Short-distance high-speed optical fiber communication equalization method and system

Abstract

The invention provides a short-distance high-speed optical fiber communication equalization method and system, and the method comprises the steps: building a Gaussian mixture model for an optical signal subjected to the quadrature amplitude modulation of a set order, and completing the parameter initialization of the Gaussian mixture model through a particle swarm optimization algorithm of which the inertia weight is firstly increased and then decreased; the method comprises the following steps: inputting data features of two dimensions of an in-phase component and an orthogonal component of data into a Gaussian mixture model, then solving parameters of the Gaussian mixture model by using an expectation maximization algorithm, and determining the membership degree of each data point and a clustering label corresponding to each unary Gaussian distribution through the parameters; and through the maximum posterior probability, the data are distributed to the corresponding clustering labels, and nonlinear equalization is completed. According to the method, the memory property of signal nonlinear impairment can be accurately captured, the nonlinear clustering precision is improved, and the calculation complexity of a nonlinear clustering model is reduced under the same clustering effect.

Description

Technical Field

[0001] The present invention relates to the field of optical communication technology, and in particular to a short-distance high-speed optical fiber communication equalization method and system. Background Art

[0002] In the context of the global information age, the development of information digitization is inseparable from data transmission. As one of the important components of the transmission network, the fiber-optic communication system is facing increasing transmission capacity and transmission distance requirements. As a mature technology, single-mode optical fiber has been widely used in short-distance high-speed communications such as network center interconnection. In short-distance optical fiber communication, compensation and equalization for various damages have always been a research hotspot. Various damages in high-speed coherent optical communication systems seriously inhibit the transmission performance of the system. Since both optoelectronic devices and optical fiber transmission links have certain nonlinear effects, when the number of devices is large, the nonlinear effects they produce are superimposed on each other, which will cause nonlinear phase shift of the transmission signal and crosstalk between code elements, thereby seriously affecting the signal quality. How to deal with nonlinear damage is a major problem that cannot be avoided in the implementation of high-speed coherent optical communication systems.

[0003] The prior art proposes many nonlinear equalization methods for receiving end digital signal processors (DSPs), including digital back propagation (DBP), Volterra series approximation, and nonlinear equalization based on perturbation. The above nonlinear equalization methods have been widely studied for their effectiveness in eliminating the system, but in high-speed communication systems, these methods have low performance and huge complexity. With the development of machine learning, the application of a large number of machine learning algorithms has provided a new route for compensating nonlinear damage in optical fiber communication systems. Among them, supervised learning algorithms such as support vector machines, support vector regression, and K-means, as well as unsupervised learning algorithms such as k-nearest neighbors and expectation maximization (EM) have attracted much attention due to their good equalization capabilities. In high-speed communication systems, due to the influence of nonlinearity, pulse broadening occurs during the transmission process of signals, and multiple adjacent signals will affect each other, that is, memory effects appear and become time series. Compared with algorithms based on Volterra functions and digital back propagation, machine learning algorithms do not require parameters of optical fiber transmission links and have better flexibility and adaptability. In addition, unsupervised learning algorithms do not require training sequence assistance and have the characteristics of low data redundancy and low computational complexity. Nonlinear equalizers for this algorithm have received great attention.

[0004] The use of polarization multiplexing technology in single-mode optical fiber can improve transmission capacity. However, in short-distance high-speed transmission systems, the nonlinearity of the optical fiber and the nonlinearity of the receiving device are superimposed, making subsequent compensation much more difficult. Conventional electrical domain equalization algorithms cannot completely eliminate these mixed interferences. Therefore, compensation technology for nonlinear damage in optical fiber communication systems also needs to consider the superposition and timing correlation of damage to maximize system performance, which is also a part neglected in existing technologies. Summary of the invention

[0005] In view of this, an embodiment of the present invention provides a short-distance high-speed optical fiber communication equalization method and system to eliminate or improve one or more defects existing in the prior art, and to solve the problem that the prior art causes high complexity of the nonlinear clustering model and difficulty in eliminating mixed interference due to nonlinear superposition of optical fibers and receiving end devices in short-distance high-speed transmission systems.

[0006] One aspect of the present invention provides a short-distance high-speed optical fiber communication equalization method, the method is used to be performed at a receiving end in a single-mode optical fiber communication system, and the method comprises the following steps:

[0007] The optical signal of the transmitting end is received after orthogonal amplitude modulation of a set order, and after polarization splitting, photoelectric detection and four-channel digital sampling are performed on the optical signal, a linear equalization process is performed using a preset method to obtain a complete data set in two polarization directions; each symbol in the symbol sequence of the data set expresses the real part and the imaginary part of the data point based on a two-dimensional vector;

[0008] For the complete data set, using a particle swarm optimization algorithm to establish initial parameters of a Gaussian mixture model, the Gaussian mixture model comprising a plurality of Gaussian distributions, each Gaussian distribution corresponding to the set order to describe the distribution of each constellation point;

[0009] Iterating the Gaussian mixture model under the initial parameter state by using an expectation maximization algorithm until the parameters converge;

[0010] Based on the Gaussian mixture model after parameter convergence, the posterior probability that each data point belongs to each Gaussian distribution is calculated as the membership degree, and clustering is performed according to the membership degree to divide the clusters into the same cluster, thereby completing nonlinear equalization.

[0011] In some embodiments, a linear equalization process is performed using a preset method, including:

[0012] The sampled data is subjected to I / Q orthogonalization, low-pass filtering, Gardner time synchronization, dispersion compensation, frequency offset compensation and phase offset compensation.

