Signal compensation method of optical fiber communication system

Abstract

The invention discloses a signal compensation method of an optical fiber communication system, which relates to the technical field of optical fiber communication and comprises the following steps: acquiring a baseband signal output after carrier phase recovery in a receiving end; a triple in the baseband signal is extracted, the triple is input into a target CLDNN network to obtain optical fiber nonlinear damage, the target CLDNN network is obtained by migrating a CLDNN network in a source domain to a target domain through migration learning, and the CLDNN network in the source domain is a network of a cascade CNN network, an LSTM network and a DNN network; and compensating the baseband signal through the nonlinear damage of the optical fiber. According to the technical scheme, the calculation complexity of the algorithm can be reduced, meanwhile, the high system transmission performance can be obtained under the low calculation complexity, and the compensation scheme can obtain high universality.

Description

Technical Field

[0001] This invention relates to the field of optical fiber communication technology, and more specifically to a signal compensation method for optical fiber communication systems. Background Technology

[0002] Optical communication is the cornerstone of modern information technology. In recent years, with the rapid growth of bandwidth-intensive services such as high-definition video streaming, the Internet of Things, and big data, optical networks are developing towards a more dynamic and complex direction. However, for high-speed, long-distance optical fiber transmission systems, optical signals are susceptible to various transmission impairments. Among them, non-linear impairments (NLI) have long been fundamental bottlenecks in long-distance optical transmission systems. Fortunately, due to the rapid development of DSP (Digital Signal Processing) technology in recent years, linear impairments can be adequately compensated for. Therefore, NLI has become the main limitation of long-distance optical fiber transmission systems.

[0003] To address nonlinear impairments in optical fibers, the DBP (Digital Backpropagation) algorithm effectively suppresses fiber nonlinearity, but it requires multiple FFT (Fast Fourier Transform) and IFFT (Inverse Fast Fourier Transform) processes, increasing the computational overhead and resulting in poor real-time performance. The VSNE (Volterra series nonlinear equalizer) algorithm suffers from unstable compensation performance due to tap coefficients and also incurs significant computational overhead. The PCTM (Phase Conjugated Twin Wave) algorithm sacrifices 50% spectral efficiency for compensation performance. The PNC (Perturbative Nonlinear Compensation) algorithm, due to its single-stage and symbol-rate processing, is considered a powerful and promising method. However, in practical applications, it involves numerous multiplications of perturbation terms, particularly in high-dispersion links, leading to substantial computational overhead.

[0004] In recent years, machine learning technology has been widely applied to various aspects of optical communication due to its powerful self-learning capabilities. Among them, deep learning algorithms have achieved good performance in non-linear compensation (NLI) in coherent fiber optic transmission systems. Based on perturbation-based equalization and deep neural networks (DNNs), the DNN-NLC algorithm (Deep Neural Network based Non-Linear Compensation) has been proposed. However, the DNN-NLC algorithm requires calculating perturbation coefficients to select the triplet with the largest contribution. As a feedforward neural network, DNN is a memoryless deep learning algorithm. To meet the feature requirements of the input layer, DNN needs to increase the dimension of the input layer, leading to higher computational complexity. To reduce computational complexity, an NLC scheme based on recurrent neural networks (RNNs) has been proposed. However, this algorithm has poor real-time performance, and the NLC performance is lower than that of the DBP algorithm. Summary of the Invention

[0005] The purpose of this invention is to provide a signal compensation method for optical fiber communication systems to solve the technical problems existing in related technologies.

[0006] To achieve the above objectives, the present invention provides a signal compensation method for an optical fiber communication system, comprising:

[0007] The baseband signal output from the receiving end after carrier phase recovery is obtained, wherein the receiving end is the optical signal receiving end in the optical fiber communication system;

[0008] Triples are extracted from the baseband signal and input into the target CLDNN network to obtain fiber nonlinear impairment. The target CLDNN network is obtained by transferring the CLDNN network in the source domain to the target domain through transfer learning. The target CLDNN network is a cascaded CNN network, LSTM network and DNN network.

