A training method and device for an optical communication receiving model
By constructing and training an optical communication reception model, using Transformer encoder and deep fully connected network structure, the shortcomings of the existing optical reception algorithm in general and signal restoration capabilities are solved, efficient signal compensation and judgment are achieved, and reception is improved general performance and signal reduction rate of the machine.
Patent Information
- Application Number
- CN202210228714.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-03-09
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-03-09
AI Technical Summary
The existing optical reception algorithm lacks complete optimization and algorithm integration characteristics, which makes it difficult for various functional modules to coordinate, limiting the universality of the receiver and signal reduction capabilities, especially in the aspect of dispersion and nonlinear compensation, narrowband characteristics are insufficient.
Using a training method of optical communication reception model, by obtaining multiple optical fiber transmission signals, three-dimensional tensor data and two-dimensional label data are constructed, and Transformer encoder and deep fully connected network structure are used, and combined with the signal judgment layer, optical communication reception is trained. The model is used to optimize signal reduction capabilities.
It improves the general performance of the receiver and the high reduction rate of the source signal, and can effectively compensate for dispersion and nonlinear effects in different transmission scenarios, achieving accurate signal judgment and recovery.
Smart Images

Figure CN114722905B_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of optical fiber communication and machine learning technology, and in particular to a training method, restoration method, device, electronic device, storage medium and program product of an optical communication receiving model. Background Art
[0002] As an important research field in optical fiber communication, optical receiving algorithms play a vital role in the final extraction of optical information. In order to restore the transmission signal as accurately as possible, the optical receiving algorithm not only needs to compensate for the dispersion in optical fiber transmission, but also needs to equalize the nonlinearity therein. In addition, the optical receiving algorithm also has good anti-noise performance. Therefore, the research network of traditional optical receiving algorithms and their optimization is centered on the above-mentioned functional modules, and is studied and optimized in blocks. Although this research method is conducive to breaking down the problem into parts and optimizing it in blocks according to different tasks and scenarios, it lacks complete optimization and algorithm integration characteristics for the entire receiver system, resulting in the various functional module algorithms operating independently and difficult to coordinate. As for dispersion, the narrowband characteristics of traditional optical compensation or electrical compensation also limit the versatility of the receiver. Summary of the invention
[0003] In order to solve the problems existing in the prior art, the embodiments of the present invention provide a training method, restoration method, device, electronic device, storage medium and program product of an optical communication receiving model, which aims to improve the general performance of the receiver and the high restoration rate of the source signal.
[0004] A first aspect of the present disclosure provides a training method for an optical communication receiving model, comprising: obtaining N strings of source communication symbols and N optical fiber transmission signals; wherein N≥2, and the N optical fiber transmission signals are signals of each string of source communication symbols in the N strings of source communication symbols after being transmitted through different optical fiber distances; normalizing the N optical fiber transmission signals to obtain a normalized model input signal; constructing a three-dimensional tensor of the normalized model input signal and a two-dimensional tensor of the N strings of source communication symbols to obtain three-dimensional model input data and two-dimensional label data; and using part of the three-dimensional model input data and the two-dimensional label data as training data to train the optical communication receiving model, the optical communication receiving model being used to restore the optical fiber transmission signal to the corresponding source communication symbol.
[0005] Furthermore, part of the three-dimensional model input data and the two-dimensional label data are used as training data to train the optical communication receiving model, including: setting the initial training parameters of the optical communication receiving model, wherein the initial training parameters at least include a maximum iteration parameter threshold and a minimum loss function threshold; inputting part of the three-dimensional model input data and the two-dimensional label data into the optical communication receiving model for training; determining whether the number of generations of the current iteration is greater than the maximum iteration parameter threshold or less than the minimum loss function threshold; if so, ending the iteration and fixing the model parameters of the trained optical communication receiving model.
[0006] Furthermore, the method also includes: using another part of the three-dimensional model input data and the two-dimensional label data as test data to test the optical communication receiving model to obtain a tested optical communication receiving model; wherein the test data is data other than the training data in the three-dimensional model input data and the two-dimensional label data.
[0007] Furthermore, another part of the three-dimensional model input data and the two-dimensional label data are used as test data to test the optical communication receiving model to obtain the tested optical communication receiving model, including: using another part of the three-dimensional model input data and the two-dimensional label data as test data to test the optical communication receiving model; comparing and analyzing the results obtained by model training with another part of the two-dimensional label data to obtain the bit error rate of the optical communication receiving model training; and adjusting the initial training parameters of the optical communication receiving model according to the bit error rate to optimize the optical communication receiving model.