[0013] In some embodiments, for the complete data set, initial parameters of the Gaussian mixture model are established using a particle swarm optimization algorithm, including:

[0014] Initialize a particle swarm, each particle representing a solution to the Gaussian mixture model parameters;

[0015] In each iteration, the fitness value of each particle is calculated and recorded; the individual extreme value and the group extreme value are determined based on the fitness value of each particle; the speed and position of each particle itself are updated according to the position and speed corresponding to the individual extreme value and the group extreme value;

[0016] When the set number of iterations is reached or the group extremum reaches the set value, the iteration is stopped, and the solution of the particle corresponding to the group extremum is output as the initial parameter of the Gaussian mixture model.

[0017] In some embodiments, the fitness value is calculated using the log-likelihood function of the Gaussian mixture model, and the calculation formula is:

[0018]

[0019] Among them, α k represents the mixing coefficient of the kth Gaussian distribution, μ k represents the mean of the kth Gaussian distribution, represents the covariance of the kth Gaussian distribution, represents the probability density of the kth Gaussian distribution; N represents the number of data points, x i represents the i-th data point; K represents the number of Gaussian distributions;

[0020] In updating the speed and position of each particle according to the position and speed corresponding to the individual extreme value and the group extreme value, the speed update formula is:

[0021]

[0022] Among them, w is the inertia weight; k is the current iteration number, c1 is the individual learning factor, c2 is the social learning factor; r1 and r2 are random numbers distributed between [0,1]; represents the position of particle i at the current k-th iteration; represents the individual optimal position of particle i up to the current iteration round; The global optimal position of the entire particle up to the current iteration round; represents the velocity of particle i at the k+1th iteration, represents the velocity of particle i at the kth iteration;

[0023] The particle position update formula is:

[0024]

[0025] in, Represents the position of particle i in the next iteration round; is the position of particle i in the current iteration round, is the velocity of particle i in the next iteration round.

[0026] In some embodiments, the Gaussian mixture model under the initial parameter state is iterated using an expectation maximization algorithm until the parameters converge, including:

[0027] Define the hidden variable z ik , reflecting the data point x i From the kth Gaussian distribution, the expression is:

[0028]

[0029] Wherein, i = 1, 2, ..., N; k = 1, 2, ..., K; the latent variables and the data points constitute complete data, denoted as (x i , z i1 , z i2 , …, z iK )i=1,2,…,N;

[0030] The likelihood function expression of the Gaussian mixture model is:

[0031]

[0032] in, μ k represents the mean of the kth Gaussian distribution, represents the covariance of the kth Gaussian distribution; n k Represents the number of N observation points from the kth Gaussian distribution;

[0033] Then the log-likelihood function of the Gaussian mixture model is:

[0034]

[0035] In the E step, the expectation maximization algorithm calculates each data point x according to the current parameters. i The posterior probability of belonging to the kth Gaussian distribution Note θ 0 is the initial value of the iterative parameter, denoted by θ t is the estimated value of the parameter θ in the t-th iteration. By taking the expectation of the log-likelihood function, we get the Q function:

[0036] Q(θ,θ t ) = E(logP(x, x|θ), θt );

[0037]

[0038] Wherein, i=1, 2, ..., N; k=1, 2, ..., K;

[0039] In step M, find the value of θ when the Q function reaches its maximum value, as θ t+1 Participate in the next iteration, the expression is:

[0040]

[0041] Wherein, θ contains the mixing coefficient α of each Gaussian distribution in the Gaussian mixture model k , mean μ k and covariance The expression is:

[0042] θ=(α1,α2,...,α K ; μ1, μ2, ..., μ K ; σ1, σ2, ..., σ K );

[0043] Then for the mixing coefficient α k , mean θ k and covariance The estimate is:

[0044]

[0045] The E step and the M step are repeated until the parameters converge.

[0046] In some embodiments, the transmitting end in the method modulates the optical signal using 16-QAM, 64-QAM or 256-QAM.

[0047] On the other hand, the present invention also provides a short-distance high-speed optical fiber communication system, the system comprising:

[0048] A transmitting end, the transmitting end comprising a transmitting end digital signal processor, an arbitrary waveform generator, a digital-to-analog converter and an optical modulator connected in sequence, the optical modulator being provided with a light source by a first local oscillator laser; the transmitting end is used to generate an optical signal subjected to orthogonal amplitude modulation of a set order;

[0049] an optical amplifier, used to amplify the optical signal output by the optical modulator;

[0050] A single-mode optical fiber, used for transmitting the amplified optical signal;

[0051] The receiving end includes an optical mixer, a photodiode, a sampling oscilloscope and a receiving end digital signal processor connected in sequence; the optical mixer is connected to a second local oscillator laser; and the receiving end executes the above short-distance high-speed optical fiber communication equalization method.

[0052] In some embodiments, the fiber amplifier is an erbium-doped fiber amplifier.

[0053] On the other hand, the present invention further provides a computer-readable storage medium having a computer program / instruction stored thereon, which implements the steps of the above method when the computer program / instruction is executed by a processor.

[0054] On the other hand, the present invention also provides a computer program product, comprising a computer program / instruction, which implements the steps of the above method when executed by a processor.