[0009] Compensation for baseband signals is achieved through nonlinear damage to optical fibers.

[0010] Optionally, the extraction of triples from the baseband signal includes:

[0011] Calculate the correlation coefficient of the baseband signal, and determine the triplet index coefficient based on the correlation coefficient;

[0012] The fiber nonlinear perturbation coefficient is determined based on the triplet index coefficient;

[0013] Based on the triplet index coefficient, fiber nonlinear perturbation coefficient, and perturbation equalization algorithm, the baseband signal is calculated in... Transmission signals on polarization and in Transmission signals on polarization, for The complex conjugate of the transmitted signal on polarization and its sum in The conjugate complex numbers of the transmitted signals on the polarization sides are subjected to semi-degeneracy processing to obtain the results in... Triples in polarization and in Triples in polarization;

[0014] In Triples in polarization and in The triplet in polarization is determined as the triplet of the baseband signal.

[0015] Optionally, the calculation of the baseband signal based on triplet index coefficients, fiber nonlinear perturbation coefficients, and perturbation equalization algorithms... Transmission signals on polarization and in Transmission signals on polarization, for The complex conjugate of the transmitted signal on polarization and its sum in The conjugate complex numbers of the transmitted signals on the polarization sides are subjected to semi-degeneracy processing to obtain the results in... Triples in polarization and in A polarization triplet includes:

[0016] Calculate the triplet index coefficient in the baseband signal. The product of polarized transmitted signals and in The product of polarized transmitted signals will be... The polarization-biased transmission signal conjugate complex half-degenerates into a simplified form with QPSK notation and the origin, which will be... The polarization-biased transmitted signal conjugate complex number is half-degenerate into a simplified form with QPSK notation and the origin, resulting in... Triples in polarization and in A triplet in polarization.

[0017] Optionally, the fiber nonlinear damage is expressed by the following calculation formula:

[0018] ;

[0019] in, This refers to the nonlinear damage in the x-polarization direction of the optical fiber at time k. This refers to the nonlinear damage in the y-polarization direction of the optical fiber at time k. For transmission power, These are the fiber nonlinear perturbation coefficients after N-order quantization, where N is the order. The nonlinear perturbation coefficient of the optical fiber. and The index coefficients of the triplet. In the first The transmission symbol of polarization at time x. In the first The transmission symbol of polarization at time x. In the first The transmission symbol of y-polarization at time t, In the first The transmission symbol of y-polarization at time t, Logical operations that represent symbolic degeneration.

[0020] Optionally, the fiber nonlinear perturbation coefficient This can be expressed by the following calculation formula:

[0021] ;

[0022] in, The imaginary unit, For group velocity dispersion, For fiber nonlinearity coefficients, The pulse width. For transmission distance, For integration variables, It is a first-order exponential integral function. The symbol period.

[0023] Optionally, the triples of the baseband signal are expressed by the following formula:

[0024] ;

[0025] in, In order to be in Triples in polarization In order to be in A triplet in polarization.

[0026] Optionally, determining the triplet index coefficient based on the correlation coefficient includes:

[0027] The target range of triplet coefficients is determined by the correlation coefficient.

[0028] The triplet index coefficients are determined based on the target value range and the product constraint threshold of the triplet coefficients.

[0029] Optionally, the triplet index coefficient is determined by the following formula:

[0030] ;

[0031] in, The threshold for the product constraint. The upper limit of the target value range. and is the index coefficient of the triplet.

[0032] Optionally, the target CLDNN network is obtained by the following method:

[0033] The CLDNN network trained in the previous round is identified as the source domain in transfer learning. Through the multi-source domain transfer learning method, the CLDNN network trained in the previous round is transferred to the target domain of the current round of training to obtain the target CLDNN network.

