Channel state information estimation method and device, terminal equipment and storage medium

Abstract

The invention discloses a channel state information estimation method and device, terminal equipment and a storage medium, and belongs to the technical field of communication, and the method comprises the following steps: collecting an original optical signal, and carrying out exploratory data analysis processing on the original optical signal to obtain preprocessed data; inputting the preprocessed data into an auto-encoder, and compressing the preprocessed data into low-dimensional potential features through the auto-encoder; inputting the low-dimensional potential features and the historical channel state information labels into an initial random forest model for model training to obtain a random forest model for estimating channel state information; and inputting a to-be-measured optical signal into the random forest model to obtain a channel state information estimation value. According to the method, the complex data relation can be processed in scenes such as nonlinear signal propagation or rapid environment change, and the method is suitable for complex communication scenes, so that the accuracy of channel state information estimation can be effectively improved, and the communication reliability can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a channel state information estimation method, apparatus, terminal equipment and storage medium. Background Art

[0002] With the rapid growth of internet traffic in the big data era, optical communication (OptiCom) systems are being upgraded to ultra-high speeds to meet the growing demand from high-bandwidth devices such as cloud computing platforms. As the demand for high-speed, high-capacity data transmission continues to rise, optical communication networks are reshaping the way information is transmitted and received over long distances. For traditional communication networks, problems such as aging components, broken lines, and cable anomalies can cause communication interruptions. At the same time, increasingly complex communication scenarios such as environmental changes and signal attenuation further complicate the design and optimization of traditional networks. To address these challenges, optical communication technology uses light to transmit information. Its high bandwidth and low latency make it an ideal solution for data transmission. Optical communication networks can dynamically adapt to evolving network conditions and ensure optimal performance.

[0003] Existing channel state information estimation methods are usually based on static rules to achieve channel state information estimation. They are unable to handle complex data relationships in scenarios such as nonlinear signal propagation or rapid environmental changes, resulting in low accuracy of channel state information estimation. Summary of the Invention

[0004] The present invention provides a channel state information estimation method, apparatus, terminal device and storage medium, which can solve the technical problem in the prior art that complex data relationships cannot be processed in scenarios such as nonlinear signal propagation or rapid environmental changes, resulting in low accuracy of channel state information estimation.

[0005] The present invention provides a channel state information estimation method, comprising:

[0006] Collecting original optical signals, performing exploratory data analysis on the original optical signals to obtain preprocessed data;

[0007] Inputting the preprocessed data into an autoencoder, and compressing the preprocessed data into low-dimensional latent features through the autoencoder;

[0008] Inputting the low-dimensional potential features and historical channel state information labels into an initial random forest model for model training to obtain a random forest model for estimating channel state information;

[0009] The optical signal to be measured is input into the random forest model to obtain a channel state information estimation value.

[0010] Furthermore, performing exploratory data analysis on the original optical signal to obtain preprocessed data includes:

[0011] Determining outliers in the original optical signal using an interquartile range method;

[0012] Abnormal values ​​in the original optical signal are eliminated to obtain cleaned data, and correlation analysis is performed on the cleaned data. Parameters in the cleaned data with correlation coefficients greater than a preset value are used as preprocessed data.

[0013] Furthermore, compressing the preprocessed data into low-dimensional potential features by the autoencoder includes:

[0014] The preprocessed data is mapped to a low-dimensional potential representation, and the preprocessed data is reconstructed in the low-dimensional potential representation to obtain low-dimensional potential features.

[0015] Furthermore, the reconstructing the preprocessed data in the low-dimensional potential representation to obtain low-dimensional potential features includes:

[0016] Minimizing divergence is introduced into the low-dimensional latent representation. During the reconstruction process, the difference between the latent space distribution and the standard normal distribution is minimized based on the minimizing divergence to obtain the low-dimensional latent features.

[0017] Furthermore, the low-dimensional potential features and historical channel state information labels are input into an initial random forest model for model training to obtain a random forest model for estimating channel state information, including:

[0018] Constructing a training set based on the low-dimensional potential features and corresponding historical channel state information labels, and generating multiple sub-datasets based on the training set;

[0019] A corresponding decision tree is obtained by training each of the sub-data sets, and a random forest model for estimating channel state information is constructed based on multiple decision trees.