[0055] The beneficial effects of the present invention are at least:

[0056] The short-distance high-speed optical fiber communication equalization method and system of the present invention constructs a Gaussian mixture model for an optical signal modulated by orthogonal amplitude modulation of a set order, first uses a particle swarm optimization (PSO) algorithm with an inertia weight that increases first and then decreases to complete the parameter initialization of the Gaussian mixture model, inputs the data features of two dimensions of the in-phase component and the orthogonal component of the data into the Gaussian mixture model, and then uses the expectation maximization (EM) algorithm to solve the parameters of the Gaussian mixture model, determines the membership of each data point with the cluster label corresponding to each univariate Gaussian distribution through the parameters, and assigns it to the corresponding cluster label through the maximum a posteriori probability to complete nonlinear equalization. The present invention can accurately capture the memory properties of nonlinear damage of the signal, improve the accuracy of nonlinear clustering, and reduce the computational complexity of the nonlinear clustering model under the same clustering effect.

[0057] Additional advantages, purposes, and features of the present invention will be described in part in the following description, and will become apparent to those skilled in the art after studying the following, or may be learned from the practice of the present invention. The purposes and other advantages of the present invention may be achieved and obtained by the structures specifically indicated in the specification and the accompanying drawings.

[0058] Those skilled in the art will appreciate that the objectives and advantages that can be achieved with the present invention are not limited to the above specific description, and the above and other objectives that can be achieved by the present invention will be more clearly understood from the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The drawings described herein are used to provide a further understanding of the present invention, constitute a part of the present application, and do not constitute a limitation of the present invention. In the drawings:

[0060] Figure 1The figure is a flow chart of a short-distance high-speed optical fiber communication equalization method according to an embodiment of the present invention.

[0061] Figure 2 This is a logical schematic diagram of a short-distance high-speed optical fiber communication equalization method according to another embodiment of the present invention.

[0062] Figure 3 It is a schematic structural diagram of a short-distance high-speed optical fiber communication system according to another embodiment of the present invention.

[0063] Figure 4 This is a graph showing the variation of bit error rate with receiving end optical power when clustering operation is performed on optical signals based on the PSO-GMM model according to an embodiment of the present invention.

[0064] Figure 5 This is a graph showing how the quality factor varies with the optical power at the receiving end when clustering analysis of optical signals is performed based on the PSO-GMM model, K-Means, DBSCAN and KNN according to an embodiment of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments and the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0066] It should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only structures and / or processing steps closely related to the solutions according to the present invention are shown in the accompanying drawings, while other details that are not closely related to the present invention are omitted.

[0067] It should be emphasized that the term “include / comprises” when used herein refers to the presence of features, elements, steps or components, but does not exclude the presence or addition of one or more other features, elements, steps or components.

[0068] It should also be noted that, unless otherwise specified, the term “connection” herein may refer not only to a direct connection but also to an indirect connection involving an intermediate.

[0069] During the transmission process, optical signals will produce nonlinear phase shift, inter-symbol interference (ISI) and other problems due to nonlinear interference between optical fibers and receiving ends, which will affect the quality of the transmitted signal. In particular, the "memory effect" that occurs during high-speed and large-capacity transmission exacerbates the nonlinear distortion. Traditional nonlinear equalization methods have limitations in performance or computational complexity and are difficult to meet the needs of high-speed communication. To this end, the present invention proposes to use a PSO-GMM clustering model and utilize a particle swarm optimization (PSO) algorithm to improve the convergence speed of a Gaussian mixture model (GMM) to more accurately capture the nonlinear characteristics of the signal and achieve efficient nonlinear equalization, thereby improving the transmission performance of the system.

[0070] Specifically, one aspect of the present invention provides a short-distance high-speed optical fiber communication equalization method, the method is used to be performed at a receiving end in a single-mode optical fiber communication system, such as Figure 1 and 2 As shown, the method includes the following steps S101 to S104:

[0071] Step S101: receiving an optical signal from a transmitter that has undergone orthogonal amplitude modulation of a set order, performing polarization splitting, photoelectric detection, and four-channel digital sampling on the optical signal, and then performing linear equalization processing using a preset method to obtain a complete data set in two polarization directions; each symbol in the data set symbol sequence expresses the real part and imaginary part of a data point based on a two-dimensional vector.

[0072] Step S102: for the complete data set, the initial parameters of the Gaussian mixture model are established using the particle swarm optimization algorithm. The Gaussian mixture model includes multiple Gaussian distributions, each Gaussian distribution corresponds to a set order to describe the distribution of each constellation point.

[0073] Step S103: using the expectation maximization algorithm to iterate the Gaussian mixture model under the initial parameter state until the parameters converge.

[0074] Step S104: Based on the Gaussian mixture model after parameter convergence, the posterior probability of each data point belonging to each Gaussian distribution is calculated as the membership degree, and clustering is performed according to the membership degree to divide the clusters into the same clusters, thereby completing nonlinear equalization.

[0075] In step S101, the optical signal processed by the present invention is orthogonal amplitude modulated based on a set order at the transmitting end. Quadrature Amplitude Modulation QAM (Quadrature Amplitude Modulation): It is a modulation method that uses amplitude modulation and phase modulation at the same time. By modulating different information on two orthogonal carriers (one is a cosine wave and the other is a sine wave), more data can be transmitted simultaneously. In QAM, a signal consists of two orthogonal components, which are usually represented as I (In-phase) and Q (Quadrature-phase). The I component modulates the amplitude of one carrier, and the Q component modulates the amplitude of another carrier that is orthogonal to the I component. Specifically, the present invention can be processed for 16-QAM, 64-QAM or 256-QAM schemes.