[0008] Furthermore, the optical communication receiving model includes: a Transformer encoder, including: a first residual structure, a first data processing structure, a second residual structure and a second data processing structure; wherein the first residual structure is used to copy the three-dimensional model input data into two identical three-dimensional optical fiber data; the first data processing structure is used to perform feature extraction on one copy of the three-dimensional model input data, and perform data aggregation on the extracted feature data and another copy of the three-dimensional model input data to obtain first aggregated data; the first residual structure is used to copy the first aggregated data into two identical optical fiber data; the second data processing structure is used to perform secondary feature extraction on one copy of the data output by the first residual structure, and perform secondary data aggregation on the extracted feature data and another copy of the data to obtain second aggregated data; a deep fully connected network structure is used to perform information fusion on the second aggregated data to obtain fused data; a signal decision layer is used to restore the fused data to the corresponding source communication symbol.
[0009] Furthermore, the method also includes: using a source generator to output source communication symbols; using a waveform generator to convert the source communication symbols into electrical signals; using a modulator to load the electrical signals onto an optical carrier to obtain a modulated initial optical fiber signal; using N optical fiber links with different transmission distances to transmit the modulated initial optical fiber signal to obtain N optical fiber transmission signals.
[0010] The second aspect of the present disclosure provides a method for restoring an optical fiber transmission signal, comprising: acquiring at least one optical fiber transmission signal; restoring the at least one optical fiber transmission signal into a corresponding source communication symbol using an optical communication receiving model, wherein the optical communication receiving model is trained using the training method for the optical communication receiving model provided by the first aspect of the present disclosure.
[0011] Furthermore, the restored communication symbol of at least one optical fiber transmission signal is consistent with its corresponding source communication symbol.
[0012] The third aspect of the present disclosure provides a training device for an optical communication receiving model, comprising: a signal acquisition module, used to acquire N strings of source communication symbols and N optical fiber transmission signals; wherein N≥2, and the N optical fiber transmission signals are signals of the N strings of source communication symbols after being transmitted through N different optical fiber distances; a signal preprocessing module, used to normalize the N optical fiber transmission signals to obtain a normalized model input signal; a data construction module, used to construct a three-dimensional tensor of the normalized model input signal and a two-dimensional tensor of the N strings of source communication symbols to obtain three-dimensional model input data and two-dimensional label data; and a model training module, used to use part of the three-dimensional model input data and the two-dimensional label data as training data to train the optical communication receiving model, the optical communication receiving model being used to restore the optical fiber transmission signal to the corresponding source communication symbol.
[0013] The fourth aspect of the present disclosure provides a device for restoring optical fiber transmission signals, comprising: a data acquisition module for acquiring at least one optical fiber transmission signal; a data restoration module for using an optical communication receiving model to restore at least one optical fiber transmission signal to a corresponding source communication symbol, wherein the optical communication receiving model is trained using the training method for the optical communication receiving model provided by the first aspect of the present disclosure.
[0014] The fifth aspect of the present disclosure provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method provided in the first aspect of the present disclosure or the method provided in the second aspect of the present disclosure is implemented.
[0015] The sixth aspect of the present disclosure provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method provided by the first aspect of the present disclosure or the method provided by the second aspect of the present disclosure is implemented.
[0016] The seventh aspect of the present disclosure provides a computer program product, including a computer program, which, when executed by a processor, implements the method provided by the first aspect of the present disclosure or the method provided by the second aspect of the present disclosure.
[0017] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] For a more complete understanding of the present disclosure and its advantages, reference will now be made to the following description taken in conjunction with the accompanying drawings, in which:
[0019] Figure 1 The schematic diagram of the structure of a conventional optical communication system is shown;
[0020] Figure 2 A flowchart of a method for training an optical communication receiving model according to an embodiment of the present disclosure is schematically shown;
[0021] Figure 3 A schematic diagram schematically shows three-dimensional model input data according to an embodiment of the present disclosure;
[0022] Figure 4 A schematic diagram schematically shows two-dimensional tag data according to an embodiment of the present disclosure;
[0023] Figure 5 The structure diagram of the optical communication receiving model according to an embodiment of the present disclosure is schematically shown;
[0024] Figure 6 A training flow chart of an optical communication receiving model according to an embodiment of the present disclosure is schematically shown;
[0025] Figure 7 A block diagram schematically shows a training device for an optical communication receiving model according to an embodiment of the present disclosure;
[0026] Figure 8 A block diagram schematically shows a device for restoring optical fiber transmission signals according to an embodiment of the present disclosure;
[0027] Fig. 9 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION
[0028] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.
[0029] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.
[0030] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.
[0031] In the case of using expressions such as "at least one of A, B, and C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.). In the case of using expressions such as "at least one of A, B, or C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, or C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).