[0020] Furthermore, after inputting the optical signal to be measured into the random forest model to obtain a channel state information estimation value, the method further includes: feeding back the channel state information to the optical communication system, and adjusting corresponding transmission parameters through the optical communication system.

[0021] The present invention provides a channel state information estimation device, comprising:

[0022] A preprocessing module is used to collect original optical signals and perform exploratory data analysis on the original optical signals to obtain preprocessed data;

[0023] a data compression module, configured to input the preprocessed data into an autoencoder, and compress the preprocessed data into low-dimensional latent features through the autoencoder;

[0024] A model training module, configured to input the low-dimensional potential features and historical channel state information labels into an initial random forest model for model training to obtain a random forest model for estimating channel state information;

[0025] The information estimation module is used to input the optical signal to be measured into the random forest model to obtain a channel state information estimation value.

[0026] Furthermore, the pre-processing module is further used to:

[0027] Determining outliers in the original optical signal using an interquartile range method;

[0028] Abnormal values ​​in the original optical signal are eliminated to obtain cleaned data, and correlation analysis is performed on the cleaned data. Parameters in the cleaned data with correlation coefficients greater than a preset value are used as preprocessed data.

[0029] The present invention provides a terminal device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the channel state information estimation method as described above is implemented.

[0030] The present invention provides a computer-readable storage medium, comprising: a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the channel state information estimation method as described above.

[0031] The following beneficial effects are achieved by implementing the present invention:

[0032] The embodiment of the present invention compresses the preprocessed optical signal into low-dimensional potential features through an autoencoder, and obtains a random forest model for estimating channel state information through model training through a random forest model. The autoencoder can learn nonlinear features to solve complex linear problems of signal propagation, and dynamically integrate learning with the random forest model to cope with rapid environmental changes. It can process complex data relationships in scenarios such as nonlinear signal propagation or rapid environmental changes, and is suitable for complex communication scenarios, thereby effectively improving the accuracy of channel state information estimation, and further effectively improving communication reliability.

[0033] Furthermore, the embodiments of the present invention not only improve the robustness of the model to noise and outliers by introducing the minimized divergence constrained latent space distribution, but also enhance the cross-scenario generalization capability by normalizing the latent features, thereby significantly reducing the bit error rate prediction error in optical communication channel state estimation and improving the accuracy of channel state information estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the implementation. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0035] Figure 1 1 is a flow chart of a channel state information estimation method provided by an embodiment of the present invention;

[0036] Figure 2 The figure is a schematic diagram of the structure of a channel state information estimation device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0037] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.

[0038] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0039] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.

[0040] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0041] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0042] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0043] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.

[0044] See also Figure 1 To address the technical problem in the prior art that complex data relationships cannot be processed in scenarios such as nonlinear signal propagation or rapid environmental changes, resulting in low accuracy of channel state information estimation, an embodiment of the present invention provides a channel state information estimation method, including:

[0045] S1. Collecting original optical signals, performing exploratory data analysis on the original optical signals to obtain preprocessed data;

[0046] In the embodiment of the present invention, the original optical signal is an original optical pulse signal transmitted in an optical communication system, including light intensity, wavelength, phase, etc.

[0047] In the embodiment of the present invention, exploratory data analysis includes statistical analysis, visual analysis, outlier removal, etc.

[0048] The embodiment of the present invention obtains preprocessed data by performing exploratory data analysis and processing on the original optical signal, which can effectively improve the data cleaning effect and enhance the data quality.

[0049] S2, input the preprocessed data into the autoencoder, and compress the preprocessed data into low-dimensional potential features through the autoencoder;

[0050] In the embodiment of the present invention, the low-dimensional latent features are the core abstract representations of the data, that is, features that reduce the dimensionality while retaining key information.

[0051] The embodiments of the present invention can effectively reduce the complexity of subsequent model calculations by reducing data dimensions.

[0052] S3, inputting the low-dimensional potential features and the historical channel state information labels into the initial random forest model for model training to obtain a random forest model for estimating the channel state information;

[0053] S4. Input the optical signal to be measured into the random forest model to obtain a channel state information estimation value.

