown in FIG28, the device includes:

[0144] The signal acquisition module 210 is used to acquire the signal to be processed.

[0145] The signal processing module 220 is used to call the basic model corresponding to the signal to be processed based on the physical layer parameter configuration information and adaptive adjustment mechanism of the signal to be processed, and to determine the signal processing result.

[0146] In one embodiment, the physical layer parameter configuration information includes at least one of the following: pilot mode; modulation scheme; actual number of scheduled symbols; coexistence status of the signal to be processed and the uplink probe reference signal; and configuration information of the code division multiplexing group.

[0147] In one embodiment, the input signal of the basic model exists in the form of a three-dimensional matrix; wherein the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the number of feature layers.

[0148] In one embodiment, the input signal includes at least one of the following: the pilot position frequency domain signal of the demodulated signal to be processed; the demodulated signal of the signal to be processed; the local pilot signal; and the least squares channel estimate.

[0149] In one embodiment, the number of feature layers for the demodulated pilot position frequency domain signal and the demodulated signal of the signal to be processed is 2 × the number of receiving antennas; the number of feature layers for the least squares channel estimate is 2 × the number of receiving antennas × the number of transmission layers; and the number of feature layers for the local pilot signal is 2 × the number of transmission layers.

[0150] In one embodiment, the pilot position frequency domain signal, the local pilot signal, and the least squares channel estimate after demodulation of the signal to be processed have values ​​only at the pilot positions, and are set to zero at the other positions.

[0151] In one embodiment, the output signal of the basic model exists in the form of a three-dimensional matrix; wherein, when the basic model is a first basic model, the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the number of feature layers; wherein, when the basic matrix is ​​a second basic model, the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the modulation scheme.

[0152] In one embodiment, when the base model is a first base model, the output signal of the base model is the channel estimate of each resource unit corresponding to the signal to be processed; when the base model is a second base model, the output signal of the base model is the soft information of each resource unit corresponding to the signal to be processed.

[0153] In one embodiment, the adaptive adjustment mechanism includes at least one of the following: when the number of subcarriers of the signal to be processed is greater than the number of subcarriers of the basic model, calling the basic model through a segmentation or sliding window mechanism; when the number of subcarriers of the signal to be processed is less than the number of subcarriers of the basic model, and the waveform of the signal to be processed is a cyclic prefix-orthogonal frequency division multiplexing waveform, calling the basic model through random data copying, zero-filling, or linear extrapolation; when the number of subcarriers of the signal to be processed is less than the number of subcarriers of the basic model, and the waveform of the signal to be processed is an orthogonal frequency division multiplexing waveform based on discrete Fourier transform, calling the basic model through sequential data copying or zero-filling; extracting information belonging to the actual number of scheduled symbols in the output of the basic model according to the actual number of scheduled symbols of the signal to be processed; extracting information excluding symbols occupied by the uplink probe reference signal from the output of the basic model when the signal to be processed and the uplink probe reference signal coexist; removing information belonging to the pilot position in the output of the basic model according to the pilot mode of the signal to be processed; and pruning the output of the basic model according to the modulation method of the signal to be processed.

[0154] In one embodiment, when the base model is a second base model, the adaptive adjustment mechanism further includes at least one of the following: setting zeros to positions in the input base model that do not belong to the actual number of scheduled symbols, based on the actual number of scheduled symbols of the signal to be processed; setting zeros to symbols occupied by the uplink probe reference signal in the demodulated signal of the signal to be processed input to the base model when the signal to be processed and the uplink probe reference signal coexist; and adjusting the input and output signals of the base model according to the code division multiplexing group configuration information of the signal to be processed.

[0155] In one embodiment, when the code division multiplexing group configuration information indicates that the resource units vacated on the pilot symbols cannot be used for data transmission, the input and output signals of the base model are adjusted according to the code division multiplexing group configuration information of the signal to be processed. This includes: setting the resource units other than the pilots in the demodulated signal of the signal to be processed input to the base model to zero; and extracting information other than the pilot symbols from the output result of the base model. The second base model is trained based on the case where the code division multiplexing group configuration information indicates that the resource units vacated on the pilot symbols can be used for data transmission.

[0156] In one embodiment, if the second basic model is trained based on the configuration information of the code division multiplexing group, where the resource units vacated on the pilot symbols are not available for data transmission, then the second basic model cannot be used to process the signal to be processed when the configuration information of the code division multiplexing group is that the resource units vacated on the pilot symbols are available for data transmission.

[0157] In one embodiment, the training samples of the base model include at least one of the following: channel estimation training samples containing all pilot modes; receiver network training samples containing all pilot modes; receiver network training samples containing all modulation schemes; and receiver network training samples constructed using the maximum value of the modulation scheme.