[0034] The above technical solution acquires the baseband signal output after carrier phase recovery in an optical fiber communication system. Simultaneously, it extracts triples from the baseband signal and inputs these triples into a target CLDNN (Convolutional Long Short-Term Memory, Fully Connected Deep Neural Network) to obtain fiber nonlinear impairment. The target CLDNN is obtained by transferring a CLDNN from the source domain to the target domain through transfer learning. The source domain CLDNN is a cascaded network of CNN (Convolutional Neural Network), LSTM (Long Short-Term Memory Network), and DNN (Deep Neural Network). The baseband signal is compensated for through fiber nonlinear impairment. Obtaining fiber nonlinear impairment through equalized and simplified triples from the CLDNN network, and applying a portion of the trained parameters from the source domain to the target domain via transfer learning, reduces the computational complexity of the algorithm. This allows for higher system transmission performance with lower computational complexity and also enhances the universality of the compensation scheme.

[0035] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0036] Figure 1 This is a schematic diagram illustrating a signal compensation method for an optical fiber communication system according to an exemplary embodiment of the present invention.

[0037] Figure 2 This is a schematic diagram of an optical fiber communication system according to an exemplary embodiment of the present invention.

[0038] Figure 3 This is a schematic diagram illustrating the correlation coefficient of a signal sequence after carrier phase recovery according to an exemplary embodiment of the present invention.

[0039] Figure 4 This is a block diagram illustrating symbol degradation according to an exemplary embodiment of the present invention.

[0040] Figure 5 This is a structural diagram of the target CLDNN network according to an exemplary embodiment of the present invention. Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, so as to provide a better understanding of the concept of the present invention, the technical problem solved, the technical features constituting the technical solution, and the technical effects brought about.

[0042] like Figure 1 As shown, Figure 1 This is a schematic diagram illustrating a signal compensation method for an optical fiber communication system according to an exemplary embodiment of the present invention, with reference to... Figure 1 The method includes;

[0043] S101: Obtain the baseband signal output from the receiver after carrier phase recovery, wherein the receiver is the optical signal receiver in the optical fiber communication system;

[0044] S102: Extract the triplet from the baseband signal and input the triplet into the target CLDNN network to obtain the fiber nonlinear impairment. The target CLDNN network is obtained by transferring the CLDNN network in the source domain to the target domain through transfer learning, and the target CLDNN network is a cascaded CNN network, LSTM network and DNN network.

[0045] S103: Compensates for baseband signal through fiber nonlinear damage.

[0046] The above technical solution acquires the baseband signal output after carrier phase recovery in an optical fiber communication system. Simultaneously, it extracts triples from the baseband signal and inputs these triples into a target CLDNN network to obtain fiber nonlinear impairment. The target CLDNN network is obtained by transferring a CLDNN network from the source domain to the target domain through transfer learning. The source domain CLDNN network is a cascaded network of CNN, LSTM, and DNN. The baseband signal is compensated for through fiber nonlinear impairment. The fiber nonlinear impairment is obtained by equalizing and simplifying the triples in the CLDNN network. Applying a portion of the trained parameters from the source domain to the target domain through transfer learning reduces the computational complexity of the algorithm, achieving higher system transmission performance with lower computational complexity, and also making the compensation scheme more universal.

[0047] To enable those skilled in the art to better understand the signal compensation method for the optical fiber communication system provided by this invention, the above steps are illustrated in detail below.

