Channel estimation and equalization method and device

Abstract

The present application discloses a channel estimation and equalization method and device, which relates to the field of optical communication. When performing channel estimation, a demodulation reference signal (DMRS) may not be inserted, which can not only improve the spectrum efficiency but also improve the accuracy of channel estimation. The method includes: a first node receives a first received signal, and obtains a first transmitted signal based on the first received signal and a target neural network model. The first transmitted signal is a signal that the second node wants to send to the first node. The target neural network model is used to obtain a signal sent by the second node corresponding to the signal received by the first node based on an offset quadrature amplitude modulation (OQAM) symbol corresponding to the signal received by the first node.

Description

Technical Field

[0001] The present application relates to the field of optical communications, and in particular to a channel estimation and equalization method and device based on a residual convolutional neural network. Background Art

[0002] As we all know, Internet services need to pass through backbone networks, local area networks, and access networks before they can be transmitted to the user end. Among them, the development of backbone networks and local area networks is extremely rapid, while the development of access networks is relatively slow, becoming a bottleneck for the development of the entire network broadband. In order to increase the transmission rate of the access network, passive optical network (PON) technology was introduced.

[0003] There are many PON technologies available today, such as asynchronous transfer mode (ATM) PON (APON), Ethernet PON (EPON), gigabit PON (GPON), orthogonal frequency division multiplexing (OFDM)-based PON (OFDM-PON), or filter bank multi-carrier (FBMC)-based PON (FBMC-PON), etc.

[0004] Among them, APON, EPON and GPON use time division multiplexing technology, while OFDM-PON and FBMC-PON use frequency division multiplexing technology. The modulation format of APON, EPON and GPON is simple to implement and has low power consumption. However, since APON, EPON and GPON use time division multiplexing technology, the spectrum efficiency of APON, EPON and GPON is not high. Moreover, under the condition of high transmission rate, fiber dispersion will cause serious channel damage and limit the transmission distance. OFDM-PON can improve the spectrum efficiency to a certain extent compared with APON, EPON and GPON. However, in order to ensure the anti-interference ability of OFDM-PON, a cyclic prefix needs to be added when transmitting signals in OFDM-PON, which will also affect the spectrum efficiency of OFDM-PON. However, when transmitting signals in FBMC-PON, a cyclic prefix does not need to be added, so FBMC-PON has higher spectrum efficiency. In addition, FBMC-PON can flexibly allocate spectrum resources, and each carrier does not need to maintain synchronization and good out-of-band suppression. Therefore, the application of FBMC-PON is becoming more and more extensive. However, the problem of imaginary interference is serious when performing channel estimation in FBMC-PON, resulting in low accuracy of channel estimation. Summary of the invention

[0005] The present application provides a channel estimation and equalization method and device, which can avoid inserting a demodulation reference signal (DMRS) when performing channel estimation, thereby improving not only spectrum efficiency but also channel estimation accuracy.

[0006] To achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0007] In a first aspect, an embodiment of the present application provides a channel estimation and equalization method, the method comprising: a first node receives a first received signal; the first node obtains a first transmitted signal based on the first received signal and a target neural network model; wherein the first transmitted signal is a signal that the second node wants to send to the first node, and the target neural network model is used to obtain a signal sent by the second node corresponding to the signal received by the first node based on an offset quadrature amplitude modulation (OQAM) symbol corresponding to the signal received by the first node.

[0008] In the method provided by the first method mentioned above, after the first node receives the first received signal, it can obtain the first transmitted signal according to the first received signal and the target neural network model. In this way, the first node can obtain the signal that the second node wants to send to the first node. On the one hand, in the method provided in the first aspect, there is no need to insert DMRS, which improves the spectrum efficiency. On the other hand, in the process of obtaining the first transmitted signal, the first node does not need to perform channel estimation and equalization based on DMRS, but uses the target neural network model to perform channel estimation and equalization. The target neural network model is trained based on the marked signal, which can eliminate the problem of imaginary interference of FBMC-PON and improve the accuracy of channel estimation.

[0009] In a possible implementation, the first node obtains the first transmission signal according to the first received signal and the target neural network model, including: the first node obtains the first OQAM symbol corresponding to the first received signal according to the first received signal; the first node inputs the first OQAM symbol into the target neural network model to obtain the first transmission signal. Based on the above method, the first node can input the first OQAM symbol into the target neural network model to obtain the first transmission signal.

[0010] In a possible implementation, the method further includes: the first node obtains an initial neural network model; the first node trains the initial neural network model according to OQAM symbols corresponding to a large number of received signals and signals sent by the second node corresponding to each received signal to obtain a target neural network model. Based on the above method, the first node can train the initial neural network model according to the labeled signals to obtain the target neural network model. Subsequently, the first node can obtain the transmitted signal according to the target neural network model.

[0011] In a possible implementation, the first node obtains the initial neural network model, including: the first node obtains the initial neural network model according to the modulation format of the first received signal and / or the transmission distance of the first received signal. Based on the above method, the first node can determine a suitable initial neural network model according to the modulation format of the first received signal and / or the transmission distance of the first received signal. The initial neural network models corresponding to the first received signals with different modulation formats or different transmission distances are different.

[0012] In a possible implementation, the target neural network model includes at least two target convolutional layers, at least one target residual network layer, a target flatten layer and a target fully connected layer FCL; at least two target convolutional layers and at least one target residual network layer are alternately connected in series, and among the at least two target convolutional layers, the target convolutional layer farthest from the input end of the target neural network model is connected to one end of the target flatten layer, and the other end of the target flatten layer is connected to the target FCL; the target residual network layer is used to reduce interference of a signal received by the first node under high-order modulation.

[0013] In one possible implementation, at least one target residual network layer includes at least one target convolutional layer and an adder, at least one target convolutional layer is connected to the adder, and the adder is used to add an input signal of the target residual network layer to an output signal of the at least one target convolutional layer.

[0014] In a second aspect, an embodiment of the present application provides a communication device that can implement the method in the first aspect or any possible implementation of the first aspect. The device includes corresponding units or components for executing the above method. The units included in the device can be implemented by software and / or hardware. The device can be, for example, a terminal, or a chip, a chip system, or a processor that can support the terminal to implement the above method.

[0015] In a third aspect, an embodiment of the present application provides a communication device, comprising: a processor, the processor is coupled to a memory, the memory is used to store programs or instructions, when the program or instructions are executed by the processor, the device implements the method described in the above-mentioned first aspect, or any possible implementation method of the first aspect.

[0016] In a fourth aspect, an embodiment of the present application provides a communication device, which is used to implement the method described in the above-mentioned first aspect or any possible implementation method of the first aspect.

[0017] In a fifth aspect, an embodiment of the present application provides a computer-readable medium having a computer program or instructions stored thereon, which, when executed, enables a computer to execute the method described in the first aspect or any possible implementation of the first aspect.

[0018] In a sixth aspect, an embodiment of the present application provides a computer program product, which includes a computer program code. When the computer program code is run on a computer, the computer executes the method described in the above-mentioned first aspect or any possible implementation of the first aspect.