[0076] For example, in 64-QAM, there are 64 possible symbols (signal points), so it can transmit 6 bits of data in each symbol period, 2 6 =64. These 64 symbols are distributed in a two-dimensional coordinate system, where the horizontal axis represents the I component and the vertical axis represents the Q component. The constellation diagram of 64-QAM consists of 64 points, each of which represents a different symbol, and each symbol corresponds to 6 bits of data. The coordinates of each point in the constellation diagram are determined by I (horizontal axis) and Q (vertical axis).

[0077] The transmitter performs 64-QAM modulation on the original 0-1 bit data. Each constellation point of the signal constellation diagram represents a transmission code element, corresponding to 64 categories, numbered from 1 to 64 as cluster labels. The transmitter processes the modulated data through the arbitrary waveform generator AWG, the electrical amplifier EA, and the Mach-Zehnder modulator MZM to form an optical signal, which is then divided into two beams and enters the spatial light modulator, and the two beams are combined to enter the optical fiber transmission. At the receiving end of the coherent optical communication system, the photoelectric conversion is performed through the polarization beam splitter PBS and the photodetectors Photodetectors, and the signal is digitized using a four-channel digital sampling oscilloscope. The obtained original electrical signal is subjected to linear equalization processing, which includes row I / Q orthogonalization, low-pass filtering, Gardner time synchronization, dispersion compensation, frequency offset compensation, and phase offset compensation.

[0078] I / Q orthogonalization decomposes the received signal into two orthogonal components, namely the in-phase (I) component and the quadrature (Q) component. In the modulation system, the I / Q orthogonal components can carry different information respectively. By separating the received signal into I and Q components, the modulation characteristics of each component can be better processed to prepare for subsequent demodulation. Low-pass filtering removes high-frequency noise and interference in the signal. Since various high-frequency noises are introduced in optical signal transmission, low-pass filtering can retain the useful frequency components in the signal while eliminating unnecessary high-frequency noise, thereby improving the signal-to-noise ratio (SNR) of the signal and reducing the impact of noise on signal demodulation. Gardner time synchronization is used to achieve symbol-level time synchronization at the receiving end to ensure that the sampling time is aligned with the optimal position of the symbol. Dispersion compensation is used to compensate for the signal broadening or distortion caused by the fiber dispersion effect during transmission. Fiber dispersion causes signals of different frequency components to propagate at different speeds in the fiber, causing the signal pulse to be broadened and distorted. The dispersion compensation algorithm can correct this distortion and restore the original form of the signal, thereby reducing the inter-symbol interference (ISI) caused by broadening. Frequency offset compensation is used to correct the deviation between the carrier frequency of the receiving end and the carrier frequency of the transmitting end. At the receiving end, the received signal may cause phase and spectrum changes due to frequency offset, resulting in demodulation errors. Frequency offset compensation can correct this deviation so that the carrier frequency of the receiving end is consistent with that of the transmitting end, ensuring the accuracy of signal demodulation. Phase offset compensation is used to adjust the signal phase to align it with the phase of the transmitting end. Due to phase noise or other influences during transmission, the phase of the received signal may drift, resulting in demodulation errors. Phase offset compensation can eliminate this phase drift, allowing the receiving end to accurately restore the phase information of the transmitting end, thereby correctly decoding the signal content.

[0079] For the optical signal modulated by 64-QAM at the transmitting end, after linear uniform processing, a 64-QAM symbol sequence s = [s1, s2, ..., s N ], where the vector s i (i=1,2,…,N) represents the i-th symbol of a 64-QAM signal sequence in one polarization direction, s i =[I i ;Q i ] is a two-dimensional vector, I i and Q i Represent the real and imaginary data of the M-QAM signal respectively. The real and imaginary data in one polarization direction are taken as a group of signal points, with a total of N groups of signal points. The data in the other polarization direction are processed in the same way to obtain a complete data set X for the two polarization directions. in and Y in .

[0080] In step S102, the present invention captures the characteristics of nonlinear distortion in the optical fiber communication system through statistical modeling and cluster analysis, and classifies and balances the signal at the receiving end to offset the influence of nonlinear effects. In step S101, the characteristic data of each data point is extracted by linear equalization processing, such as the in-phase (I) and quadrature (Q) components of the 64-QAM modulated signal. These data points reflect the distribution of the signal in the current transmission state, but due to nonlinear damage, they deviate from the ideal constellation point. For each constellation point of the set order QAM, it forms a unique distribution feature during the signal transmission process, so it can be achieved by establishing a Gaussian mixture model and clustering analysis to classify and balance the signal. The present invention processes different polarization directions of the signal separately to obtain a complete nonlinear equalization data set. Further, the Gaussian mixture model (GMM) represents the complex signal data distribution as a mixture of multiple Gaussian distributions. Since the signal distribution at the receiving end becomes complicated due to nonlinear distortion, the GMM can approximate these complex distributions through multiple Gaussian components to capture different modes in the signal data. Through the GMM model, the signal data can be divided into different cluster labels, which correspond to different signal states (or different nonlinear distortion modes). In this way, the actual offset state of the received signal can be assigned to the "ideal" distribution state that is closest to it, thereby achieving error correction.

[0081] In step S102, the present invention uses a particle swarm optimization (PSO) algorithm to initialize the GMM parameters. As a global optimization algorithm, PSO finds better parameter values ​​through swarm search, so that the GMM model converges faster and reduces the number of iterations.

[0082] The probability density function (PDF) of a univariate Gaussian distribution is as follows:

[0083]

[0084] Among them, μ and σ 2 are the mean and variance of the Gaussian distribution respectively.