[0054] In the embodiment of the present invention, the channel state information estimation value includes signal-to-noise ratio, optical power attenuation, dispersion, etc.

[0055] The embodiment of the present invention compresses the preprocessed optical signal into low-dimensional potential features through an autoencoder, and obtains a random forest model for estimating channel state information through model training through a random forest model. The autoencoder can learn nonlinear features to solve complex linear problems of signal propagation, and dynamically integrate learning with the random forest model to cope with rapid environmental changes. It can process complex data relationships in scenarios such as nonlinear signal propagation or rapid environmental changes, and is suitable for complex communication scenarios, thereby effectively improving the accuracy of channel state information estimation, and further effectively improving communication reliability.

[0056] In one embodiment, step S1, performing exploratory data analysis on the original optical signal to obtain preprocessed data, includes:

[0057] S11, using the interquartile range method to determine outliers in the original optical signal;

[0058] In the embodiment of the present invention, before using the interquartile range method to determine the outliers in the desired optical signal, the method may further include:

[0059] The mean and standard deviation of the original optical signal data set are calculated to measure the degree to which the data points deviate from the average value, specifically:

[0060] The mean is calculated using the following formula:

[0061]

[0062] in, is the mean, N is the number of data, x i It should be noted that the standard deviation is a statistical indicator that describes the degree of dispersion of data distribution and is used to measure the degree to which data points deviate from the mean. For the data set X = {x (1) ,x (2) ,...,x (N)}, the standard deviation is given by:

[0063]

[0064] Among them, σ is the standard deviation. The larger its value, the more dispersed the data distribution is; the smaller its value, the more concentrated the data is around the mean.

[0065] On this basis, the embodiment of the present invention uses the interquartile range method to measure the degree of concentration and dispersion of data to identify data outliers, wherein the calculation formula is:

[0066] IQR=Q3-Q1,

[0067]

[0068] Among them, IQR is the difference between the upper quartile and the lower quartile, Q3 is the upper quartile, Q1 is the lower quartile, Outliers is a set of outliers, the upper quartile is the value that the first preset proportion of data does not exceed after the data set is sorted in ascending order, and the lower quartile is the value that the second preset proportion of data does not exceed after the data set is upgraded and sorted. The first preset proportion can be 25% and the second preset proportion can be 75%.

[0069] S12. Abnormal values ​​in the original optical signal are eliminated to obtain cleaned data, and correlation analysis is performed on the cleaned data. Parameters in the cleaned data whose correlation coefficients are greater than a preset value are used as preprocessed data.

[0070] In the embodiment of the present invention, the calculation formula of the correlation coefficient is as follows:

[0071]

[0072] Among them, x1 and x2 are two variables. is the correlation coefficient, is the standard deviation of variable x1, is the standard deviation of variable x2, Cov(x1,x2) is the covariance between x1 and x2, which indicates the degree of joint variation of the two variables.

[0073] The expression of Cov(x1,x2) is as follows:

[0074]

[0075] in, are the average values ​​of x1 and x2, The value range of is [-1, 1], specifically: ρ = 1 is completely positive correlation; ρ = -1 is completely negative correlation; ρ = 0 is nonlinear correlation. The data set contains N features, then the correlation matrix is ​​defined as:

[0076]

[0077] The diagonal elements are always 1, because the correlation coefficient of any variable with itself is 1 and is symmetric about the diagonal.

[0078] The embodiments of the present invention can effectively reduce the influence of invalid data in the channel state information estimation process by eliminating outliers and performing correlation analysis, thereby effectively improving the accuracy of channel state information estimation.

[0079] In one embodiment, during the preprocessing stage, a frequency distribution diagram may be used to represent the distribution of data in different intervals, specifically:

[0080] If the data X={x (1) ,x (2) ,...,x (N)} is divided into intervals, then the jth interval has the following frequency (or density):

[0081]

[0082] Among them, Frequency j is the frequency of the jth interval, b j , b j+1 are the upper and lower bounds of the jth interval. And ΙΙ(·) is an indicator function, which is 1 when the condition is met and 0 otherwise.

[0083] In order to make the plot height independent of the interval width and eliminate the interference of interval width on the graph height, the frequency density is defined as:

[0084]

[0085] Among them, Density j is the frequency density of the jth interval, width j = b j+1 -b j is the jth interval width.