[0019] In the seventh aspect, an embodiment of the present application provides a chip, comprising: a processor, the processor is coupled to a memory, the memory is used to store programs or instructions, when the program or instructions are executed by the processor, the chip implements the method described in the above-mentioned first aspect, or any possible implementation method of the first aspect.

[0020] It can be understood that any of the communication devices, chips, computer-readable media, or computer program products provided above are used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1A A schematic diagram of the optical network system architecture provided in an embodiment of the present application;

[0022] Figure 1B A schematic diagram of the structure of a transmitting end and a receiving end based on a target neural network model provided in an embodiment of the present application;

[0023] Figure 2 A schematic diagram of the hardware structure of a communication device provided in an embodiment of the present application;

[0024] Figure 3 Schematic diagram 1 of a flow chart of a channel estimation and equalization method provided in an embodiment of the present application;

[0025] Figure 4 Schematic diagram 1 of the target neural network model provided in the embodiment of the present application;

[0026] Figure 5 Schematic diagram of the target neural network model provided in the embodiment of the present application Figure 2 ;

[0027] Figure 6 A schematic diagram of the structure of a transmitting end and a receiving end based on digital signal processing (DSP) provided in an embodiment of the present application;

[0028] Figure 7 A schematic diagram of the sensitivity of a first node receiving a signal provided in an embodiment of the present application;

[0029] Figure 8 Schematic diagram of the process of the channel estimation and equalization method provided in the embodiment of the present application Figure 2 ;

[0030] Fig. 9 A schematic diagram of the structure of a communication device provided in an embodiment of the present application;

[0031] Fig.10 A schematic diagram of the composition of a communication system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0032] The implementation of the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0033] The method provided in the embodiment of the present application can be used in various optical network systems that require channel estimation and equalization. Figure 1A Taking the optical network system 10 as an example, the method provided in the embodiment of the present application is described.

[0034] like Figure 1A , which is a schematic diagram of the architecture of the optical network system 10 provided in an embodiment of the present application. Figure 1A In the embodiment, the optical network system 10 may include an optical line terminal (OLT) 101, an optical distribution network (ODN) 102 connected to the OLT 101, and optical network units (ONUs) 103-105 connected to the ODN 102. Figure 1A In the optical network system 10 shown, the direction from OLT 101 to ONU 103, ONU 104 or ONU 105 is called a downstream direction, and the direction from ONU 103, ONU 104 or ONU 105 to OLT 101 is called an upstream direction. Figure 1A It is only a schematic diagram and does not constitute a limitation on the applicable scenarios of the technical solution provided in this application.

[0035] Among them, OLT 101 is the core component of optical network system 10, which is equivalent to a switch or router in a traditional communication network. OLT 101 is generally placed at the central office and provides a fiber optic interface of PON for users. OLT 101 can be connected to the upper network (e.g., local area network) on the one hand to complete the uplink access of PON network, and can be connected to the user terminal equipment (e.g., ONU 103-ONU 105) through ODN 102 on the other hand to realize the control, management and ranging functions of the user terminal equipment.

[0036] The ODN 102 is a passive optical splitter, generally including a passive optical splitter, a trunk optical fiber, and a branch optical fiber, etc. The ODN 102 can aggregate and transmit the upstream signals of multiple ONUs to the OLT 101, and can also transmit the downstream signals of the OLT 101 to each ONU.

[0037] ONU 103, ONU 104 or ONU 105 is a user-end device in the optical network system 10, which is placed at the user end and used in conjunction with OLT 101 to implement Ethernet layer 2 and layer 3 functions and provide users with various services, such as voice, data or multimedia services. ONU 103, ONU 104 or ONU 105 can selectively receive signals sent by OLT 101 on the one hand, and can cache the user's Ethernet data on the other hand, and send signals in the upstream direction in the sending time slot port allocated by OLT 101.

[0038] In some embodiments, Figure 1A The ONU in the Figure 1B The method shown sends an uplink signal to the OLT. Figure 1A The OLT in Figure 1B The method shown sends a downstream signal to the ONU.

[0039] For example, Figure 1BThe transmitting end is ONU and the receiving end is OLT (that is, ONU sends an uplink signal to OLT) as an example. After ONU obtains the transmission signal to be sent to OLT, it can obtain an optical signal by quadrature amplitude modulation (QAM), offset quadrature amplitude modulation (OQAM), inverse fast Fourier transform (IFFT), filter group 1, parallel-to-serial conversion, up-conversion and Mach-Zehnder modulator (MZM). Subsequently, the optical signal is passed through an attenuator and an optoelectronic converter to obtain an electrical signal, which will be sent to OLT. After OLT receives the electrical signal, it will obtain OQAM symbols by down-conversion, serial-to-parallel conversion, filter group 2, fast Fourier transform (FFT) and OQAM demodulation, and then input the OQAM symbols into the target neural network model to obtain the transmission signal. In this way, OLT can obtain the signal that ONU wants to send to OLT. It can be understood that when the transmitting end is OLT and the receiving end is ONU (that is, OLT sends a downstream signal to ONU), the process of signal sending and receiving is similar to the case where the transmitting end is ONU and the receiving end is OLT. Therefore, the process of signal sending and receiving in the case where the transmitting end is ONU and the receiving end is OLT can be referred to and will not be repeated.

[0040] Among them, the specific introduction of QAM, OQAM, IFFT, filter bank, parallel-to-serial conversion, up-conversion, MZM, attenuator, photoelectric converter, down-conversion, serial-to-parallel conversion, filter bank 2, FFT and OQAM demodulation can be referred to the explanation in conventional technology and will not be repeated here.

[0041] The target neural network model can be used to obtain the signal sent by the ONU corresponding to the signal received by the OLT according to the OQAM symbol corresponding to the signal received by the OLT. The target neural network model can be a software function module running on a hardware device (e.g., a chip, etc.) in the OLT, or it can be a component in the OLT.

[0042] Understandably, in Figure 1A In the optical network system 10 shown, the target neural network model can be deployed on the OLT or on the ONU. The target neural network model can also be deployed independently. In other words, the target neural network model is neither deployed on the OLT nor on the ONU, but is independently deployed in the optical network system 10. For example, the optical network system 10 also includes a neural network module ( Figure 1A (not shown), the neural network module is connected to OLT 101 and / or ONU (for example, at least one of ONU103-ONU 105), and the neural network module includes the above-mentioned target neural network model.

[0043] In this case, if the ONU sends an uplink signal to the OLT, after the OLT receives the electrical signal corresponding to the uplink signal, it will down-convert the electrical signal, perform serial-to-parallel conversion, filter bank 2, FFT and OQAM demodulation to obtain the OQAM symbol. Subsequently, the OLT can send the OQAM symbol to the neural network module, and after the neural network module receives the OQAM symbol, it will input the OQAM symbol into the target neural network model to obtain the above-mentioned uplink signal, and send the uplink signal to the OLT.