[0085] The Gaussian mixture model is a model composed of multiple univariate Gaussian models, with the following probability distribution:

[0086]

[0087] Among them, K is the number of single Gaussian models in the mixed model; α k is the weight coefficient of each single Gaussian model, satisfying α k ≥0 and φ(x|θ k ) represents the probability density of the k-th single Gaussian model, called the k-th sub-model; θ krepresents the parameters (mean and variance) of the kth single Gaussian model, that is

[0088] Exemplarily, for 64-QAM modulation, the value of K is 64, and each univariate Gaussian distribution corresponds to the distribution characteristics of a constellation point.

[0089] In some embodiments, for the complete data set, initial parameters of the Gaussian mixture model are established using a particle swarm optimization algorithm, including steps S201 to S203:

[0090] Step S201: Initialize a particle swarm (PSO), where each particle represents a solution to the Gaussian mixture model parameters.

[0091] Step S202: In each iteration, the fitness value of each particle is calculated and recorded; the individual extreme value and the group extreme value are determined based on the fitness value of each particle; and the speed and position of each particle are updated according to the position and speed corresponding to the individual extreme value and the group extreme value.

[0092] Step S203: When the set number of iterations is reached or the population extremum reaches the set value, the iteration is stopped, and the solution of the particle corresponding to the population extremum is output as the initial parameter of the Gaussian mixture model.

[0093] In some embodiments, the fitness value in step S202 is calculated using the log-likelihood function of the Gaussian mixture model, and the calculation formula is:

[0094]

[0095] Among them, α k represents the mixing coefficient of the kth Gaussian distribution, μ k represents the mean of the kth Gaussian distribution, represents the covariance of the kth Gaussian distribution, represents the probability density of the kth Gaussian distribution; N represents the number of data points, x i represents the i-th data point; K represents the number of Gaussian distributions.

[0096] According to the position and speed corresponding to the individual extreme value and the group extreme value, the speed and position of each particle are updated. The speed update formula is:

[0097]

[0098] Among them, w is the inertia weight; k is the current iteration number, c1 is the individual learning factor, c2 is the social learning factor; r1 and r2 are random numbers distributed between [0,1]; represents the position of particle i at the current k-th iteration; represents the individual optimal position of particle i up to the current iteration round; The global optimal position of the entire particle up to the current iteration round; represents the velocity of particle i at the k+1th iteration, represents the velocity of particle i at the kth iteration.

[0099] The particle position update formula is:

[0100]

[0101] in, Represents the position of particle i in the next iteration round; is the position of particle i in the current iteration round, is the velocity of particle i in the next iteration round.

[0102] When the set number of iterations is reached or the swarm extremum reaches the set value, the particle swarm optimization algorithm completes the initialization construction of the Gaussian mixture model parameters.

[0103] In step S103, the main function of the EM algorithm (Expectation-Maximization) is to optimize and estimate the parameters of the Gaussian mixture model (GMM). Specifically, the EM algorithm is used to iteratively solve the hidden variables contained in the GMM, and estimate the mean, variance and weight parameters of the GMM by maximizing the likelihood function, so as to establish an accurate signal data distribution model. After the parameter optimization is completed, the EM algorithm will calculate the posterior probability of each signal point belonging to each Gaussian component based on the calculated Gaussian distribution parameters. Using these posterior probabilities, the signal point is assigned to the cluster label with the highest membership probability, so as to achieve the classification of the signal point and complete the nonlinear equalization.

[0104] In some embodiments, the Gaussian mixture model under the initial parameter state is iterated using the expectation maximization algorithm until the parameters converge, including steps S301 to S306:

[0105] Step S301: define hidden variable z ik , reflecting the data point x i From the kth Gaussian distribution, the expression is:

[0106]

[0107] Wherein, i = 1, 2, ..., N; k = 1, 2, ..., K; the latent variables and the data points constitute complete data, denoted as (x i , z i1 , z i2 , …, z iK )i=1,2,…,N;

[0108] Step S302: The likelihood function expression of the Gaussian mixture model is:

[0109]

[0110] in, μ k represents the mean of the kth Gaussian distribution, represents the covariance of the kth Gaussian distribution; n k Represents the number of N observation points coming from the kth Gaussian distribution.

[0111] Step S303: The log-likelihood function of the Gaussian mixture model is:

[0112]

[0113] Step S304: In step E, the expectation maximization algorithm calculates each data point x according to the current parameters. i The posterior probability of belonging to the kth Gaussian distribution Note θ 0 is the initial value of the iterative parameter, denoted by θ t is the estimated value of the parameter θ in the t-th iteration. By taking the expectation of the log-likelihood function, we get the Q function:

[0114] Q(θ,θ t ) = E(logP(x, x|θ), θ t );

[0115]

[0116] Wherein, i=1, 2, ..., N; k=1, 2, ..., K;

[0117] Step S305: In step M, find θ when the Q function reaches its maximum value, and use it as θ t+1 Participate in the next iteration, the expression is:

[0118]

[0119] Wherein, θ contains the mixing coefficient α of each Gaussian distribution in the Gaussian mixture model k , mean μ k and covariance The expression is:

[0120] θ=(α1,α2,...,α K ;μ1,μ2,...,μ K ;σ1,σ2,...,σ K );

[0121] Then for the mixing coefficient α k , mean μ k and covariance The estimate is:

[0122]

[0123] Step S306: Repeat step E and step M until the parameters converge.