[0158] This application embodiment also provides a communication node. Figure 29 is a structural schematic diagram of a communication node provided in this application embodiment. As shown in Figure 29, the communication node provided in this application embodiment includes a memory 320, a processor 310, and a computer program stored in the memory and executable on the processor. When the processor 310 executes the program, it implements the above-mentioned signal processing method.

[0159] The communication node may also include a memory 320; the processor 310 in the communication node may be one or more, with one processor 310 as an example in FIG29; the memory 320 is used 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 signal processing method as described in the embodiments of this application.

[0160] The communication node also includes: a communication device 330, an input device 340, and an output device 350.

[0161] The processor 310, memory 320, communication device 330, input device 340 and output device 350 in the communication node can be connected by a bus or other means. Figure 29 shows an example of connection by bus.

[0162] Input device 340 can be used to receive input digital or character information, and to generate key signal inputs related to user settings and function control of the communication node. Output device 350 may include display devices such as a display screen.

[0163] The communication device 330 may include a receiver and a transmitter. The communication device 330 is configured to perform information transmission and reception communication under the control of the processor 310.

[0164] The 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 signal processing method described in the embodiments of this application (e.g., signal acquisition module 210 and signal processing module 220). The memory 320 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the communication node, etc. Furthermore, the memory 320 may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some instances, the memory 320 may further include memory remotely located relative to the processor 310, and these remote memories can be connected to the communication node 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.

[0165] This application also provides a storage medium storing a computer program, which, when executed by a processor, implements any of the signal processing methods described in this application.

[0166] Optionally, the signal processing method includes: acquiring the signal to be processed; and, based on the physical layer parameter configuration information and adaptive adjustment mechanism of the signal to be processed, calling the basic model corresponding to the signal to be processed to determine the signal processing result.

[0167] The computer storage medium in this application embodiment can be any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. For example, a computer-readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable CD-ROM, optical storage device, magnetic storage device, or any suitable combination thereof. The computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0168] Computer-readable signal media may include data signals propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device.

[0169] Program code contained on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, radio frequency (RF), etc., or any suitable combination thereof.

[0170] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0171] Optionally, embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the signal processing method provided in any embodiment of this application.

[0172] The above description is merely an exemplary embodiment of this application and is not intended to limit the scope of protection of this application.

[0173] Those skilled in the art will understand that the term user terminal encompasses any suitable type of wireless user equipment, such as mobile phones, portable data processing devices, portable web browsers, or vehicle-mounted mobile stations.

[0174] Generally, the various embodiments of this application can be implemented in hardware or dedicated circuitry, software, logic, or any combination thereof. For example, some aspects can be implemented in hardware, while others can be implemented in firmware or software that can be executed by a controller, microprocessor, or other computing device, although this application is not limited thereto.

[0175] Embodiments of this application can be implemented by executing computer program instructions through the data processor of a mobile device, for example, in a processor entity, or through hardware, or through a combination of software and hardware. The computer program instructions can be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages.

[0176] Any block diagram of logical flow in the accompanying drawings of this application may represent program operations, or may represent interconnected logic circuits, modules, and functions, or may represent a combination of program operations and logic circuits, modules, and functions. The computer program may be stored on memory. The memory may be of any type suitable to 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 Video Disc (DVD) or Compact Disk (CD), etc.). Computer-readable media may include non-transitory storage media. The data processor may be of any type suitable to the local technical environment, such as, but not limited to, general-purpose computers, special-purpose computers, microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), and processors based on multi-core processor architectures.

Claims

1. A signal processing method, comprising: Acquire the signal to be processed; Based on the physical layer parameter configuration information and adaptive adjustment mechanism of the signal to be processed, the basic model corresponding to the signal to be processed is invoked to determine the signal processing result.

2. The signal processing method according to claim 1, wherein, The physical layer parameter configuration information includes at least one of the following: Pilot mode; Modulation method; Actual number of scheduled symbols; The coexistence of the signal to be processed and the uplink detection reference signal; Configuration information for code division multiplexing groups.

3. The signal processing method according to claim 1, wherein, The input signal of the basic model exists in the form of a three-dimensional matrix; The dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the number of feature layers.

4. The signal processing method according to claim 3, wherein, The input signal includes at least one of the following: The pilot position frequency domain signal after demodulation of the signal to be processed; The demodulated signal of the signal to be processed; Local pilot signal; Least squares channel estimate.