[0044] Similarly, if the OLT sends a downstream signal to the ONU, after the ONU receives the electrical signal corresponding to the downstream signal, it will perform down-conversion, serial-to-parallel conversion, filter group 2, FFT and OQAM demodulation on the electrical signal to obtain the OQAM symbol. Subsequently, the ONU can send the OQAM symbol to the neural network module. After receiving the OQAM symbol, the neural network module will input the OQAM symbol into the target neural network model to obtain the above-mentioned downstream signal, and send the downstream signal to the ONU.

[0045] Figure 1A The optical network system 10 shown is only used as an example and is not used to limit the technical solution of the present application. Those skilled in the art should understand that in the specific implementation process, the optical network system 10 may also include other devices, and the number of OLTs, ODNs, ONUs or neural network modules may also be determined according to specific needs without limitation.

[0046] Optionally, the present application embodiment Figure 1A Each device in the embodiment, such as OLT 101, ONU 103, ONU 104, ONU 105 or a neural network module, can be a functional module in a device. It is understandable that the functional module can be a component in a hardware device, such as a chip or component in an OLT or ONU, or a software functional module running on hardware, or a virtualized function instantiated on a platform (e.g., a cloud platform).

[0047] For example, Figure 1A Each device in the Figure 2 The communication device 200 is used to implement the communication device 200. Figure 2 The hardware structure diagram of the communication device applicable to the embodiment of the present application is shown in FIG.

[0048] The processor 201 may be a general-purpose central processing unit (CPU), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits for controlling the execution of the program of the present application.

[0049] The communication link 202 may include a path to transmit information between the above-mentioned components, such as a bus.

[0050] The communication interface 204 uses any transceiver-like device for communicating with other devices or communication networks, such as an Ethernet interface, a radio access network interface (RAN), a wireless local area network interface (WLAN), etc.

[0051] The memory 203 can be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), a disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer, but is not limited thereto. The memory can be independent and connected to the processor through a communication line 202. The memory can also be integrated with the processor. The memory provided in the embodiment of the present application can generally have non-volatility. Among them, the memory 203 is used to store the computer execution instructions involved in the execution of the present application scheme, and the execution is controlled by the processor 201. The processor 201 is used to execute the computer-executable instructions stored in the memory 203, so as to implement the method provided in the embodiment of the present application.

[0052] Optionally, the computer-executable instructions in the embodiments of the present application may also be referred to as application code, which is not specifically limited in the embodiments of the present application.

[0053] In a specific implementation, as an embodiment, the processor 201 may include one or more CPUs, such as Figure 2 CPU0 and CPU1 in.

[0054] In a specific implementation, as an embodiment, the communication device 200 may include multiple processors, such as Figure 2 201 and processor 207 in the embodiment of the present invention. Each of these processors may be a single-CPU processor or a multi-CPU processor. The processor herein may refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0055] In a specific implementation, as an embodiment, the communication device 200 may further include an output device 205 and an input device 206. The output device 205 communicates with the processor 201 and may display information in a variety of ways. For example, the output device 205 may be a liquid crystal display (LCD), a light emitting diode (LED) display device, a cathode ray tube (CRT) display device, or a projector. The input device 206 communicates with the processor 201 and may receive user input in a variety of ways. For example, the input device 206 may be a mouse, a keyboard, a touch screen device, or a sensor device.

[0056] The communication device 200 can be a general device or a dedicated device. In a specific implementation, the communication device 200 can be an OLT, an ONU or a Figure 2 The embodiment of the present application does not limit the type of the communication device 200.

[0057] Combine the following Figure 1A , Figure 1B and Figure 2 , taking the target neural network model deployed in the first node as an example, the channel estimation and equalization method provided in the embodiment of the present application is specifically explained.

[0058] It should be noted that in the embodiments of the present application, " / " can indicate that the objects associated with each other are in an "or" relationship, for example, A / B can indicate A or B; "and / or" can be used to describe that there are three relationships between associated objects, for example, A and / or B can indicate: A exists alone, A and B exist at the same time, and B exists alone, where A and B can be singular or plural.

[0059] In order to facilitate the description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the words "first", "second" and the like may be used to distinguish between technical features with the same or similar functions. The words "first", "second" and the like do not limit the quantity and execution order, and the words "first", "second" and the like do not necessarily limit the differences. In the embodiments of the present application, the words "exemplary" or "for example" are used to represent examples, illustrations or explanations, and any embodiment or design described as "exemplary" or "for example" should not be interpreted as being more preferred or more advantageous than other embodiments or design. The use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way for easy understanding.

[0060] It should be noted that in the embodiments of the present application, for a technical feature, the technical features in the technical feature are distinguished by "first", "second", "third", "A", "B", "C" and "D", and there is no order of precedence or size between the technical features described by the "first", "second", "third", "A", "B", "C" and "D".

[0061] It can be understood that the same step or steps or technical features with the same functions in the embodiments of the present application can be referenced to each other in different embodiments.

[0062] It is understandable that in the embodiment of the present application, the first node or the second node can perform some or all of the steps in the embodiment of the present application, and these steps are only examples. The embodiment of the present application can also perform other steps or variations of various steps. In addition, each step can be performed in a different order presented in the embodiment of the present application, and it is possible that not all of the steps in the embodiment of the present application need to be performed.

[0063] In the embodiment of the present application, the specific structure of the execution subject of the channel estimation and equalization method is not particularly limited in the embodiment of the present application, as long as the method of the embodiment of the present application can be executed. For example, the execution subject of the channel estimation and equalization method provided in the embodiment of the present application may be a first node, or a component applied in the first node, such as a chip, and the present application does not limit this. Alternatively, the execution subject of the channel estimation and equalization method provided in the embodiment of the present application may be a second node, or a component applied in the second node, such as a chip, and the present application does not limit this. The following embodiments are described by taking the execution subjects of the channel estimation and equalization method as the first node and the second node as examples.

[0064] like Figure 3 As shown, a channel estimation and equalization method provided in an embodiment of the present application includes steps 301 and 302.

[0065] Step 301: A first node receives a first reception signal.

[0066] The first node can be Figure 1A For example, the first node may be ONU 103, ONU 104 or ONU 105; or, the first node may be Figure 1A OLT 101 in.

[0067] In a possible implementation manner, before step 301, the second node sends a first transmission signal.

[0068] The first transmission signal is a signal that the second node wants to send to the first node. The first transmission signal may be generated by a higher layer of the second node, or may be received by the second node from other devices.

[0069] If the first node is Figure 1A Any ONU in the Figure 1A OLT 101 in. If the first node is Figure 1A The second node can be OLT 101 in Figure 1A For example, the second node may be ONU 103, ONU 104 or ONU 105.

[0070] In a possible implementation, the second node obtains the first transmission signal, and obtains the first optical signal by QAM, OQAM, IFFT, filter bank 1, parallel-to-serial conversion, up-conversion and MZM. Subsequently, the first optical signal is sent to the first node through an attenuator and an optoelectronic converter. For details, please refer to Figure 1B The corresponding description.