[0032] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some boxes or combinations thereof in the block diagrams and / or flow charts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions can create a device for implementing the functions / operations described in these block diagrams and / or flow charts when executed by the processor. The technology of the present disclosure can be implemented in the form of hardware and / or software (including firmware, microcode, etc.). In addition, the technology of the present disclosure can take the form of a computer program product on a computer-readable storage medium storing instructions, which can be used by an instruction execution system or used in conjunction with an instruction execution system.
[0033] Figure 1 The schematic diagram of the structure of the traditional optical communication system is shown schematically. Figure 1 As shown in the figure, in a traditional optical communication system, the communication symbols sent by the source are mapped to waveform signals by an arbitrary waveform generator and then modulated on an optical carrier by a modulator. After the modulated optical signal is transmitted over a long distance in the optical fiber link, it is demodulated by a photoelectric detection module at the receiving end and input into a signal processing module to recover the source information.
[0034] In such Figure 1 In the traditional signal processing module marked by the black frame, due to the dispersion and nonlinear effects of optical fiber, aliasing occurs between signal symbols. Therefore, specific dispersion compensation algorithms and nonlinear equalization algorithms for different transmission scenarios are applied to this module to eliminate the signal's inter-symbol crosstalk. In addition, this module also needs to judge the signal after crosstalk elimination to restore the information transmitted by the source as accurately as possible.
[0035] The model / device obtained by the training method of the optical communication receiving model provided in the embodiment of the present disclosure is intended to compensate for the non-ideal factors of the signal, determine the code elements and finally restore the source information.
[0036] Figure 2 The flowchart of the training method of the optical communication receiving model according to the embodiment of the present disclosure is schematically shown. Figure 2 As shown, the method includes: steps S201 to S204.
[0037] In operation S201, N strings of source communication symbols and N optical fiber transmission signals are obtained, wherein N≥2, and the N optical fiber transmission signals are signals after each string of source communication symbols in the N strings of source communication symbols is transmitted over different optical fiber distances.
[0038] In the embodiments of the present disclosure, according to Figure 1The structure shown builds an optical fiber communication system. The transmission distance of the optical fiber link selects N different transmission distances, such as: 0.5km, 1km, 1.5km, 2km, ..., 99.5km and 100km. The symbol rates can be 10GBaud, 20GBaud and 40GBaud respectively. The sampling rate is K times the symbol rate. The modulation format can be OOK, PAM, etc.
[0039] Furthermore, if Figure 1 As shown, N strings of source communication symbols can be N strings of different or identical communication symbols sent by the source. The waveform generator is used to convert the N strings of source communication symbols into electrical signals, and then the electrical signals are loaded onto the optical carrier through the modulator to obtain the modulated initial optical fiber signal, and then the modulated initial optical fiber signal is transmitted using N optical fiber links with different transmission distances to obtain N optical fiber transmission signals. Following the above embodiment, the N strings of source communication symbols are processed in a series and then transmitted through optical fiber links of 0.5km, 1km, 1.5km, 2km, ..., 99.5km and 100km respectively. For example, the first string of source communication symbols in the N strings of source communication symbols is transmitted through a 0.5km optical fiber link, the second string of source communication symbols is transmitted through a 1km optical fiber link, and the third string of source communication symbols is transmitted through a 1.5km optical fiber link, and so on, to obtain N optical fiber transmission signals corresponding to the N strings of source communication symbols. Among them, N optical fiber transmission signals are used as the input data source for model training, and N strings of source communication symbols are used as label signals for subsequent model training and comparative analysis of training results.
[0040] It should be noted that the transmission distance range of the optical fiber link in the above embodiment is 0km to 100km and the different transmission distances are only exemplary descriptions, and do not constitute a limitation of the embodiments of the present disclosure. In actual application scenarios, when training or restoring signals in a long-distance optical communication system, the transmission distance of the optical fiber link can be 500km, 1000km, etc., and the signals obtained by the equally spaced transmission distances can be signals separated by 20km, 50km or 100km, etc.
[0041] In operation S202, normalization processing is performed on N optical fiber transmission signals to obtain a normalized model input signal.
[0042] In the embodiments of the present disclosure, before the signal is input into the model training, the signal needs to be preprocessed, including normalizing the signal, to facilitate the subsequent construction of tensors for the signal and generation of a model training data set.
[0043] Specifically, the N optical fiber transmission signals are complex signals, and the N optical fiber transmission signals are normalized to obtain a normalized model input signal, including: dividing the N optical fiber transmission signals by the maximum value of the corresponding module, and respectively obtaining the real part and the imaginary part of the N optical fiber transmission signals to obtain the normalized model input signal.
[0044] In operation S203, a three-dimensional tensor of a normalized model input signal and a two-dimensional tensor of N strings of source communication symbols are constructed to obtain three-dimensional model input data and two-dimensional label data.