[0048] For example, an optical fiber communication system can be a DP-16QAM (Dual-Polarization 16-Quadrature Amplitude Modulation) coherent optical fiber communication system. The receiver can be a device for receiving optical fiber signals transmitted in an optical fiber link. At the receiver, IQ (In-phase & Quadrature) imbalance compensation, dispersion compensation, polarization demultiplexing, frequency offset estimation, and carrier phase recovery processing can be performed on the received optical signal. Specifically, such as... Figure 2 As shown, at the transmitting end of the optical fiber communication system, the bit sequence consists of a length of 2... 16 The pseudo-random bit sequence is generated. First, a 30GBaud (30G baud rate) single-channel DP-16QAM symbol with root-raised cosine 0.01 Nyquist pulse shaping is generated using a 60Gsample / s (60G sampling rate) arbitrary waveform generator. Then, a polarization beam splitter splits the 1550nm optical carrier generated by the external cavity laser into two polarized beams, which are then modulated by corresponding IQ modulators to generate modulation signals. Finally, a polarization beam coupler combines the two modulation signals into a single optical signal, which is then amplified by an erbium-doped fiber amplifier and transmitted into the fiber optic link. On the fiber optic link side, the link is implemented through a fiber optic loop consisting of 100km of standard single-mode fiber, an erbium-doped fiber amplifier, a bandpass filter, and a loop controller. At the receiving end, the signal is first detected by a coherent receiver, then digitized by an 80GSa / s (80G sampling rate) oscilloscope, and finally subjected to IQ imbalance compensation, dispersion compensation, polarization demultiplexing, frequency offset estimation, and carrier phase recovery. For x-polarized in-phase components, For the y-polarized in-phase component, These are the orthogonal components of x-polarization. These are the orthogonal components of the y-polarization.

[0049] In this embodiment of the invention, the acquired baseband signal is the baseband signal output after carrier phase processing at the receiving end. Compared with the original input optical fiber signal, the baseband signal may have distortion or other abnormalities. Therefore, it is necessary to perform compensation processing on the baseband signal.

[0050] For example, a triplet can be a third-order product operation unit formed by combining dual-polarized light signals and their conjugate forms with different time delay orders in coherent optical communication nonlinear impairment equalization. After obtaining the baseband signal, triplets can be extracted from the baseband signal and then input into the target CLDNN network to obtain the fiber nonlinear impairment. Transfer learning can be a deep learning optimization method that transfers model knowledge trained in one task / scenario to another similar but different task / scenario.

[0051] In one possible manner, the extraction of triples from the baseband signal includes:

[0052] Calculate the correlation coefficient of the baseband signal, and determine the triplet index coefficient based on the correlation coefficient;

[0053] The fiber nonlinear perturbation coefficient is determined based on the triplet index coefficient;

[0054] Based on the triplet index coefficient, fiber nonlinear perturbation coefficient, and perturbation equalization algorithm, the baseband signal is calculated in... Transmission signals on polarization and in Transmission signals on polarization, for The complex conjugate of the transmitted signal on polarization and its sum in The conjugate complex numbers of the transmitted signals on the polarization sides are subjected to semi-degeneracy processing to obtain the results in... Triples in polarization and in Triples in polarization;

[0055] In Triples in polarization and in The triplet in polarization is determined as the triplet of the baseband signal.

[0056] It should be understood that the correlation coefficient of the baseband signal can be used as a quantitative indicator of the timing correlation of complex baseband IQ signals at different delay orders in optical communication. In this embodiment of the invention, the correlation coefficient can be calculated using the normalized autocorrelation coefficient method for complex baseband IQ signals in the prior art, which will not be elaborated here. After obtaining the correlation coefficient, the fiber nonlinear perturbation coefficient can be determined based on the correlation coefficient. The fiber nonlinear perturbation coefficient is a key parameter for quantifying the signal distortion intensity caused by nonlinear effects in the fiber. After obtaining the fiber nonlinear perturbation coefficient, the baseband signal can be calculated based on the triplet index coefficient, the fiber nonlinear perturbation coefficient, and the perturbation equalization algorithm. Transmission signals on polarization and in Transmission signals on polarization, for The complex conjugate of the transmitted signal on polarization and its sum in The conjugate complex numbers of the transmitted signals on the polarization sides are subjected to semi-degeneracy processing to obtain the results in... Triples in polarization and in The triplet in polarization. Among them, the perturbation equalization algorithm is a dedicated equalization algorithm designed for signal perturbations caused by fiber nonlinearity in the DSP of coherent optical communication receiver. The core logic is to first model / fit the amplitude and distribution of fiber nonlinearity perturbation, and then cancel the perturbation from the distorted received signal to restore the original transmitted signal.