[0071] It is understandable that there are various interferences or signal losses in the process from the second node sending a signal to the first node receiving the signal, such as channel interference, loss caused by optical dispersion, or nonlinear effects of optical devices, so the signal received by the first node is different from the signal sent by the second node. In other words, when the first transmitted signal reaches the first node after QAM, OQAM, IFFT, filter bank 1, parallel-to-serial conversion, up-conversion, MZM, attenuator, and photoelectric converter, it is a signal different from the first transmitted signal. In the embodiment of the present application, the signal that reaches the first node after multiple processing of the first transmitted signal, for example, after QAM, OQAM, IFFT, filter bank 1, parallel-to-serial conversion, up-conversion, MZM, attenuator, and photoelectric converter, is called the first received signal.

[0072] It can be understood that the first transmission signal is not only a signal that the second node wants to send to the first node, but also a signal that the second node wants to send to the first node and other nodes. Figure 1A In OLT101, the first node is Figure 1A Taking ONU 105 in FIG. 1 as an example, the first transmission signal may be a signal that OLT 101 wants to send to ONU 104 and ONU 105; or, the first transmission signal may be a signal that OLT 101 wants to send to ONU 103, ONU 104 and ONU 105; or, the first transmission signal may be a signal that OLT 101 wants to send to ONU 103 and ONU 105. The embodiment of the present application is introduced by taking the first transmission signal as a signal that the second node wants to send to the first node as an example. For the case where the first transmission signal is a signal that the second node wants to send to the first node and other nodes, reference may be made to the description of the first transmission signal as a signal that the second node wants to send to the first node in the embodiment of the present application, which will not be described in detail.

[0073] Step 302: The first node obtains a first transmission signal according to the first received signal and the target neural network model.

[0074] The target neural network model can be used to obtain a signal sent by a second node corresponding to the signal received by the first node according to the OQAM symbol corresponding to the signal received by the first node. The OQAM symbol corresponding to the signal received by the first node is a symbol obtained by multiple processing of the signal received by the first node, for example, after down-conversion, serial-to-parallel conversion, filter bank 2, FFT and OQAM demodulation.

[0075] For example, the target neural network model can be used to obtain a first transmission signal according to a first OQAM symbol corresponding to a first received signal. The first OQAM symbol is a symbol obtained by processing the first received signal multiple times, for example, after down-conversion, serial-to-parallel conversion, filter bank 2, FFT, and OQAM demodulation. For details, please refer to Figure 1B The corresponding description.

[0076] Further, the first OQAM symbol includes M subcarriers, each subcarrier is a complex number, that is, each subcarrier includes a real part and an imaginary part. M is the number of subcarriers occupied by the first received signal.

[0077] A possible implementation method is that the first node obtains a first transmission signal based on the first received signal and the target neural network model, including: the first node obtains a first OQAM symbol based on the first received signal; the first node inputs the first OQAM symbol into the target neural network model to obtain a first transmission signal.

[0078] Exemplarily, after receiving the first received signal, the first node processes the first received signal multiple times to obtain a first OQAM symbol. The first node inputs the first OQAM symbol into the target neural network model to obtain a first transmitted signal. In this way, the first node can obtain the signal that the second node wants to send to the first node, and then process the first transmitted signal, or process the service according to the first transmitted signal.

[0079] It can be understood that before the first node inputs the first OQAM symbol into the target neural network model, the first OQAM symbol can be transformed into a first vector and the first vector can be input into the target neural network model. The dimension of the first vector is M*2. The element a in the first vector is i,1 is the real part of the i-th subcarrier among the M subcarriers included in the first OQAM symbol, a i,2 is the imaginary part of the i-th subcarrier among the M subcarriers included in the first OQAM symbol; or, i,1 is the imaginary part of the i-th subcarrier among the M subcarriers included in the first OQAM symbol, a i,2 is the real part of the ith subcarrier among the M subcarriers included in the first OQAM symbol. i is an integer greater than 0 and less than M.

[0080] It can be understood that the first node inputs the first OQAM symbol into the target neural network model to obtain the second vector; the first node obtains the first transmission signal according to the second vector. The dimension of the second vector is 2*M. The element b in the second vector 1,j is the real part of the jth subcarrier among the M subcarriers occupied by the first transmitted signal, b 2,j is the imaginary part of the jth subcarrier among the M subcarriers occupied by the first transmitted signal; or, b 1,j is the imaginary part of the jth subcarrier among the M subcarriers occupied by the first transmitted signal, b 2,j is the real part of the jth subcarrier among the M subcarriers occupied by the first transmitted signal. j is an integer greater than 0 and less than M.

[0081] In a possible implementation, the target neural network model includes at least two target convolutional layers, at least one target residual network layer, a target flatten layer and a target fully connected layer (fully connected layer, FCL). Further, at least two target convolutional layers and at least one target residual network layer are alternately connected in series. Among the at least two target convolutional layers, the target convolutional layer farthest from the input end of the target neural network model is connected to one end of the target flatten layer, and the other end of the target flatten layer is connected to the target FCL.

[0082] It can be understood that the signal between the target convolution layer and the target residual network layer, and the signal between the target convolution layer and the target flatten layer can be processed by the first activation function and the first normalization operation. That is to say, for the target convolution layer, the target residual network layer and the target flatten layer, the output signal of one layer can be processed by the first activation function and the first normalization operation and then input to the other layer. Taking the serial connection of the target convolution layer and the target residual network layer as an example, the output signal of the target convolution layer is processed by the first activation function and the first normalization operation and then input to the target residual network layer. The first activation function between each two layers can be the same or different. The first normalization operation between each two layers can be the same or different.

[0083] It can be understood that the output signal of the target FCL can be processed by the second activation function and then output as the target neural network model. The second activation function is different from the first activation function.

[0084] Among them, the introduction of at least two target convolutional layers, the target flatten layer and the target FCL can refer to the explanation of the convolutional layer, the flatten layer and the FCL in the conventional technology, which will not be repeated here.

[0085] The target residual network layer is used to reduce interference of the signal received by the first node under high-order modulation. For example, the target residual network layer is used to reduce the loss caused by optical dispersion or nonlinear effects of optical devices under high-order modulation of the signal received by the first node.

[0086] In one possible implementation, at least one target residual network layer includes at least one target convolutional layer and an adder. At least one target convolutional layer is connected to the adder. The adder is used to add an input signal of the target residual network layer to an output signal of at least one target convolutional layer. It can be understood that if at least one target convolutional layer includes two or more convolutional layers, the convolutional layers in at least one target convolutional layer are connected in series.

[0087] It can be understood that in each target residual network layer, the signal between the target convolution layer and the target convolution layer, and the signal between the target convolution layer and the adder can be processed by the third activation function and the second normalization operation. Taking the target convolution layer and the adder as an example, the output signal of the target convolution layer is processed by the third activation function and the second normalization operation and then input into the adder. The third activation function in at least one target residual network layer can be the same or different, and the second normalization operation in at least one target residual network layer can be the same or different. For example, the third activation function between the target convolution layer and the target convolution layer can be the same or different from the third activation function between the target convolution layer and the adder, and the second normalization operation between the target convolution layer and the target convolution layer can be the same or different from the second normalization operation between the target convolution layer and the adder. It should be understood that the third activation function and the first activation function can be the same or different, the third activation function and the second activation function are different, and the second normalization operation and the first normalization operation can be the same or different. In the present application, the activation function (eg, the first activation function, the second activation function, or the third activation function) can enable the target neural network model to have better nonlinear expression capabilities.