[0086] The kernel density estimation then uses the smoothed data point distribution to generate a continuous probability density curve, which is expressed as:

[0087]

[0088] in, The probability density function is B, which is a smoothing parameter used to control the width of the kernel function. K(·) is the kernel function that determines the contribution of each point (determines the contribution weight of each data point to the density). Common kernel functions include:

[0089] (1) Gaussian kernel:

[0090]

[0091] Among them, K(u) is the kernel function.

[0092] (2) Unified kernel:

[0093]

[0094] Kernel density estimation approximates the data distribution through a continuous function, effectively eliminating the discontinuity introduced in the histogram due to binning.

[0095] In one embodiment, step S2, compressing the preprocessed data into low-dimensional latent features by an autoencoder, includes:

[0096] S21. Map the preprocessed data to a low-dimensional latent representation, reconstruct the preprocessed data in the low-dimensional latent representation, and obtain low-dimensional latent features.

[0097] In an embodiment of the present invention, divergence minimization is introduced in the low-dimensional latent representation. During the reconstruction process, the difference between the latent space distribution and the standard normal distribution is minimized based on the divergence minimization to obtain the low-dimensional latent features.

[0098] In the embodiment of the present invention, the autoencoder is an unsupervised learning algorithm, which is mainly used for feature extraction and dimensionality reduction.

[0099] The overall process of the embodiment of the present invention can be divided into two core parts: first, the input data is pre-processed data Mapped to a low-dimensional potential representation Z, the input data X is then reconstructed from the potential representation Z to obtain the reconstructed data The reconstructed data is the low-dimensional latent feature. The goal of the autoencoder is to achieve efficient data representation by minimizing the reconstruction error between input and output.

[0100] 1) Encoding process: The encoder takes the input data Convert to low-dimensional latent representation (where M < D), the specific form is as follows:

[0101] Z=T encoder (X) = σ(W1X+b1)

[0102] Where W1 is the weight matrix, b1 is the bias term in the encoder, and T encoder (X) is the encoder function, and function σ is a nonlinear activation function. It can help the model capture complex nonlinear relationships.

[0103] 2) Decoding process: The decoder will extract the latent representation from the low-dimensional latent representation Z and recover the input data, that is:

[0104]

[0105] Among them, W2 is the weight matrix, f decoder (Z) is the decoder function and b2 is the bias term in the decoder.

[0106] 3) Optimization goal: The training goal of the autoencoder is to minimize the input data X and the reconstructed data The commonly used loss function is the reconstruction error, which is usually measured using the mean squared error (MSE), as shown below:

[0107]

[0108] Where N is the number of samples in the training set, X i is the first input data of the i-th sample, is the first reconstructed data of the i-th sample.

[0109] In the embodiment of the present invention, KL divergence is also introduced, that is, by minimizing the Kullback-Leibler (KL) divergence, the autoencoder can learn effective representations and ensure that the latent space has a more regular geometric structure, thereby improving the generalization ability of the model. Specifically, given a latent representation set Z = {z (1) ,z (2) ,...,z (N)} and input data set X={x (1) ,x (2) ,...,x (N)}, KL divergence is regularized by forcing the encoder distribution q(z|x) to approach the preset prior distribution (usually the standard normal distribution p(z)), which is defined as:

[0110]

[0111] Among them, q(z|x) is an approximate distribution, and the core goal of KL divergence is to narrow the difference between the approximate distribution q(z|x) and the true prior distribution p(z).

[0112] The embodiment of the present invention defines the expected information content to measure the uncertainty of the distribution. The expected information content under the probability distribution q(z|x) is defined as follows:

[0113]

[0114] in, is the expected amount of information under the q(z|x) distribution, I(Z) is the information function, that is, the entropy calculated by H(q) according to the distribution q(z|x), expressed as:

[0115] H(q)=-∫q(z|x)logq(z|x)dz.

[0116] where H(q) is the entropy of the q(z|x) distribution.

[0117] In this embodiment of the present invention, the true distribution p(z) is used instead of the approximate distribution q(z|x) to calculate the true distribution p(z) under the expected amount of information:

[0118]

[0119] in, represents the expected amount of information under the distribution of p(z).