[0057] Specifically, based on the Manakov equation, the dual-polarization transfer function of a single-mode fiber can be represented by the NLSE (Non-Linear Schrödinger Equation). According to the NLSE representation, based on the perturbation equalization algorithm, the in-channel fiber nonlinear distortion with dual polarization symbols is approximately expressed as:

[0058] ;

[0059] In the formula, This refers to the nonlinear damage in the x-polarization direction of the optical fiber at time k. This refers to the nonlinear damage of the fiber on the y-polarization at time k in the context of fiber nonlinear damage. It is the transmission power; In the first The transmission symbol of polarization at time x. In the first The transmission symbol for y-polarization at time; and They represent the first time. The conjugate of the transmission symbols for x-polarization and y-polarization at time; m and n are symbol indices; The nonlinear perturbation coefficient of the optical fiber is expressed by the following formula:

[0060] ;

[0061] in, The imaginary unit, For group velocity dispersion, For fiber nonlinearity coefficients, The pulse width. For transmission distance, For integration variables, It is a first-order exponential integral function. The sign period is given. At this point, the formula for calculating the triplet is:

[0062] ;

[0063] In the formula, and Let represent the triplet in x-polarization and the triplet in y-polarization at time k, respectively, without simplification.

[0064] In one possible manner, determining the triplet index coefficient based on the correlation coefficient includes:

[0065] The target range of triplet coefficients is determined by the correlation coefficient.

[0066] The triplet index coefficients are determined based on the target value range and the product constraint threshold of the triplet coefficients.

[0067] It should be understood that the target value range can be used to specify the upper and lower limits of the coefficients in the triplet index. The value of L depends on the size of the channel memory. If L is too large, it will introduce overfitting and low convergence; if it is too small, it will introduce underfitting. To reduce the complexity of triplet calculation, the triplet with the largest contribution is selected in this embodiment of the invention. To further reduce the complexity of triplet calculation, selection criteria for the index coefficients m and n in the triplet need to be established, as follows:

[0068] ;

[0069] in, It serves as a product constraint threshold and can be used to balance the performance and computational complexity of fiber nonlinear equalization. The upper limit of the target value range. and For the triple index coefficient. Specifically, as... Figure 3 As shown, L can be set to 15.

[0070] In one possible manner, the baseband signal is calculated based on triplet index coefficients, fiber nonlinear perturbation coefficients, and perturbation equalization algorithms. Transmission signals on polarization and in Transmission signals on polarization, for The complex conjugate of the transmitted signal on polarization and its sum in The conjugate complex numbers of the transmitted signals on the polarization sides are subjected to semi-degeneracy processing to obtain the results in... Triples in polarization and in A polarization triplet includes:

[0071] Calculate the triplet index coefficient in the baseband signal. The product of polarized transmitted signals and in The product of polarized transmitted signals will be... The polarization-biased transmitted signal conjugate complex half-degenerates into a simplified form of the QPSK symbol (Quadrature Phase Shift Keying Symbol) and the origin, which will be... The polarization-biased transmitted signal conjugate complex number is half-degenerate into a simplified form with QPSK notation and the origin, resulting in... Triples in polarization and in A triplet in polarization.

[0072] It should be understood that, in order to further reduce the computational complexity of triples, firstly, based on the symmetry of the triplet symbol indices m and n, the calculated time k is... The product of polarized transmitted signals and exist The product of polarized transmitted signals as well as Share it with other times, such as: The triples at time k and time k respectively contain and That is, a triplet of multiple signals contains Then, according to the semi-degeneracy theory, and The symbol degenerates into a QPSK-like symbol and origin. The symbol degeneration diagram is as follows: Figure 4 As shown. Symbols within the dashed lines are degenerated to the origin, and symbols outside the dashed lines are degenerated to the QPSK symbol of their respective quadrant. Therefore, logical operations can be used instead. , , as well as and and Complex multiplication operations are performed between them, where the QPSK symbol can be a basic signal unit of quaternary phase modulation, and the constellation diagram of the QPSK symbol is four symmetrical points on the two-dimensional IQ plane. At this time, in Triples in polarization and in The polarization triple can be expressed by the following formula:

[0073] ;

[0074] in, In order to be in Triples in polarization In order to be in A triplet in polarization.