[0088] Exemplarily, taking the case where the target neural network model includes two target convolutional layers, a target residual network layer, a target flatten layer and FCL, and the target residual network layer includes a target convolutional layer and an adder, the target neural network model may be as shown in FIG. 4 . Figure 4 In FIG. 4 , the target neural network model includes a target convolution layer 1, a target residual network layer, a target convolution layer 2, a target flatten layer and an FCL. Among them, the target residual network layer includes a target convolution layer 3 and an adder. After the first vector is input into the target neural network model shown in FIG. 4 , after passing through the target convolution layer 1, it will be processed by the first activation function (e.g., Leaky Relu) and the first normalization operation (e.g., Batch Normalization) to obtain a third vector. After passing through the target convolution layer 3, the third vector will be processed by the third activation function (e.g., Leaky Relu) and the second normalization operation (e.g., Batch Normalization) to obtain a fourth vector. After passing through the target convolution layer 2, the sum of the third vector and the fourth vector will be processed by the first activation function and the first normalization operation, and the processed vector will pass through the target flatten layer to obtain a one-dimensional vector. After passing through the target FCL, the one-dimensional vector will pass through the second activation function (e.g., sigmoid) to obtain a first transmission signal.

[0089] in, Figure 4The target convolution layer in can adopt one-dimensional convolution, and the dimension of the convolution kernel of the convolution layer can be set as needed. For example, the dimension of the convolution kernel can be 2, 3 or 4. The number of bits output by the target convolution layer 1 is the same as the number of bits N of the first received data. Figure 4 The number of bits output by other target convolutional layers in can be equal to N or greater than N. For example, the number of bits output by target convolutional layer 3 and target convolutional layer 2 is 2N.

[0090] For example, taking the target neural network model including three target convolutional layers, two target residual network layers, a target flatten layer and FCL, each target residual network layer including two target convolutional layers and an adder (it should be understood that the number of target convolutional layers included in each target residual network layer may be different), the target neural network model may be as follows: Figure 5 shown. Figure 5 In , the target neural network model includes target convolution layer 1, target residual network layer 1, target convolution layer 2, target residual network layer 2, target convolution layer 3, target flatten layer and FCL. Among them, the target residual network layer 1 includes target convolution layer 4, target convolution layer 5 and adder 1. The target residual network layer 2 includes target convolution layer 6, target convolution layer 7 and adder 2. The first vector is input to Figure 5 After the target neural network model shown, after passing through the target convolution layer 1, it will be processed by the first activation function (such as Leaky Relu) and the first normalization operation (such as Batch Normalization) to obtain a third vector. After the third vector passes through the target convolution layer 4, it will be processed by the third activation function (such as Leaky Relu) and the second normalization operation (such as Batch Normalization). After the processed vector passes through the target convolution layer 5, it will be processed by the third activation function and the second normalization operation to obtain a fourth vector. After the sum of the third vector and the fourth vector passes through the target convolution layer 2, it will be processed by the first activation function and the first normalization operation to obtain a fifth vector. After the fifth vector passes through the target convolution layer 6, it will be processed by the third activation function and the second normalization operation. After the processed vector passes through the target convolution layer 7, it will be processed by the third activation function and the second normalization operation to obtain a sixth vector. After the sum of the fifth vector and the sixth vector passes through the target convolution layer 3, it will be processed by the first activation function and the first normalization operation. The processed vector will pass through the target flatten layer to obtain a one-dimensional vector. After the one-dimensional vector passes through the target FCL, it will obtain the first transmission signal through the second activation function (such as sigmoid).

[0091] in, Figure 5The target convolution layer in can adopt one-dimensional convolution, and the dimension of the convolution kernel of the convolution layer can be set as needed. For example, the dimension of the convolution kernel can be 2, 3 or 4. The number of bits output by the target convolution layer 1 is the same as the number of bits N of the first received data. Figure 5 The number of bits output by other target convolutional layers in can be equal to N or greater than N. For example, the number of bits output by target convolutional layer 5 and target convolutional layer 2 is 2N.

[0092] It should be noted that the number of target convolutional layers, the number of target residual network layers, and the number of target convolutional layers included in the target neural network model can be set as needed. Generally, the more target convolutional layers, the more target residual network layers, and the more target convolutional layers included in the target residual network layer, the higher the accuracy of the target neural network model. However, if the number of target convolutional layers, the number of target residual network layers, and the number of target convolutional layers included in the target residual network model exceed the threshold, overfitting may occur, in which case the accuracy of the target neural network model will decrease.

[0093] It should be noted that the target neural network model does not include a pooling layer. The pooling layer uses downsampling to reduce the size of the grid in the target neural network to simplify the computational complexity. Therefore, if the target neural network model includes a pooling layer, it will have a significant impact on the accuracy of the target neural network model, resulting in an inaccurate first transmission signal obtained by the target neural network model.

[0094] In addition, in the case where the target neural network model is not used for channel estimation and equalization, the second node needs to insert DMRS when sending the first transmission signal.

[0095] For example, the first node and the second node use DSP to perform channel estimation and equalization. Figure 6 As shown, after the second node obtains the transmission signal to be sent to the first node, it can insert the transmission signal into DMRS after QAM and OQAM, and then obtain the optical signal by IFFT, filter group 1, parallel-to-serial conversion, up-conversion and MZM. Subsequently, the optical signal is passed through an attenuator and an optoelectronic converter to obtain an electrical signal, which is sent to the first node. After receiving the electrical signal, the first node will equalize the signal obtained by down-conversion, serial-to-parallel conversion, filter group 2 and FFT of the electrical signal, as well as the signal obtained by the first node after channel estimation according to DMRS. Subsequently, the first node will demodulate the equalized signal through OQAM and OAM to obtain the transmission signal.

[0096] Figure 6In the case of using DSP for channel estimation and equalization, DMRS will occupy certain frequency domain resources. Therefore, the channel estimation and equalization method provided by the embodiment of the present application has higher spectrum efficiency than the method using DSP for channel estimation and equalization.

[0097] Please refer to Figure 7 , Figure 7 The following figure shows the use of DSP for channel estimation and equalization at different transmission distances, as well as the use of Figure 5 The sensitivity of the first node receiving the signal when the target neural network model performs channel estimation and equalization. Figure 7 In FIG. 1 , the horizontal axis represents the received optical power (received optical rower), and the vertical axis represents the negative logarithm of the bit error rate (BER) (-Log(BER)). Figure 7 The BER in is 10 -3 .