[0045] In the embodiment of the present disclosure, for the model input signal, a three-dimensional tensor is constructed to satisfy the data format of the model training input data. Figure 3 As shown, the model input data is a three-dimensional tensor of m×n×2k, which can be specifically regarded as an m×n storage of each communication symbol V ij An m×k two-dimensional matrix, where the elements V ij is a vector with 2k elements, such as Figure 3 a1, a2, .., a k , b1, b2, ..., b k The matrix is a matrix of m rows and n columns, and the number of rows m and the number of columns n are determined by the scale of batch calculation of the model and the scale of neurons, respectively, which is not limited in the embodiments of the present disclosure.
[0046] like Figure 3 As shown, for each element V of the matrix ij The signal waveform corresponding to the communication symbol represented by it is specifically described by a vector of 2k elements, and the length parameter k of the description vector is determined by the sampling rate of the communication signal. In this vector, the elements 1 to k describe the real part of the sampled signal value, such as: a1, a2, ..., a k ; (k+1)~2k elements describe the imaginary part of the sampled signal value, such as b1, b2, ..., b k Although the model has the function of processing and judging received signals at different distances, the model’s input data does not contain transmission distance information because of its good distance generalization performance.
[0047] In the embodiment of the present disclosure, for the label data, a two-dimensional tensor is constructed to obtain two-dimensional label data. The data format of the two-dimensional label data is as follows: Figure 4 It should be noted that the output data format of the three-dimensional tensor data after training with the model provided by the present disclosure is the same as Figure 4 The data format shown is consistent, so that the model output results can be compared and analyzed with the label data to obtain the training results or test results of the model, and then error analysis can be performed.
[0048] Specifically, Figure 4 As shown, the physical meaning of the data format is the communication symbol obtained by judgment, which is an m×n matrix. The element c corresponding to the i-th row and j-th column of the matrix ij The corresponding waveform can be regarded as the element V in the i-th row and j-th column of the input data. ij The 2k sampled signal points are represented by . In digital communication systems, each element of the tag data generally takes a discrete value.
[0049] It should be noted that in the above embodiments, m, n, k, i, and j are all any positive integers, and i is less than or equal to m, and j is less than or equal to n. In addition, the examples of the real part, imaginary part, etc. in the above embodiments are only exemplary descriptions and do not constitute limitations on the embodiments of the present disclosure.
[0050] In operation S204, a portion of the three-dimensional model input data and the two-dimensional label data are used as training data to train an optical communication receiving model, wherein the trained optical communication receiving model is used to restore the optical fiber transmission signal to the corresponding source communication symbol.
[0051] In the embodiment of the present disclosure, the structure of the constructed optical communication receiving model is as follows: Figure 5 As shown in the figure, it specifically includes: Transformer encoding module, deep fully connected network structure and signal decision layer.
[0052] Specifically, the Transformer encoding module includes: a first residual structure, a first data processing structure, a second residual structure and a second data processing structure. The first residual structure is used to copy the three-dimensional model input data into two identical three-dimensional optical fiber data. The first data processing structure includes: a multi-head attention mechanism layer and an addition and normalization layer, wherein the multi-head attention mechanism layer is used to extract features from a three-dimensional model input data, specifically extracting multiple aliasing features of the aliased signal; the addition and normalization layer is used to aggregate the extracted feature data with another three-dimensional model input data to obtain first aggregated data. The second residual structure is used to copy the first aggregated data into two identical optical fiber data. The second data processing structure includes: a forward calculation structure, an addition and normalization layer, wherein the forward calculation structure is used to perform secondary feature extraction on a piece of data output by the second residual structure, and the addition and normalization layer is used to perform secondary data aggregation on the extracted feature data with another piece of data to obtain second aggregated data.
[0053] Furthermore, a deep fully connected network structure is used to compensate for signal aliasing caused by various dispersions and nonlinear effects during the transmission process. Specifically, it includes: a first fully connected layer, a first nonlinear activation layer, a second fully connected layer, a second nonlinear activation layer and a third fully connected layer. The three fully connected layers are used to deeply fuse the preliminary features extracted by the Transformer encoder using the connections between the neurons in each network. The deeper the number of layers, the higher the degree of feature fusion.
[0054] Furthermore, the signal decision layer is used to restore the fused data to the corresponding source communication symbols.
[0055] In the embodiments of the present disclosure, a deep fully connected network structure performs information fusion to better eliminate crosstalk between code elements and restore the signal before aliasing to the maximum extent. The setting of the residual network structure can effectively avoid the phenomenon of gradient disappearance during training. The signal decision layer at the end of the network model maps the signal after de-aliasing to the source communication information. The embodiments of the present disclosure are different from the traditional decision algorithm designed based on the optimal decision probability of matched filtering. The decision network can continuously optimize the decision threshold based on the learning of a large amount of data, so as to approach or even reach the performance of the optimal decision.