5. The signal processing method according to claim 4, wherein, The number of feature layers of the pilot position frequency domain signal after demodulation of the signal to be processed and the demodulated signal of the signal to be processed is 2 × the number of receiving antennas; The number of feature layers in the least squares channel estimate is 2 × the number of receiving antennas × the number of transmission layers; The number of characteristic layers of the local pilot signal is 2 × the number of transmission layers.

6. The signal processing method according to claim 4, wherein, The frequency domain signal of the pilot position after demodulation of the signal to be processed, the local pilot signal, and the least squares channel estimate have values ​​only at the pilot positions, and are set to zero at the other positions.

7. The signal processing method according to claim 1, wherein, The output signal of the basic model exists in the form of a three-dimensional matrix. Wherein, when the basic model is the first basic model, the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the number of feature layers; Wherein, when the basic matrix is ​​the second basic model, the dimensions of the three-dimensional matrix include the number of resource block symbols, the number of subcarriers, and the modulation scheme.

8. The signal processing method according to claim 7, wherein, When the base model is the first base model, the output signal of the base model is the channel estimate value of multiple resource units corresponding to the signal to be processed; When the base model is the second base model, the output signal of the base model is soft information of multiple resource units corresponding to the signal to be processed.

9. The signal processing method according to claim 1, wherein, The adaptive adjustment mechanism includes at least one of the following: If the number of subcarriers of the signal to be processed is greater than the number of subcarriers of the basic model, the basic model is invoked through a segmentation or sliding window mechanism. When the number of subcarriers of the signal to be processed is less than the number of subcarriers of the basic model, and the waveform of the signal to be processed is a cyclic prefix-orthogonal frequency division multiplexing waveform, the basic model is invoked by randomly copying data, filling zeros, or linear extrapolation. When the number of subcarriers of the signal to be processed is less than the number of subcarriers of the basic model, and the waveform of the signal to be processed is an orthogonal frequency division multiplexing waveform based on discrete Fourier transform, the basic model is called by sequentially copying data or filling with zeros. Based on the actual number of scheduled symbols of the signal to be processed, extract the information belonging to the actual number of scheduled symbols from the output of the basic model; In the case where the signal to be processed and the uplink probe reference signal coexist, the information excluding the symbols occupied by the uplink probe reference signal is extracted from the output of the basic model. Based on the pilot pattern of the signal to be processed, remove the information belonging to the pilot position from the output of the basic model; The output of the basic model is pruned according to the modulation scheme of the signal to be processed.

10. The signal processing method according to claim 9, wherein, When the base model is a second base model, the adaptive adjustment mechanism further includes at least one of the following: Based on the actual number of scheduled symbols of the signal to be processed, set the positions in the input basic model that do not belong to the actual number of scheduled symbols to zero; When the signal to be processed and the uplink probe reference signal coexist, the symbol occupied by the uplink probe reference signal in the demodulated signal of the signal to be processed input to the basic model is set to zero; The input and output signals of the basic model are adjusted according to the configuration information of the code division multiplexing group of the signal to be processed.

11. The signal processing method according to claim 10, wherein, When the configuration information of the code division multiplexing group indicates that the resource units vacated on the pilot symbols cannot be used for data transmission, adjusting the input and output signals of the basic model according to the configuration information of the code division multiplexing group of the signal to be processed includes: Set all resource units except the pilot in the demodulated signal of the signal to be processed input into the basic model to zero; Extract information from the output of the basic model, excluding the pilot symbols; The second basic model is trained based on the configuration information of the code division multiplexing group, where the resource units vacated on the pilot symbol by the pilot can be used for data transmission.

12. The signal processing method according to claim 10, wherein, If the second basic model is trained based on the configuration information of the code division multiplexing group, where the resource units vacated on the pilot symbols are not available for data transmission, then the second basic model cannot be used to process the signal to be processed when the configuration information of the code division multiplexing group is that the resource units vacated on the pilot symbols are available for data transmission.

13. The signal processing method according to any one of claims 1-12, wherein, The training samples for the base model include at least one of the following: Channel estimation training samples containing all pilot modes; Receiver network training samples containing all pilot modes; Training samples for receiver networks containing all modulation schemes; Receiver network training samples constructed using the maximum value of the modulation scheme.

14. A communication node, comprising: The program includes a memory, a processor, a program stored in the memory and executable on the processor, and a data bus configured to enable communication between the processor and the memory, wherein the program, when executed by the processor, implements the signal processing method as described in any one of claims 1-13.

15. A storage medium configured as a computer-readable storage medium storing at least one program that can be executed by at least one processor to implement the signal processing method as claimed in any one of claims 1-13.

16. A computer program product comprising a computer program that, when executed by a processor, implements the signal processing method as described in any one of claims 1-13.

Images