[0098] Figure 7 (a) shows that when the modulation format is 4QAM, DSP is used for channel estimation and equalization, and Figure 5 The sensitivity of the first node receiving the signal when the target neural network model performs channel estimation and equalization. Figure 7 In (a), curve 1 is used to indicate the back-to-back (B2B) communication. Figure 5 The sensitivity of the first node receiving the signal when the target neural network model is used for channel estimation and equalization. Curve 2 is used to indicate that the transmission distance is 20km and the Figure 5 The sensitivity of the first node receiving the signal when the target neural network model is used for channel estimation and equalization. Curve 3 is used to indicate that the transmission distance is 50km and the Figure 5 The sensitivity of the first node receiving the signal when the target neural network model is used for channel estimation and equalization. Curve 4 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 20km and DSP is used for channel estimation and equalization. Curve 5 is used to indicate the sensitivity of the first node receiving the signal when DSP is used for channel estimation and equalization in B2B communication. Curve 6 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 80km and DSP is used for channel estimation and equalization. Figure 5 The sensitivity of the first node receiving the signal when the target neural network model shown is used for channel estimation and equalization. Curve 7 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 50km and the DSP is used for channel estimation and equalization. Curve 8 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 80km and the DSP is used for channel estimation and equalization.

[0099] For B2B communication, the value of curve 1 is higher than that of curve 5. In B2B communication, signal transmission is still affected by devices and noise. Figure 5 The target neural network model shown in the figure can bring bit mapping gain when performing channel estimation and equalization. When the transmission distance is 20km, the value of curve 2 is higher than that of curve 4. When the transmission distance is 50km, the value of curve 3 is higher than that of curve 7. When the transmission distance is 80km, the value of curve 6 is higher than that of curve 8. It can be seen from the above that when the modulation format is 4QAM, under the same conditions, the value of curve 2 is higher than that of curve 4. Figure 5 When the target neural network model shown in the figure performs channel estimation and equalization, the sensitivity of the first node receiving the signal is greater than that when the DSP is used for channel estimation and equalization. Figure 5 The advantages of the target neural network model shown in the figure for channel estimation and equalization are more obvious.

[0100] Figure 7 (b) shows that when the modulation format is 16QAM, DSP is used for channel estimation and equalization, and Figure 5 The sensitivity of the first node receiving the signal when the target neural network model performs channel estimation and equalization. Figure 7 In (b), curve 1 is used to indicate the B2B communication. Figure 5 The sensitivity of the first node receiving the signal when the target neural network model is used for channel estimation and equalization. Curve 2 is used to indicate that the transmission distance is 20km and the Figure 5 The sensitivity of the first node receiving the signal when the target neural network model is used for channel estimation and equalization. Curve 3 is used to indicate that the transmission distance is 50km and the Figure 5 Curve 4 is used to indicate the sensitivity of the first node receiving the signal when the DSP is used for channel estimation and equalization in B2B communication. Curve 5 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 20km and the DSP is used for channel estimation and equalization. Curve 6 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 80km and the DSP is used for channel estimation and equalization. Figure 5 The sensitivity of the first node receiving the signal when the target neural network model shown is used for channel estimation and equalization. Curve 7 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 50km and the DSP is used for channel estimation and equalization. Curve 8 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 80km and the DSP is used for channel estimation and equalization.

[0101] For B2B communication, the value of curve 1 is higher than that of curve 4. In B2B communication, signal transmission is still affected by devices and noise. Figure 5 The target neural network model shown in the figure can bring bit mapping gain when performing channel estimation and equalization. When the transmission distance is 20km, the value of curve 2 is higher than that of curve 5. When the transmission distance is 50km, the value of curve 3 is higher than that of curve 7. When the transmission distance is 80km, the value of curve 6 is higher than that of curve 8. It can be seen from the above that when the modulation format is 16QAM, under the same conditions, the use of Figure 5 When the target neural network model shown in FIG5 is used for channel estimation and equalization, the sensitivity of the first node receiving the signal is greater than that when the DSP is used for channel estimation and equalization. Moreover, the greater the transmission distance, the more obvious the advantage of using the target neural network model shown in FIG5 for channel estimation and equalization.

[0102] It is understandable that the longer the signal transmission distance, the more serious the signal power attenuation. Usually, when the signal is transmitted on an 80km optical fiber, the signal power attenuation is already quite serious. If DSP is used for channel estimation and equalization, error-free transmission may not be achieved. In this case, it is necessary to amplify the optical signal through an erbium doped optical fiber amplifier (EDFA). Figure 7 (c) shows the sensitivity of the first node receiving the signal when the transmission distance is 80 km, DSP is used for channel estimation and equalization, the gain in the EDFA parameters is set to 13 dB, and the noise figure (NF) is set to 5.0 dB. Figure 7 (c) in Figure 1 also shows that the transmission distance is 80 km. Figure 5 The sensitivity of the first node receiving the signal when the target neural network model performs channel estimation and equalization.

[0103] Figure 7 In (c), curve 1 is used to indicate that the transmission distance is 80 km. Figure 5The sensitivity of the first node receiving the signal when the target neural network model shown is used for channel estimation and equalization. Curve 2 is used to indicate the sensitivity of the first node receiving the signal when the transmission distance is 80km, DSP is used for channel estimation and equalization, and the gain in the EDFA parameters is set to 13dB and the NF is set to 5.0dB. Obviously, the value of curve 1 is greater than the value of curve 2. That is to say, when the transmission distance is 80km, when DSP is used for channel estimation and equalization, even if the signal is amplified to a certain extent, the sensitivity of the first node receiving the signal is less than that when the DSP is used for channel estimation and equalization. Figure 5 The sensitivity of the first node receiving the signal when the target neural network model performs channel estimation and equalization.

[0104] based on Figure 3 In the method shown, after the first node receives the first received signal, it can obtain the first transmitted signal according to the first received signal and the target neural network model. In this way, the first node can obtain the signal that the second node wants to send to the first node. Figure 3 In the method shown, there is no need to insert DMRS, which improves spectrum efficiency. On the other hand, in the process of obtaining the first transmission signal, the first node does not need to perform channel estimation and equalization based on DMRS, but uses the target neural network model to perform channel estimation and equalization. The target neural network model is trained based on the marked signal, which can eliminate the problem of imaginary interference of FBMC-PON and improve the accuracy of channel estimation.

[0105] Further optional, such as Figure 8 As shown, in Figure 3 In one possible implementation of the method shown, Figure 3 The method shown also includes step 801 and step 802 .

[0106] Step 801: The first node obtains an initial neural network model.

[0107] In a possible implementation, the initial neural network model includes at least two initial convolutional layers, at least one initial residual network layer, an initial flatten layer and an initial FCL. Furthermore, at least two initial convolutional layers and at least one initial residual network layer are alternately connected in series. Among the at least two initial convolutional layers, the initial convolutional layer farthest from the input end of the initial neural network model is connected to one end of the initial flatten layer, and the other end of the initial flatten layer is connected to the initial FCL.

[0108] It can be understood that the signal between the initial convolution layer and the initial residual network layer, and the signal between the initial convolution layer and the initial flatten layer can be processed by the first activation function and the first normalization operation. That is to say, for the two layers connected in series among the initial convolution layer, the initial residual network layer and the initial flatten layer, the output signal of one layer can be processed by the first activation function and the first normalization operation and then input to the other layer.