[0056] In the embodiment of the present disclosure, the three-dimensional model input data obtained in step S203 is matched with the two-dimensional label data to establish a total data set for model training, and then part of the data is selected from the total data set as the test set data to train the above-constructed model. Following the above embodiment, data with a transmission distance of 1km, 2km, 3km, ..., 100km in the total data set can be selected as a training data set for model training; data with a transmission distance of 0.5km, 1.5km, 2.5km, 3.5km, ..., 99.5km can be selected as a test data set for model testing.
[0057] It should be noted that the above division of the data set is only an exemplary description, and it can be other forms of data selection, which is not limited in the embodiments of the present disclosure.
[0058] According to the embodiments of the present disclosure, Figure 6 As shown, part of the three-dimensional model input data and the two-dimensional label data are used as training data to train the optical communication receiving model, which specifically includes: steps S601 to S606.
[0059] In operation S601, initial training parameters of the optical communication receiving model are set, wherein the initial training parameters at least include the number of multiple heads and nonlinear activation function, model training optimizer, learning rate, maximum iteration parameter threshold and minimum loss function threshold.
[0060] In the embodiments of the present disclosure, the initial training parameters of the model can be directly configured by the default configuration of the model or manually set according to the requirements of the data type and model accuracy to meet the accuracy requirements of the data training results in the actual application process. For example, the maximum iteration parameter threshold can take values of 100, 500, 1000 or larger / smaller values to prevent overfitting or underfitting of the model, so that the training results are as close to the actual results as possible.
[0061] In operation S602, part of the three-dimensional model input data and the two-dimensional label data are input into the optical communication receiving model for training.
[0062] In operation S603, it is determined whether the number of generations of the current iteration is greater than the maximum iteration parameter threshold or less than the minimum loss function threshold. If not, step S604 is executed; if yes, step S606 is executed.
[0063] In operation S604, a portion of data is randomly extracted from the training data set and input into the model for training, and the loss function value and the mean square error of the label are calculated point by point.
[0064] In the embodiments of the present disclosure, some data are randomly selected from the training data set and input into the model for training again. The loss function value after model training and the mean square error of the label are calculated point by point. At this time, the results after model training are compared and analyzed with the selected label data, and the mean square error is calculated to determine the accuracy of the model calculation results, which is convenient for model optimization.
[0065] In operation S605, the gradient of the loss function with respect to the trainable parameters of the network is calculated according to the numerical value of the loss function, and the parameters in the model such as the Q, K, V matrices in the Transformer encoding structure and the weights between the remaining structures are updated according to the training optimizer algorithm. The number of iteration parameters is increased by one, and the process returns to step S603.
[0066] In operation S606, the model training is completed, the iteration is ended, and the model parameters after the optical communication receiving model training are fixed.
[0067] In the embodiments of the present disclosure, the trained model is tested to determine the error rate of the trained model, which also includes a process of testing the model. Figure 6 As shown, in operation S607: after the model training is completed, the data in the test data is input into the model for testing, and the error between the final model and the traditional method is calculated, and the error result is output. Among them, the test data is the data other than the training data in the three-dimensional model input data and the two-dimensional label data, and step S607 is used to evaluate the training effect and generalization performance of the optical communication receiving model.
[0068] According to an embodiment of the present disclosure, another part of the three-dimensional model input data and the two-dimensional label data are used as test data to test the optical communication receiving model to obtain the tested optical communication receiving model, which specifically includes: using another part of the three-dimensional model input data and the two-dimensional label data as test data to test the optical communication receiving model; comparing and analyzing the result obtained by the model training with the two-dimensional label data corresponding to another part of the test data to obtain the bit error rate of the optical communication receiving model training; and adjusting the initial training parameters of the optical communication receiving model according to the bit error rate to optimize the optical communication receiving model, thereby improving the model performance.
[0069] It should be noted that the examples in the above embodiments are merely exemplary descriptions and do not constitute limitations of the embodiments of the present disclosure.
[0070] The training method of the optical communication receiving model provided by the embodiment of the present disclosure can gradually grasp the law of signal de-aliasing and optimal decision in different transmission scenarios under the training of data and gradient descent algorithm. After that, the model will be able to process transmission signals with different transmission lengths and signal-to-noise ratios, and recover the source information with a lower bit error rate.
[0071] Another embodiment of the present disclosure provides a method for restoring an optical fiber transmission signal, comprising: acquiring at least one optical fiber transmission signal; restoring the at least one optical fiber transmission signal into a corresponding source communication symbol using an optical communication receiving model, wherein the optical communication receiving model is trained using the method provided in the above embodiment.