[0075] At this point, the nonlinear damage to the optical fiber can be approximately expressed as:

[0076] ;

[0077] in, This refers to the nonlinear damage in the x-polarization direction of the optical fiber at time k. This refers to the nonlinear damage in the y-polarization direction of the optical fiber at time k. For transmission power, The nonlinear perturbation coefficient of the optical fiber after N-level quantization. The nonlinear perturbation coefficient of the optical fiber. and The index coefficients of the triplet. In the first The transmission symbol of polarization at time x. In the first The transmission symbol of polarization at time x. In the first The transmission symbol of y-polarization at time t, In the first The transmission symbol of y-polarization at time t, Logical operations that represent symbolic degeneration.

[0078] In one possible manner, the target CLDNN network is obtained by the following method:

[0079] The CLDNN network trained in the previous round is identified as the source domain in transfer learning. Through the multi-source domain transfer learning method, the CLDNN network trained in the previous round is transferred to the target domain of the current round of training to obtain the target CLDNN network.

[0080] It should be understood that the network structure diagram of the target CLDNN network is as follows: Figure 5 As shown, it includes a cascaded input layer, a convolutional network, a memory network, a tiling layer, and a linear layer. It can be the first Each signal transmission power, for The received symbol at that moment, for The received symbol at a given time. First, after carrier phase recovery, the triplet of the baseband signal is calculated; then, CNN, LSTM, and DNN networks are cascaded to construct a unified CLDNN network, and the calculated triplet is equalized to obtain the fiber nonlinear impairment. and Then, using transfer learning, the pre-trained CLDNN network is used as the source domain. A multi-source domain-based transfer learning approach is employed, applying a subset of pre-trained parameters from the source domain to the target domain. The target domain uses a small number of features to quickly reconstruct the target network, and fiber nonlinearity impairments are subtracted from the baseband signal. and This compensates for the coherent optical fiber communication system. Finally, the bit error rate (BRE) and Q-factor are calculated to obtain the performance of the fiber nonlinear equalization based on the TL-CLDNN algorithm.

[0081] Using the above technical solution, a single-channel DP-16QAM coherent optical fiber communication system is first constructed. After carrier phase recovery at the receiving end, the baseband signal is obtained. Triple selection, value sharing, and a semi-degenerate algorithm are then used to calculate the baseband signal. Triples in polarization and in The polarization triple; then the current time in Triples in polarization and in Polarization triplets are input into the target CLDNN network. After equalization of the target CLDNN network, the estimated fiber nonlinear impairment is obtained. The target CLDNN network can be obtained by transferring the CLDNN network from the source domain to the target domain through transfer learning. Multi-source domain-based transfer learning (TL) is employed to apply a subset of parameters trained in the source domain to the target domain, rapidly reconstructing the target network. Finally, this value is used to compensate for fiber nonlinear impairment in the coherent fiber communication system. This invention simplifies the calculation of triplets by utilizing triplet index scaling, value sharing, and sign degradation algorithms. Furthermore, it reduces the computational complexity of the experiment by applying a subset of parameters trained in the source domain to the target domain through transfer learning. This scheme not only achieves high system transmission performance with lower computational complexity but also makes the compensation scheme highly universal.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A signal compensation method for an optical fiber communication system, characterized in that, include: The baseband signal output from the receiving end after carrier phase recovery is obtained, wherein the receiving end is the optical signal receiving end in the optical fiber communication system; Triples are extracted from the baseband signal and input into the target CLDNN network to obtain fiber nonlinear impairment. The target CLDNN network is obtained by transferring the CLDNN network in the source domain to the target domain through transfer learning, and the target CLDNN network is a cascaded CNN network, LSTM network and DNN network. Compensation for baseband signals is achieved through nonlinear damage to optical fibers.