[0109] It can be understood that the output signal of the initial FCL can be processed by the second activation function and then output as the initial neural network model.

[0110] In one possible implementation, at least one initial residual network layer includes at least one initial convolutional layer and an adder. At least one initial convolutional layer is connected to the adder. The adder is used to add the input signal of the initial residual network layer to the output signal of at least one initial convolutional layer. It can be understood that if at least one initial convolutional layer includes two or more convolutional layers, the convolutional layers in at least one initial convolutional layer are connected in series.

[0111] It can be understood that in each initial residual network layer, the signal between the initial convolution layer and the initial convolution layer, and the signal between the initial convolution layer and the adder can be processed by the third activation function and the second normalization operation.

[0112] In a possible implementation, the first node obtains the initial neural network model, including: the first node obtains the initial neural network model according to the modulation format of the first received signal and / or the transmission distance of the first received signal. In other words, the first node can determine the number of initial convolutional layers in the initial neural network model, the number of initial residual network layers, and the number of initial convolutional layers in the initial residual network layer according to the modulation format of the first received signal and / or the transmission distance of the first received signal.

[0113] It can be understood that the higher the modulation order of the modulation format of the first received signal, the greater the number of initial convolutional layers in the initial neural network model, the number of initial residual network layers, or the number of initial convolutional layers in the initial residual network layer. The lower the modulation order of the modulation format of the first received signal, the smaller the number of initial convolutional layers in the initial neural network model, the number of initial residual network layers, or the number of initial convolutional layers in the initial residual network layer. The longer the transmission distance of the first received signal, the greater the number of initial convolutional layers in the initial neural network model, the number of initial residual network layers, or the number of initial convolutional layers in the initial residual network layer. The shorter the transmission distance of the first received signal, the smaller the number of initial convolutional layers in the initial neural network model, the number of initial residual network layers, or the number of initial convolutional layers in the initial residual network layer.

[0114] Exemplarily, when the modulation format is 4QAM, the initial neural network model may include two initial convolutional layers, an initial residual network layer, an initial flatten layer and an initial FCL. The initial residual network layer may include two initial convolutional layers and an adder.

[0115] Exemplarily, when the modulation format is 16QAM, the initial neural network model may include three initial convolutional layers, two initial residual network layers, an initial flatten layer and an initial FCL. Each initial residual network layer may include an initial convolutional layer and an adder.

[0116] It should be noted that when determining the number of initial convolutional layers in the initial neural network model, the number of initial residual network layers, and the number of initial convolutional layers in the initial residual network layers, overfitting of the target neural network model should be avoided.

[0117] It is understandable that the initial neural network models corresponding to different modulation formats and / or different transmission distances may be different.

[0118] Step 802: The first node trains an initial neural network model based on OQAM symbols corresponding to a large number of received signals and a signal sent by the second node corresponding to each received signal to obtain a target neural network model.

[0119] In one possible implementation, the first node updates the initial neural network model using a back-propagation method based on OQAM symbols corresponding to a large number of received signals and a signal sent by the second node corresponding to each received signal to obtain a target neural network model.

[0120] Exemplarily, the first node uses the back propagation method to determine the gradient vector according to the OQAM symbols corresponding to a large number of received signals and the signal sent by the second node corresponding to each received signal, and adjusts the weight of the initial neural network model through the gradient vector. If the adjusted initial neural network model converges, the adjusted initial neural network model is determined as the target neural network model. If the adjusted initial neural network model does not converge, the back propagation method is continued to be used to determine the gradient vector, and the weight of the newly obtained neural network model is adjusted through the gradient vector until the adjusted neural network model converges.

[0121] It can be understood that different modulation formats and / or different transmission distances correspond to different received signals. Therefore, different modulation formats and / or different transmission distances correspond to different target neural network models.

[0122] It can be understood that step 801 and step 802 can also be implemented as a separate embodiment.

[0123] based on Figure 8 In the method shown, the first node can obtain an initial neural network model, and train the initial neural network model based on OQAM symbols corresponding to a large number of received signals and a signal sent by the second node corresponding to each received signal, to obtain a target neural network model. In this way, the first node can obtain a signal sent by the second node corresponding to the signal received by the first node based on the signal received by the first node and the target neural network model.

[0124] The above mainly introduces the solution provided by the embodiment of the present application from the perspective of the interaction between various network elements. It can be understood that in order to realize the above functions, the above-mentioned first node, etc., includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm operations of each example described in the embodiments disclosed in this document, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0125] The embodiment of the present application can divide the functional modules of the first node according to the above method example. For example, each functional module can be divided corresponding to each function, or two or more functions can be integrated into one processing module. The above integrated module can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiment of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation.

[0126] For example, when the functional modules are divided in an integrated manner, Fig. 9 FIG. 9 is a schematic diagram showing the structure of a communication device 90. Exemplarily, the communication device 90 is used to implement the function of the first node. The communication device 90 is, for example, Figure 3 The embodiment shown, or Figure 8 The first node described in the embodiment shown.

[0127] The communication device 90 includes a transceiver module 901 and a processing module 902. Exemplarily, the communication device 90 may be a first node, or a chip applied to the first node, or other combined devices or components having the above-mentioned first node function. When the communication device 90 is the first node, the transceiver module 901 may be a transceiver, which may include an antenna and a radio frequency circuit, etc., and the processing module 902 may be a processor (or a processing circuit), such as a baseband processor, which may include one or more CPUs. When the communication device 90 is a component having the above-mentioned first node function, the transceiver module 901 may be a radio frequency unit, and the processing module 902 may be a processor (or a processing circuit), such as a baseband processor. When the communication device 90 is a chip system, the transceiver module 901 may be an input and output interface of a chip (such as a baseband chip), and the processing module 902 may be a processor (or a processing circuit) of the chip system, which may include one or more central processing units. It should be understood that the transceiver module 901 in the embodiment of the present application may be implemented by a transceiver or a transceiver-related circuit component, and the processing module 902 may be implemented by a processor or a processor-related circuit component (or, referred to as a processing circuit).

[0128] For example, the transceiver module 901 can be used to perform Figure 3 In the embodiment shown, all the sending and receiving operations performed by the first node, such as step 301, and / or other processes used to support the technology described herein. The processing module 902 can be used to perform Figure 3 In the illustrated embodiment, all operations except the sending and receiving operations performed by the first node, such as step 302, and / or other processes for supporting the technology described herein.

[0129] The transceiver module 901 is used to receive a first reception signal.

[0130] The processing module 902 is used to obtain a first transmission signal based on the first received signal and the target neural network model; wherein the first transmission signal is a signal that the second node wants to send to the communication device 90, and the target neural network model is used to obtain a signal sent by the second node corresponding to the signal received by the communication device according to the OQAM symbol corresponding to the signal received by the communication device 90.