[0072] In the embodiments of the present disclosure, the model trained by the method provided in the above embodiments is applied to Figure 1 The signal processing module is replaced in the optical communication system shown, thereby realizing rapid and accurate recovery of the information transmitted by the source from the optical fiber transmission signal.
[0073] In an embodiment of the present disclosure, the present disclosure provides a method for restoring optical fiber transmission signals, which can judge and restore optical fiber transmission signals in different scenarios to the source information before transmission with high accuracy. The restoration method effectively utilizes the generalization ability of neural networks to improve the general performance of traditional receivers, so that it can be used in different transmission scenarios.
[0074] Figure 7 A block diagram of a training device for an optical communication receiving model according to an embodiment of the present disclosure is schematically shown.
[0075] like Figure 7 As shown, the optical communication receiving model training device includes: a signal acquisition module 710, a signal preprocessing module 720, a data construction module 730 and a model training module 740. The device 700 can be used to implement reference Figure 2A training method for an optical communication reception model is described.
[0076] The signal acquisition module 710 is used to acquire N strings of source communication symbols and N optical fiber transmission signals; wherein N ≥ 2, and the N optical fiber transmission signals are signals after the N strings of source communication symbols are transmitted through N different optical fiber distances. The signal acquisition module 710 can be used, for example, to execute the above reference Figure 2 The described step S201 will not be repeated here.
[0077] The signal preprocessing module 720 is used to perform normalization processing on the N optical fiber transmission signals to obtain a normalized model input signal. The signal preprocessing module 720 can be used to perform the above reference Figure 2 The described step S202 will not be repeated here.
[0078] The data construction module 730 is used to construct a three-dimensional tensor of the normalized model input signal and a two-dimensional tensor of N strings of source communication symbols to obtain three-dimensional model input data and two-dimensional label data. The data construction module 730 can be used to execute the above reference Figure 2 The described step S203 will not be repeated here.
[0079] The model training module 740 is used to use part of the three-dimensional model input data and the two-dimensional label data as training data to train the optical communication receiving model, which is used to restore the optical fiber transmission signal to the corresponding source communication symbol. The model training module 740 can be used to perform the above reference Figure 2 The described step S204 will not be repeated here.
[0080] Figure 8 The block diagram schematically shows a device for restoring optical fiber transmission signals according to an embodiment of the present disclosure.
[0081] like Figure 8 As shown, the optical fiber transmission signal restoration device includes: a data acquisition module 810 and a data restoration module 820. The device 800 can be used to implement the optical fiber transmission signal restoration method described with reference to the above embodiments.
[0082] The data acquisition module 810 is used to acquire at least one optical fiber transmission signal.
[0083] The data restoration module 820 is used to restore at least one optical fiber transmission signal into a corresponding source communication symbol using an optical communication receiving model, wherein the optical communication receiving model is trained using the method shown in the above embodiment.
[0084] According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits, or at least part of the functions of any one of them can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), on-chip devices, devices on substrates, devices on packages, application-specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging circuits, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, submodules, units, and subunits can be at least partially implemented as computer program modules, and when the computer program modules are run, the corresponding functions can be performed.
[0085] For example, any of the signal acquisition module 710, the signal preprocessing module 720, the data construction module 730 and the model training module 740 or the data acquisition module 810 and the data restoration module 820 can be combined in one module for implementation, or any of the modules can be split into multiple modules. Alternatively, at least part of the functions of one or more of these modules can be combined with at least part of the functions of other modules and implemented in one module. According to an embodiment of the present disclosure, at least one of the signal acquisition module 710, the signal preprocessing module 720, the data construction module 730 and the model training module 740 or the data acquisition module 810 and the data restoration module 820 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), an on-chip device, a device on a substrate, a device on a package, an application-specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or by any one of the three implementation methods of software, hardware and firmware or by an appropriate combination of any of them. Alternatively, at least one of the signal acquisition module 710, the signal preprocessing module 720, the data construction module 730 and the model training module 740 or the data acquisition module 810 and the data restoration module 820 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is executed.
[0086] Fig. 9 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. Fig. 9The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.
[0087] like Fig. 9 As shown, the electronic device 900 described in this embodiment includes: a processor 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage part 908 into a random access memory (RAM) 903. The processor 901 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), and the like. The processor 901 may also include an onboard memory for caching purposes. The processor 901 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.
[0088] In RAM 903, various programs and data required for the operation of electronic device 900 are stored. Processor 901, ROM 902 and RAM 903 are connected to each other via bus 904. Processor 901 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 902 and / or RAM 903. It should be noted that the program can also be stored in one or more memories other than ROM 902 and RAM 903. Processor 901 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.