[0120] That is, the entropy H(p) under the distribution p(z) is calculated based on the distribution p(z) and is expressed as:

[0121] H(p)=-∫p(z)logp(z)dz.

[0122] Therefore, the KL divergence reflects the statistical distance between the true distribution p(z) and the approximate distribution q(z|x) by quantifying the difference in the expected amount of information:

[0123]

[0124] Among them, D KL (q||p) is the KL discretization of distribution q with respect to distribution p.

[0125] Substituting the above formula into the equation, we can get:

[0126] D KL (q||p)=-∫p(z)logp(z)dz+∫q(z|x)logq(z|x)dz.

[0127] Rewritten as:

[0128] D KL (q||p)=∫q(z|x)logq(z|x)dz-∫q(z|x)logP(z)dz.

[0129] Further rewritten as:

[0130]

[0131] Among them, D KL (q(z|x)||p(z))≥0. The KL divergence is zero if and only if q(z||x)=p(z).

[0132] In an embodiment of the present invention, KL divergence can be added to the loss function to minimize the reconstruction error while constraining the potential space distribution to obtain better generalization ability. That is, the comprehensive loss function of the data before and after the introduction of KL divergence is:

[0133] L total = L MSE + βD KL (q(z|x)||p(z)),

[0134] where, L total is the total loss function, L MSE is the mean squared error, and β is the balance coefficient for adjusting the weights of the reconstruction loss (MSE) and the KL divergence.

[0135] By introducing the minimization of divergence to constrain the latent space distribution, the embodiments of the present invention can not only improve the robustness of the model to noise and outliers, but also enhance the cross-scene generalization ability by normalizing the latent features, thereby significantly reducing the prediction error of the bit error rate and improving the accuracy of channel state information estimation in optical communication channel state estimation.

[0136] In one embodiment, step S3: input the low-dimensional latent features and the historical channel state information labels into the initial random forest model for model training to obtain a random forest model for estimating the channel state information, including:

[0137] S31: Construct a training set according to the low-dimensional latent features and the corresponding historical channel state information labels, and generate multiple sub-datasets according to the training set;

[0138] S32: Train a corresponding decision tree according to each sub-dataset, and construct a random forest model for estimating the channel state information according to multiple decision trees.

[0139] In the embodiments of the present invention, the generation of the decision tree includes random feature selection and tree growth, where the random feature selection is:

[0140] For each split node, randomly select s features (s < S, S is the total number of features) from all features, and then select the best split point from them to divide the data. The tree growth is: start recursively splitting the data from the root node until the stop condition is met (such as the node purity reaches the standard or the maximum depth is reached).

[0141] In the embodiments of the present invention, the random feature selection mechanism can enhance the robustness of the model to feature perturbations and improve the diversity of the base learners.

[0142] In the embodiments of the present invention, after training to obtain a random forest model for estimating the channel state information, obtain the optical signal to be evaluated and input it into the random forest model, and obtain the final result by combining the outputs of each tree. For a regression problem, obtain the final prediction by averaging the outputs of all k decision trees to minimize the mean squared error:

[0143]

[0144] Among them, h j (x) represents the prediction result of the j-th decision tree for the input x, and k is the total number of decision trees.

[0145] In the embodiment of the present invention, the goal of the regression task is to reduce the overall prediction error by integrating and averaging the prediction errors of individual trees, thereby ensuring the robustness and accuracy of the prediction results.

[0146] In addition, the performance of random forests depends on the tuning of the following parameters: (1) the number of trees: the more trees there are in a random forest, the lower the variance of the model, but the more computational effort it also requires; (2) the maximum number of features: the number of features randomly selected for each node; (3) the maximum depth: controls the maximum depth of a single tree to prevent overfitting caused by excessive tree depth; (4) the minimum number of sample splits: specifies the minimum number of samples required for a node to be split, which is used to control the complexity of the tree; and (5) the minimum number of leaf node samples: limits the minimum number of samples that must be retained for a leaf node, which is used to suppress excessive tree growth.