[0131] A possible implementation method is that the first node obtains the first transmission signal according to the first received signal and the target neural network model, including: the first node obtains the first OQAM symbol corresponding to the first received signal according to the first received signal; the first node inputs the first OQAM symbol into the target neural network model to obtain the first transmission signal.

[0132] A possible implementation method further includes: the first node obtains an initial neural network model; the first node trains the initial neural network model based on OQAM symbols corresponding to a large number of received signals and a signal sent by the second node corresponding to each received signal to obtain a target neural network model.

[0133] In a possible implementation, the first node obtains the initial neural network model, including: the first node obtains the initial neural network model according to the modulation format of the first received signal and / or the transmission distance of the first received signal.

[0134] In a possible implementation, the target neural network model includes at least two target convolutional layers, at least one target residual network layer, a target flatten layer and a target fully connected layer FCL; at least two target convolutional layers and at least one target residual network layer are alternately connected in series, and among the at least two target convolutional layers, the target convolutional layer farthest from the input end of the target neural network model is connected to one end of the target flatten layer, and the other end of the target flatten layer is connected to the target FCL; the target residual network layer is used to reduce interference of a signal received by the first node under high-order modulation.

[0135] In one possible implementation, at least one target residual network layer includes at least one target convolutional layer and an adder, at least one target convolutional layer is connected to the adder, and the adder is used to add an input signal of the target residual network layer to an output signal of the at least one target convolutional layer.

[0136] When used to implement the function of the first node, for other functions that the communication device 90 can implement, refer to Figure 3 The embodiment shown, or Figure 8 The relevant introduction of the illustrated embodiment will not be repeated in detail.

[0137] Fig.10 A schematic diagram of the composition of a communication system is shown in FIG. Fig.10 As shown, the communication system 100 may include: a node 1001 and a node 1002. It should be noted that: Fig.10 This is only an exemplary figure, and the embodiments of this application are not limited to Fig.10 The communication system 100 shown includes network elements and the number of network elements.

[0138] Among them, node 1001 has the above Fig. 9 The function of the communication device 90 shown can be used to receive a first received signal, and obtain a first transmitted signal based on the first received signal and the target neural network model.

[0139] The node 1002 may be configured to send a first transmission signal.

[0140] It should be noted that all relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding network element of the communication system 100, and will not be repeated here.

[0141] Through the description of the above implementation methods, technical personnel in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In specific applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0142] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0143] The units described as separate components may or may not be physically separated, and the components shown as units may be one physical unit or multiple physical units, that is, they may be located in one place or distributed in multiple different places. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0144] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0145] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiment of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium, including several instructions to enable a device (which can be a single-chip microcomputer, chip, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard drives, ROM, RAM, magnetic disks or optical disks.

[0146] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A channel estimation and equalization method, characterized in that: The method comprises: The first node receives a first reception signal; The first node obtains a first transmission signal according to the first received signal and a target neural network model; Among them, the first transmitted signal is a signal that the second node wants to send to the first node, and the target neural network model is used to obtain the signal sent by the second node corresponding to the signal received by the first node according to the interleaved orthogonal amplitude modulation OQAM symbols corresponding to the signal received by the first node; the target neural network model includes at least two target convolutional layers, at least one target residual network layer, a target flatten layer and a target fully connected layer FCL; the target residual network layer is used to reduce the interference of the signal received by the first node under high-order modulation.

2. The method according to claim 1, characterized in that The first node obtains a first transmission signal according to the first received signal and the target neural network model, including: The first node acquires, according to the first received signal, a first OQAM symbol corresponding to the first received signal; The first node inputs the first OQAM symbol into the target neural network model to obtain the first transmit signal.

3. The method according to claim 1, characterized in that: The method further comprises: The first node obtains an initial neural network model; The first node trains the initial neural network model based on OQAM symbols corresponding to a large number of received signals and a signal sent by the second node corresponding to each of the received signals to obtain the target neural network model.

4. The method according to claim 3, characterized in that The first node obtains an initial neural network model, including: The first node obtains the initial neural network model according to a modulation format of the first received signal and / or a transmission distance of the first received signal.

5. The method according to any one of claims 1 to 4, characterized in that The at least two target convolutional layers and the at least one target residual network layer are alternately connected in series. Among the at least two target convolutional layers, the target convolutional layer farthest from the input end of the target neural network model is connected to one end of the target flatten layer, and the other end of the target flatten layer is connected to the target FCL.

6. The method according to claim 5, characterized in that The at least one target residual network layer includes at least one target convolutional layer and an adder, the at least one target convolutional layer is connected to the adder, and the adder is used to add an input signal of the target residual network layer to an output signal of the at least one target convolutional layer.

7. A communication device, characterized in that: The communication device comprises: a transceiver module and a processing module; The transceiver module is used to receive a first reception signal; The processing module is used to obtain a first transmission signal according to the first received signal and the target neural network model; Among them, the first transmitted signal is a signal that the second node wants to send to the communication device, and the target neural network model is used to obtain the signal sent by the second node corresponding to the signal received by the communication device according to the interleaved orthogonal amplitude modulation OQAM symbol corresponding to the signal received by the communication device; the target neural network model includes at least two target convolutional layers, at least one target residual network layer, a target flatten layer and a target fully connected layer FCL; the target residual network layer is used to reduce the interference of the signal received by the communication device under high-order modulation.

8. The communication device according to claim 7, characterized in that: The processing module is specifically configured to obtain, according to the first received signal, a first OQAM symbol corresponding to the first received signal; The processing module is also specifically used to input the first OQAM symbol into the target neural network model to obtain the first transmission signal.

9. The communication device according to claim 7, characterized in that: The processing module is also used to obtain an initial neural network model; The processing module is also used to train the initial neural network model based on OQAM symbols corresponding to a large number of received signals and the signal sent by the second node corresponding to each of the received signals to obtain the target neural network model.

10. The communication device according to claim 9, characterized in that: The processing module is specifically used to obtain the initial neural network model according to the modulation format of the first received signal and / or the transmission distance of the first received signal.

11. The communication device according to any one of claims 7 to 10, characterized in that: The at least two target convolutional layers and the at least one target residual network layer are alternately connected in series. Among the at least two target convolutional layers, the target convolutional layer farthest from the input end of the target neural network model is connected to one end of the target flatten layer, and the other end of the target flatten layer is connected to the target FCL.

12. The communication device according to claim 11, characterized in that: The at least one target residual network layer includes at least one target convolutional layer and an adder, the at least one target convolutional layer is connected to the adder, and the adder is used to add an input signal of the target residual network layer to an output signal of the at least one target convolutional layer.

13. A communication device, characterized in that: include: A processor, wherein the processor is coupled to a memory, wherein the memory is used to store programs or instructions, and when the programs or instructions are executed by the processor, the device executes the method according to any one of claims 1 to 6.

14. A chip, characterized in that: include: A processor, wherein the processor is coupled to a memory, wherein the memory is used to store programs or instructions, and when the programs or instructions are executed by the processor, the chip executes the method as claimed in any one of claims 1 to 6.

15. A computer readable medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instructions are executed, the computer is caused to perform the method according to any one of claims 1 to 6.

Images