[0124] In step S104, the Gaussian mixture model with optimized parameters is used to calculate each data point x i The posterior probability of belonging to the kth Gaussian distribution The membership degree is used as the clustering division to complete the nonlinear equalization of the signal.

[0125] On the other hand, the present invention also provides a short-distance high-speed optical fiber communication system, such as Figure 3 As shown, the system comprises:

[0126] The transmitting end includes a transmitting end digital signal processor, an arbitrary waveform generator, a digital-to-analog converter and an optical modulator connected in sequence, and the optical modulator is provided with a light source by a first local oscillator laser; the end is used to generate an optical signal with a set order of orthogonal amplitude modulation.

[0127] The optical amplifier is used to amplify the optical signal output by the optical modulator.

[0128] The single-mode optical fiber is used to transmit the amplified optical signal.

[0129] The receiving end includes an optical mixer, a photodiode, a sampling oscilloscope and a receiving end digital signal processor connected in sequence; the optical mixer is connected to a second local oscillator laser; the receiving end executes the short-distance high-speed optical fiber communication equalization method described in steps S101 to S104.

[0130] In some embodiments, the optical fiber amplifier is an erbium-doped optical fiber amplifier. An optical power meter is also provided after the optical modulator.

[0131] Exemplarily, the process of performing nonlinear equalization of optical signals for 64-QAM modulation in the short-distance high-speed optical fiber communication system includes steps S401 to S403:

[0132] Step S401: At the transmitting end, the original 0-1 bit data is modulated by 64-QAM. Each constellation point of the signal constellation diagram represents a transmission symbol, corresponding to 64 categories, numbered from 1 to 64 as the category label; the data is sequentially passed through an arbitrary waveform generator (AWG), an electrical amplifier (EA), and a Mach-Zehnder modulator (MZM) to form an optical signal, which is then divided into two beams and enters a spatial light modulator, and finally merged into optical fiber transmission. At the receiving end of the coherent optical communication system, the polarization beam splitter (PBS) and photodetectors are used for photoelectric conversion, and a four-channel digital sampling oscilloscope is used to digitize the received signal, and the obtained original electrical signal is linearly equalized. For the 64-QAM signal at the current moment, two features, the co-directional component and the orthogonal component of each polarization direction, are selected to construct a data set.

[0133] The high-speed coherent optical communication system is as follows Figure 3 As shown, the basic block diagram of the communication system is shown, which includes an arbitrary waveform generator (AWG) with a rate of 100Ga / s, a digital-to-analog converter, a 40GHz high-bandwidth IQ modulator, a 1550nm laser, an erbium-doped fiber amplifier (EDFA), a G652 standard single-mode fiber (SSMF), an optical mixer, a balanced photodetector, a 256GS / a digital sampling oscilloscope, and digital signal processing (DSP) modules at the transmitting and receiving ends. In the DSP module at the transmitting end, pseudo-random binary sequence (PRBS), QAM symbol mapping, upsampling, root raised cosine filter shaping, and resampling are executed; in the DSP module at the receiving end, low-pass filter (LPF), power normalization, IQ orthogonalization, clock recovery, chromatic dispersion compensation (CDC), polarization mode dispersion compensation (MCMA), carrier recovery (FOE, CPE) and PSO-GMM nonlinear equalization processing, 64QAM codeword inverse mapping and bit error rate calculation are executed.

[0134] Step S402: Build a PSO-GMM clustering model for nonlinear equalization of high-speed coherent optical communication system. The PSO-GMM algorithm model includes an input layer, a PSO algorithm submodule, an EM algorithm submodule and a linear output layer. The input layer sends the prepared 64-QAM data set to the PSO algorithm submodule. The data set X inis a two-dimensional vector with a dimension of N×2, where N represents the length of the data set and 2 represents the number of data features, including two dimensional features of in-phase components and orthogonal components. The PSO algorithm is a heuristic optimization algorithm in the field of intelligent computing. Its core idea is to use the information sharing of individuals in the group to make the movement of the entire group evolve from disorder to order in the problem-solving space, so as to quickly obtain the optimal solution to the problem. It is often used for multi-objective optimization to find the global optimal solution. The PSO algorithm submodule uses the particle swarm optimization algorithm with inertia weight first increasing and then decreasing to complete the parameter initialization of the GMM model to accelerate the convergence process and reduce the number of iterations. After the PSO algorithm submodule completes the parameter initialization, the EM algorithm is used to iterate the parameters to complete the parameter estimation. The EM algorithm is an unsupervised expectation maximization algorithm, which is mainly used to solve the parameter estimation of the mixture model with hidden variables. The EM algorithm solves the maximum likelihood estimation of the parameters through iteration. The output layer uses the estimated parameters of the Gaussian mixture model obtained by iteration of the EM algorithm to divide the signal points into categories, and selects the maximum membership of the current signal point as its final clustering result.

[0135] Step S403: For the PSO-GMM algorithm model for nonlinear equalization of high-speed coherent optical communication systems built in step 2, configure the required parameters, including the parameters of the PSO algorithm and the parameters of the GMM model, wherein the parameters of the PSO algorithm include the particle swarm size and particle dimension. Number of particle iterations, inertia weight, and learning factor; the parameters of the GMM model include the number of Gaussian single models, the mean and variance of each model, and the number of iterations. Use the 64-QAM training data set produced in step 1 to initialize the parameter values ​​of the GMM model, use the PSO algorithm to initialize the parameter values ​​of the GMM model, and then use the GM algorithm to estimate the parameters. After iteration, the optimal parameter estimate is obtained, and the parameters are used to cluster the data points and divide them into 64 clusters of the same cluster to complete the nonlinear equalization of the signal.