[0147] The embodiment of the present invention uses decision tree integration guided by sub-datasets, which not only effectively solves the overfitting problem of traditional models in dynamic and nonlinear scenarios, but also significantly improves the model's adaptability to complex channel environments through parallel training and feature random selection mechanisms.

[0148] In one embodiment, after step 4, inputting the optical signal to be measured into the random forest model to obtain the channel state information estimation value, the method further includes: feeding back the channel state information to the optical communication system, and adjusting the corresponding transmission parameters through the optical communication system.

[0149] In an embodiment of the present invention, the control module in the optical communication system can adjust the transmission parameters of the transmitting end of the optical communication system, including the modulation format, coding rate, and transmission power, according to the channel state information, so as to be able to quickly adapt to channel changes in various complex scenarios.

[0150] The implementation of the present invention has the following beneficial effects:

[0151] The embodiment of the present invention compresses the preprocessed optical signal into low-dimensional potential features through an autoencoder, and obtains a random forest model for estimating channel state information through model training through a random forest model. The autoencoder can learn nonlinear features to solve complex linear problems of signal propagation, and dynamically integrate learning with the random forest model to cope with rapid environmental changes. It can process complex data relationships in scenarios such as nonlinear signal propagation or rapid environmental changes, and is suitable for complex communication scenarios, thereby effectively improving the accuracy of channel state information estimation, and further effectively improving communication reliability.

[0152] Furthermore, the embodiments of the present invention not only improve the robustness of the model to noise and outliers by introducing the minimized divergence constrained latent space distribution, but also enhance the cross-scenario generalization capability by normalizing the latent features, thereby significantly reducing the bit error rate prediction error in optical communication channel state estimation and improving the accuracy of channel state information estimation.

[0153] See also Figure 2 Based on the same inventive concept as the above embodiment, an embodiment of the present invention provides a channel state information estimation device, including:

[0154] The preprocessing module 10 is used to collect the original optical signal and perform exploratory data analysis on the original optical signal to obtain preprocessed data;

[0155] A data compression module 20 is used to input the preprocessed data into the autoencoder and compress the preprocessed data into low-dimensional potential features through the autoencoder;

[0156] A model training module 30 is configured to input low-dimensional potential features and historical channel state information labels into an initial random forest model for model training to obtain a random forest model for estimating channel state information;

[0157] The information estimation module 40 is used to input the optical signal to be measured into the random forest model to obtain a channel state information estimation value.

[0158] In one embodiment, exploratory data analysis is performed on the original optical signal to obtain preprocessed data, including:

[0159] The interquartile range method was used to identify outliers in the raw optical signal;

[0160] Abnormal values ​​in the original optical signal are eliminated to obtain cleaned data, and correlation analysis is performed on the cleaned data. Parameters in the cleaned data with correlation coefficients greater than a preset value are used as preprocessed data.

[0161] In one embodiment, the preprocessed data is compressed into low-dimensional latent features by an autoencoder, including:

[0162] The preprocessed data is mapped to a low-dimensional latent representation, and the preprocessed data is reconstructed in the low-dimensional latent representation to obtain low-dimensional latent features.

[0163] In one embodiment, the preprocessed data is reconstructed in a low-dimensional latent representation to obtain low-dimensional latent features, including:

[0164] Minimizing divergence is introduced in the low-dimensional latent representation. During the reconstruction process, the difference between the latent space distribution and the standard normal distribution is minimized based on minimizing divergence to obtain low-dimensional latent features.

[0165] In one embodiment, low-dimensional potential features and historical channel state information labels are input into an initial random forest model for model training to obtain a random forest model for estimating channel state information, including:

[0166] A training set is formed based on low-dimensional potential features and corresponding historical channel state information labels, and multiple sub-datasets are generated based on the training set;

[0167] A corresponding decision tree is obtained by training each sub-data set, and a random forest model for estimating channel state information is constructed based on multiple decision trees.

[0168] In one embodiment, a parameter adjustment module is further included, which is used to: feed back the channel state information to the optical communication system, and adjust the corresponding transmission parameters through the optical communication system.

[0169] It can be understood that the above-mentioned device embodiment corresponds to the method embodiment of the present invention, and can implement any one of the above-mentioned method embodiments of the present invention to provide a channel state information estimation method.