2. The signal compensation method for an optical fiber communication system according to claim 1, characterized in that, The extraction of triples from the baseband signal includes: Calculate the correlation coefficient of the baseband signal, and determine the triplet index coefficient based on the correlation coefficient; The fiber nonlinear perturbation coefficient is determined based on the triplet index coefficient; Based on the triplet index coefficient, fiber nonlinear perturbation coefficient, and perturbation equalization algorithm, the baseband signal is calculated in... Transmission signals on polarization and in Transmission signals on polarization, for The complex conjugate of the transmitted signal on polarization and its sum in The conjugate complex numbers of the transmitted signals on the polarization sides are subjected to semi-degeneracy processing to obtain the results in... Triples in polarization and in Triples in polarization; In Triples in polarization and in The triplet in polarization is determined as the triplet of the baseband signal.

3. The signal compensation method for an optical fiber communication system according to claim 2, characterized in that, The algorithm, based on triplet index coefficients, fiber nonlinear perturbation coefficients, and perturbation equalization, calculates the baseband signal at... Transmission signals on polarization and in Transmission signals on polarization, for The complex conjugate of the transmitted signal on polarization and its sum in The conjugate complex numbers of the transmitted signals on the polarization sides are subjected to semi-degeneracy processing to obtain the results in... Triples in polarization and in A polarization triplet includes: Calculate the triplet index coefficient in the baseband signal. The product of polarized transmitted signals and in The product of polarized transmitted signals will be... The polarization-biased transmission signal conjugate complex half-degenerates into a simplified form with QPSK notation and the origin, which will be... The polarization-biased transmitted signal conjugate complex number is half-degenerate into a simplified form with QPSK notation and the origin, resulting in... Triples in polarization and in A triplet in polarization.

4. The signal compensation method for an optical fiber communication system according to claim 3, characterized in that, The nonlinear damage to the optical fiber is expressed by the following calculation formula: ； in, This refers to the nonlinear damage in the x-polarization direction of the optical fiber at time k. This refers to the nonlinear damage in the y-polarization direction of the optical fiber at time k. For transmission power, These are the fiber nonlinear perturbation coefficients after N-order quantization, where N is the order. The nonlinear perturbation coefficient of the optical fiber. and The index coefficients of the triplet. In the first The transmission symbol of polarization at time x. In the first The transmission symbol of polarization at time x. In the first The transmission symbol of y-polarization at time t, In the first The transmission symbol of y-polarization at time t, Logical operations that represent symbolic degeneration.

5. The signal compensation method for an optical fiber communication system according to claim 4, characterized in that, The nonlinear perturbation coefficient of the optical fiber This can be expressed by the following calculation formula: ； in, The imaginary unit, For group velocity dispersion, For fiber nonlinearity coefficients, The pulse width. For transmission distance, For integration variables, It is a first-order exponential integral function. The symbol period.

6. The signal compensation method for an optical fiber communication system according to claim 4, characterized in that, The triplet of the baseband signal is expressed by the following formula: ； in, In order to be in Triples in polarization In order to be in A triplet in polarization.

7. The signal compensation method for an optical fiber communication system according to claim 2, characterized in that, The process of determining the triplet index coefficient based on the correlation coefficient includes: The target range of triplet coefficients is determined by the correlation coefficient. The triplet index coefficients are determined based on the target value range and the product constraint threshold of the triplet coefficients.

8. The signal compensation method for an optical fiber communication system according to claim 7, characterized in that, The triplet index coefficient is determined by the following formula: ； in, The threshold for the product constraint. The upper limit of the target value range. and is the index coefficient of the triplet.

9. The signal compensation method for an optical fiber communication system according to any one of claims 1-8, characterized in that, The target CLDNN network was obtained through the following method: The CLDNN network trained in the previous round is identified as the source domain in transfer learning. Through the multi-source domain transfer learning method, the CLDNN network trained in the previous round is transferred to the target domain of the current round of training to obtain the target CLDNN network.

Images