[0089] According to an embodiment of the present disclosure, the electronic device 900 may further include an input / output (I / O) interface 905, which is also connected to the bus 904. The electronic device 900 may further include one or more of the following components connected to the I / O interface 905: an input portion 906 including a keyboard, a mouse, etc.; an output portion 907 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 908 including a hard disk, etc.; and a communication portion 909 including a network interface card such as a LAN card, a modem, etc. The communication portion 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to the I / O interface 905 as needed. A removable medium 911, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 910 as needed, so that a computer program read therefrom is installed into the storage portion 908 as needed.
[0090] According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, the above-mentioned functions defined in the device of the embodiment of the present disclosure are executed. According to an embodiment of the present disclosure, the apparatus, equipment, device, module, unit, etc. described above can be implemented by a computer program module.
[0091] The embodiment of the present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / device described in the above embodiment; or may exist independently without being assembled into the device / apparatus / device. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the training method of the optical communication receiving model or the restoration method of the optical fiber transmission signal according to the embodiment of the present disclosure is implemented.
[0092] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, for example, may include but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In an embodiment of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution device, an apparatus or a device. For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 902 and / or RAM 903 described above and / or one or more memories other than ROM 902 and RAM 903.
[0093] The embodiment of the present disclosure also includes a computer program product, which includes a computer program, and the computer program contains program code for executing the method shown in the flowchart. When the computer program product is run in a computer device, the program code is used to enable the computer device to implement the training method of the optical communication receiving model or the restoration method of the optical fiber transmission signal provided in the embodiment of the present disclosure.
[0094] The above functions defined in the apparatus / device of the embodiment of the present disclosure are performed when the computer program is executed by the processor 901. According to the embodiment of the present disclosure, the apparatus, means, modules, units, etc. described above can be implemented by computer program modules.
[0095] In one embodiment, the computer program may be based on a tangible storage medium such as an optical storage device, a magnetic storage device, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of a signal on a network medium, and downloaded and installed through the communication part 909, and / or installed from a removable medium 911. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.
[0096] In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 909, and / or installed from the removable medium 911. When the computer program is executed by the processor 901, the above functions defined in the apparatus of the embodiment of the present disclosure are performed. According to the embodiment of the present disclosure, the apparatus, device, means, module, unit, etc. described above can be implemented by a computer program module.
[0097] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).
[0098] It should be noted that the functional modules in the various embodiments of the present disclosure may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention may essentially or in other words, the part that contributes to the prior art or all or part of the technical solution may be embodied in the form of a software product.
[0099] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the devices, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of boxes in the block diagram or flow chart, can be implemented with a dedicated hardware-based device that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0100] It will be appreciated by those skilled in the art that the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways, even if such combinations and / or combinations are not explicitly described in the present disclosure. In particular, the features described in the various embodiments and / or claims of the present disclosure may be combined and / or combined in a variety of ways without departing from the spirit and teachings of the present disclosure. All of these combinations and / or combinations fall within the scope of the present disclosure.
[0101] Although the present disclosure has been shown and described with reference to specific exemplary embodiments of the present disclosure, it should be understood by those skilled in the art that various changes in form and details may be made to the present disclosure without departing from the spirit and scope of the present disclosure as defined by the appended claims and their equivalents. Therefore, the scope of the present disclosure should not be limited to the above-mentioned embodiments, but should be determined not only by the appended claims, but also by the equivalents of the appended claims.
Claims
1. A training method for an optical communication receiving model, characterized in that: include: Obtaining N strings of source communication symbols and N optical fiber transmission signals; wherein N≥2, the N optical fiber transmission signals are signals after each string of source communication symbols in the N strings of source communication symbols is transmitted over different optical fiber distances; Normalizing the N optical fiber transmission signals to obtain a normalized model input signal; Constructing a three-dimensional tensor of the normalized model input signal and a two-dimensional tensor of the N strings of source communication symbols to obtain three-dimensional model input data and two-dimensional label data; and Using part of the three-dimensional model input data and the two-dimensional label data as training data to train an optical communication receiving model, wherein the optical communication receiving model is used to restore the optical fiber transmission signal to a corresponding source communication symbol; The optical communication receiving model comprises: The Transformer encoder comprises: a first residual structure, a first data processing structure, a second residual structure and a second data processing structure; wherein the first residual structure is used to copy the three-dimensional model input data into two identical three-dimensional optical fiber data; the first data processing structure is used to perform feature extraction on one copy of the three-dimensional model input data, and perform data aggregation on the extracted feature data and another copy of the three-dimensional model input data to obtain first aggregated data; the first residual structure is used to copy the first aggregated data into two identical optical fiber data; the second data processing structure is used to perform secondary feature extraction on one copy of the data output by the first residual structure, and perform secondary data aggregation on the extracted feature data and another copy of the data to obtain second aggregated data; A deep fully connected network structure, used for fusing information of the second aggregated data to obtain fused data; The signal decision layer is used to restore the fused data to the corresponding source communication symbols.