[0170] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. Furthermore, in the drawings of the device embodiments provided by the present invention, the connection relationship between modules indicates that they have a communication connection, which may be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement the present invention without inventive effort.

[0171] Based on the above-mentioned embodiment of the channel state information estimation method, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the channel state information estimation method of any embodiment of the present invention.

[0172] For example, in this embodiment, the computer program may be divided into one or more modules, one or more of which are stored in a memory and executed by a processor to implement the present invention. One or more module elements may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in a terminal device.

[0173] The terminal device can be a computing device such as a desktop computer, notebook computer, PDA, or cloud server. The terminal device may include, but is not limited to, a processor and memory.

[0174] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.

[0175] Based on the above-mentioned method embodiments, another embodiment of the present invention provides a computer-readable storage medium, including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the channel state information estimation method described in any one of the above-mentioned method embodiments of the present invention.

[0176] Wherein, the module / unit integrated in the device / terminal equipment, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0177] The above is a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A channel state information estimation method, characterized in that: include: Collecting original optical signals, performing exploratory data analysis on the original optical signals to obtain preprocessed data; Inputting the preprocessed data into an autoencoder, and compressing the preprocessed data into low-dimensional latent features through the autoencoder; Inputting the low-dimensional potential features and historical channel state information labels into an initial random forest model for model training to obtain a random forest model for estimating channel state information; The optical signal to be measured is input into the random forest model to obtain a channel state information estimation value.

2. The channel state information estimation method according to claim 1, wherein The performing exploratory data analysis on the original optical signal to obtain preprocessed data includes: Determining outliers in the original optical signal using an interquartile range method; Abnormal values ​​in the original optical signal are eliminated to obtain cleaned data, and correlation analysis is performed on the cleaned data. Parameters in the cleaned data with correlation coefficients greater than a preset value are used as preprocessed data.

3. The channel state information estimation method according to claim 1, wherein: The compressing the pre-processed data into low-dimensional potential features by the autoencoder comprises: The preprocessed data is mapped to a low-dimensional potential representation, and the preprocessed data is reconstructed in the low-dimensional potential representation to obtain low-dimensional potential features.

4. The channel state information estimation method according to claim 3, wherein: The reconstructing the preprocessed data in the low-dimensional potential representation to obtain low-dimensional potential features includes: Minimizing divergence is introduced into the low-dimensional latent representation. During the reconstruction process, the difference between the latent space distribution and the standard normal distribution is minimized based on the minimizing divergence to obtain the low-dimensional latent features.

5. The channel state information estimation method according to claim 1, wherein: The step of inputting the low-dimensional potential features and the historical channel state information labels into an initial random forest model for model training to obtain a random forest model for estimating channel state information includes: Constructing a training set based on the low-dimensional potential features and corresponding historical channel state information labels, and generating multiple sub-datasets based on the training set; A corresponding decision tree is obtained by training each of the sub-data sets, and a random forest model for estimating channel state information is constructed based on multiple decision trees.

6. The channel state information estimation method according to claim 1, wherein: After inputting the optical signal to be measured into the random forest model to obtain a channel state information estimation value, the method further includes: feeding back the channel state information to an optical communication system, and adjusting corresponding transmission parameters through the optical communication system.

7. A channel state information estimation device, characterized in that: include: A preprocessing module is used to collect original optical signals and perform exploratory data analysis on the original optical signals to obtain preprocessed data; a data compression module, configured to input the preprocessed data into an autoencoder, and compress the preprocessed data into low-dimensional latent features through the autoencoder; A model training module, configured to input the low-dimensional potential features and historical channel state information labels into an initial random forest model for model training to obtain a random forest model for estimating channel state information; The information estimation module is used to input the optical signal to be measured into the random forest model to obtain a channel state information estimation value.

8. The channel state information estimation device according to claim 7, wherein: The preprocessing module is further used for: Determining outliers in the original optical signal using an interquartile range method; Abnormal values ​​in the original optical signal are eliminated to obtain cleaned data, and correlation analysis is performed on the cleaned data. Parameters in the cleaned data with correlation coefficients greater than a preset value are used as preprocessed data.

9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the channel state information estimation method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that include: A stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to perform the channel state information estimation method according to any one of claims 1 to 6.

Images