[0136] In order to find the best parameters, parameter control is required. First, various parameters of the PSO algorithm submodule are determined. In this example, the particle swarm size is set to 64 and the particle dimension is set to 2. The number of particle iterations is 150, and the inertia weight is initialized to 0.8. It changes linearly with the increase in the number of iterations. The learning factor is 0.01. After the iteration is completed, the inertia weight is 0.443. At the same time, the initial values ​​of the GMM model parameters are obtained. Then, the optimal values ​​of the GMM model parameters are determined through the EM algorithm iteration. In this example, there are 64 single Gaussian models and their corresponding parameters.

[0137] After determining the initial parameters of the GMM model, the iteration begins. When the iteration error remains basically unchanged, the iteration ends and a complete clustering model is obtained. At the same time, the clustering of 64QAM signals is completed and the bit error rate is calculated to be 0.0187. The PSO-GMM model is used to cluster the signals under different receiving end optical power ROP. The results are as follows: Figure 4 The PSO-GMM model is compared with the traditional GMM, K-mean, and KNN clustering models. The Q factor is used to analyze the clustering effect of different clustering models on 64-QAM signals within the range of -15dBm to -5dBm at the system optical transmit power (ROP). The results are shown in Figure 5 shown.

[0138] The system processes the nonlinear damage of the 64-QAM signal at the receiving end. For the signal sequence after conventional linear equalization processing, the 64-QAM signal at each moment has four characteristic dimensions, including real and imaginary data in two polarization directions. The nonlinear interference of optical fiber and high-speed optical devices causes nonlinear effects between symbols. For the 64-QAM symbol at each moment, the PSO-GMM algorithm is used for iteration, and the parameters of the GMM model are used to divide the signal into the cluster with the largest membership, so as to complete the nonlinear equalization of the signal. The PSO algorithm submodule is used to initialize the parameters of the GMM model to improve the convergence speed of the GMM model and reduce the number of iterations. At the same time, the dynamic inertia weight is used to make the particle swarm algorithm have a strong global convergence ability in the early stage and a strong local convergence ability in the later stage. The EM algorithm is used to perform maximum likelihood estimation on the parameters of the GMM model, and the obtained parameters are used to cluster the signal.

[0139] Corresponding to the above method, the present invention also provides an apparatus / system, which includes a computer device, the computer device includes a processor and a memory, the memory stores computer instructions, the processor is used to execute the computer instructions stored in the memory, and when the computer instructions are executed by the processor, the apparatus / system implements the steps of the method described above.

[0140] The embodiment of the present invention also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the aforementioned edge computing server deployment method are implemented. The computer-readable storage medium can be a tangible storage medium, such as a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a floppy disk, a hard disk, a removable storage disk, a CD-ROM, or any other form of storage medium known in the technical field.

[0141] In summary, the short-distance high-speed optical fiber communication equalization method and system described in the present invention constructs a Gaussian mixture model for an optical signal modulated by orthogonal amplitude modulation of a set order, first uses a particle swarm optimization (PSO) algorithm with an inertia weight that increases first and then decreases to complete the parameter initialization of the Gaussian mixture model, inputs the data features of two dimensions of the in-phase component and the orthogonal component of the data into the Gaussian mixture model, and then uses the expectation maximization (EM) algorithm to solve the parameters of the Gaussian mixture model, determines the membership of each data point with the clustering label corresponding to each univariate Gaussian distribution through the parameters, and assigns it to the corresponding clustering label through the maximum a posteriori probability to complete nonlinear equalization. The present invention can accurately capture the memory properties of nonlinear damage to the signal, improve the accuracy of nonlinear clustering, and reduce the computational complexity of the nonlinear clustering model under the same clustering effect.

[0142] It should be understood by those skilled in the art that the exemplary components, systems and methods described in conjunction with the embodiments disclosed herein can be implemented in hardware, software or a combination of the two. Whether it is performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention. When implemented in hardware, it can be, for example, an electronic circuit, an application specific integrated circuit (ASIC), appropriate firmware, a plug-in, a function card, etc. When implemented in software, the elements of the present invention are programs or code segments used to perform the required tasks. The program or code segment can be stored in a machine-readable medium, or transmitted on a transmission medium or a communication link via a data signal carried in a carrier.

[0143] It should be clear that the present invention is not limited to the specific configuration and processing described above and shown in the figures. For the sake of simplicity, a detailed description of the known method is omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of the present invention is not limited to the specific steps described and shown, and those skilled in the art can make various changes, modifications and additions, or change the order between the steps after understanding the spirit of the present invention.

[0144] In the present invention, features described and / or illustrated for one embodiment may be used in the same or similar manner in one or more other embodiments, and / or combined with features of other embodiments or replace features of other embodiments.

[0145] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the embodiments of the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A short-distance high-speed optical fiber communication equalization method, characterized in that: The method is used to be performed at a receiving end in a single-mode optical fiber communication system, and the method comprises the following steps: The optical signal of the transmitting end is received after orthogonal amplitude modulation of a set order, and after polarization splitting, photoelectric detection and four-channel digital sampling are performed on the optical signal, a linear equalization process is performed using a preset method to obtain a complete data set in two polarization directions; each symbol in the symbol sequence of the data set expresses the real part and the imaginary part of the data point based on a two-dimensional vector; For the complete data set, using a particle swarm optimization algorithm to establish initial parameters of a Gaussian mixture model, the Gaussian mixture model comprising a plurality of Gaussian distributions, each Gaussian distribution corresponding to the set order to describe the distribution of each constellation point; Iterating the Gaussian mixture model under the initial parameter state by using an expectation maximization algorithm until the parameters converge; Based on the Gaussian mixture model after parameter convergence, the posterior probability that each data point belongs to each Gaussian distribution is calculated as the membership degree, and clustering is performed according to the membership degree to divide the clusters into the same cluster, thereby completing nonlinear equalization.