2. The training method of the optical communication receiving model according to claim 1, characterized in that: The using part of the three-dimensional model input data and the two-dimensional label data as training data to train the optical communication receiving model includes: Setting initial training parameters of the optical communication receiving model, wherein the initial training parameters at least include a maximum iteration parameter threshold and a minimum loss function threshold; Inputting part of the three-dimensional model input data and the two-dimensional label data into the optical communication receiving model for training; Determine whether the number of generations of the current iteration is greater than the maximum iteration parameter threshold or less than the minimum loss function threshold; If so, the iteration is terminated and the model parameters of the optical communication receiving model after training are fixed.
3. The training method of the optical communication receiving model according to claim 1, characterized in that: The method further includes: Use another part of the three-dimensional model input data and the two-dimensional label data as test data to test the optical communication receiving model to obtain the tested optical communication receiving model; wherein the test data is data other than the training data in the three-dimensional model input data and the two-dimensional label data.
4. The optical communication receiving model training method according to claim 3, characterized in that: The method of using another part of the three-dimensional model input data and the two-dimensional label data as test data to test the optical communication receiving model to obtain the optical communication receiving model after the test includes: Using another portion of the three-dimensional model input data and the two-dimensional label data as test data to test the optical communication receiving model; Comparing and analyzing the results obtained by the model training with another part of the two-dimensional label data, to obtain the bit error rate of the optical communication receiving model training; According to the bit error rate, initial training parameters of the optical communication receiving model are adjusted to optimize the optical communication receiving model.
5. The optical communication receiving model training method according to claim 1, characterized in that: The method further includes: Outputting source communication symbols using a source generator; Converting the source communication symbol into an electrical signal using a waveform generator; Using a modulator to load the electrical signal onto an optical carrier to obtain a modulated initial optical fiber signal; The modulated initial optical fiber signal is transmitted using N optical fiber links with different transmission distances to obtain the N optical fiber transmission signals.
6. A method for restoring optical fiber transmission signals, characterized in that: include: Acquiring at least one optical fiber transmission signal; The at least one optical fiber transmission signal is restored to a corresponding source communication symbol using an optical communication receiving model, wherein the optical communication receiving model is trained using the method according to any one of claims 1 to 5.
7. The method for restoring optical fiber transmission signals according to claim 6, characterized in that: The restored communication symbol of the at least one optical fiber transmission signal is consistent with its corresponding source communication symbol.
8. A training device for an optical communication receiving model, characterized in that: include: A signal acquisition module, used to acquire N strings of source communication symbols and N optical fiber transmission signals; wherein N≥2, the N optical fiber transmission signals are signals of the N strings of source communication symbols after being transmitted through N different optical fiber distances; A signal preprocessing module, used for normalizing the N optical fiber transmission signals to obtain a normalized model input signal; A data construction module, used to construct a three-dimensional tensor of the normalized model input signal and a two-dimensional tensor of the N strings of source communication symbols to obtain three-dimensional model input data and two-dimensional label data; and A model training module, used to use part of the three-dimensional model input data and the two-dimensional label data as training data to train an optical communication receiving model, wherein the optical communication receiving model is used to restore the optical fiber transmission signal to a corresponding source communication symbol; The optical communication receiving model comprises: The Transformer encoder comprises: a first residual structure, a first data processing structure, a second residual structure and a second data processing structure; wherein the first residual structure is used to copy the three-dimensional model input data into two identical three-dimensional optical fiber data; the first data processing structure is used to perform feature extraction on one copy of the three-dimensional model input data, and perform data aggregation on the extracted feature data and another copy of the three-dimensional model input data to obtain first aggregated data; the first residual structure is used to copy the first aggregated data into two identical optical fiber data; the second data processing structure is used to perform secondary feature extraction on one copy of the data output by the first residual structure, and perform secondary data aggregation on the extracted feature data and another copy of the data to obtain second aggregated data; A deep fully connected network structure, used for fusing information of the second aggregated data to obtain fused data; The signal decision layer is used to restore the fused data to the corresponding source communication symbols.
9. A device for restoring optical fiber transmission signals, characterized in that: include: A data acquisition module, used to acquire at least one optical fiber transmission signal; A data restoration module, used to restore the at least one optical fiber transmission signal to a corresponding source communication symbol using an optical communication receiving model, wherein the optical communication receiving model is trained using the method according to any one of claims 1 to 5.
10. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 6.
11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.