2. The short-distance high-speed optical fiber communication equalization method according to claim 1, characterized in that: Perform linear equalization using preset methods, including: The sampled data is subjected to I / Q orthogonalization, low-pass filtering, Gardner time synchronization, dispersion compensation, frequency offset compensation and phase offset compensation.

3. The short-distance high-speed optical fiber communication equalization method according to claim 2, characterized in that: For the complete data set, the initial parameters of the Gaussian mixture model are established using the particle swarm optimization algorithm, including: Initialize a particle swarm, each particle representing a solution to the Gaussian mixture model parameters; In each iteration, the fitness value of each particle is calculated and recorded; the individual extreme value and the group extreme value are determined based on the fitness value of each particle; the speed and position of each particle itself are updated according to the position and speed corresponding to the individual extreme value and the group extreme value; When the set number of iterations is reached or the group extremum reaches the set value, the iteration is stopped, and the solution of the particle corresponding to the group extremum is output as the initial parameter of the Gaussian mixture model.

4. The short-distance high-speed optical fiber communication equalization method according to claim 3, characterized in that: The fitness value is calculated using the log-likelihood function of the Gaussian mixture model, and the calculation formula is: Among them, α k represents the mixing coefficient of the kth Gaussian distribution, μ k represents the mean of the kth Gaussian distribution, represents the covariance of the kth Gaussian distribution, represents the probability density of the kth Gaussian distribution; N represents the number of data points, x i represents the i-th data point; K represents the number of Gaussian distributions; In updating the speed and position of each particle according to the position and speed corresponding to the individual extreme value and the group extreme value, the speed update formula is: Among them, w is the inertia weight; k is the current iteration number, c1 is the individual learning factor, c2 is the social learning factor; r1 and r2 are random numbers distributed between [0,1]; represents the position of particle i at the current k-th iteration; represents the individual optimal position of particle i up to the current iteration round; The global optimal position of the entire particle up to the current iteration round; represents the velocity of particle i at the k+1th iteration, represents the velocity of particle i at the kth iteration; The particle position update formula is: in, Represents the position of particle i in the next iteration round; is the position of particle i in the current iteration round, is the velocity of particle i in the next iteration round.

5. The short-distance high-speed optical fiber communication equalization method according to claim 4, characterized in that: The Gaussian mixture model under the initial parameter state is iterated using an expectation maximization algorithm until the parameters converge, including: Define the hidden variable z ik , reflecting the data point x i From the kth Gaussian distribution, the expression is: Wherein, i = 1, 2, ..., N; k = 1, 2, ..., K; the latent variables and the data points constitute complete data, denoted as (x i , z i1 , z i2 , …, z iK )i=1,2,…,N; The likelihood function expression of the Gaussian mixture model is: in, μ k represents the mean of the kth Gaussian distribution, represents the covariance of the kth Gaussian distribution; n k Represents the number of N observation points from the kth Gaussian distribution; Then the log-likelihood function of the Gaussian mixture model is: In the E step, the expectation maximization algorithm calculates each data point x according to the current parameters. i The posterior probability of belonging to the kth Gaussian distribution Note θ 0 is the initial value of the iterative parameter, denoted by θ t For the estimated value of the parameter θ in the t-th iteration, the expectation of the log-likelihood function is obtained to obtain the Q function: Q(θ,θ t )=E(log P(x,x|θ),θ t ); Wherein, i=1, 2, ..., N; k=1, 2, ..., K; In step M, find the value of θ when the Q function reaches its maximum value, as θ t+1 Participate in the next iteration, the expression is: Wherein, θ contains the mixing coefficient α of each Gaussian distribution in the Gaussian mixture model k , mean μ k and covariance The expression is: θ=(α1,α2,...,α K ;μ1,μ2,...,μ K ;σ1,σ2,...,σ K ); Then for the mixing coefficient α k , mean μ k and covariance The estimate is: The E step and the M step are repeated until the parameters converge.

6. The short-distance high-speed optical fiber communication equalization method according to claim 1, characterized in that: In the method, the transmitting end uses 16-QAM, 64-QAM or 256-QAM to modulate the optical signal.

7. A short-distance high-speed optical fiber communication system, characterized in that: The system comprises: A transmitting end, the transmitting end comprising a transmitting end digital signal processor, an arbitrary waveform generator, a digital-to-analog converter and an optical modulator connected in sequence, the optical modulator being provided with a light source by a first local oscillator laser; the transmitting end is used to generate an optical signal subjected to orthogonal amplitude modulation of a set order; an optical amplifier, used to amplify the optical signal output by the optical modulator; A single-mode optical fiber, used for transmitting the amplified optical signal; The receiving end includes an optical mixer, a photodiode, a sampling oscilloscope and a receiving end digital signal processor connected in sequence; the optical mixer is connected to a second local oscillator laser; the receiving end executes the short-distance high-speed optical fiber communication equalization method as described in any one of claims 1 to 6.

8. The short-distance high-speed optical fiber communication system according to claim 7, characterized in that: The optical fiber amplifier is an erbium-doped optical fiber amplifier.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images