Polarization mode dispersion measurement method, device, equipment, storage medium and product thereof

Abstract

The application discloses a polarization mode dispersion measurement method and device, equipment, storage medium and product thereof, and relates to the technical field of polarization measurement. The method comprises the following steps: acquiring interference spectrum data of a to-be-measured optical fiber based on a fixed analyzer; performing dimension reduction processing on the interference spectrum data to obtain initial feature data containing a feature vector of the to-be-measured optical fiber, and determining a credibility score representing spectral physical consistency based on the interference spectrum data; adaptively reweighting the initial feature data based on the credibility score to generate spectral feature data; inputting the spectral feature data into a preset back propagation neural network to obtain a polarization mode dispersion value of the to-be-measured optical fiber, wherein the back propagation neural network is obtained based on a nonlinear mapping relationship between the spectral feature and the polarization mode dispersion value. While retaining the real physical signal features, the influence of the pseudo structure caused by the link noise is automatically weakened, so that the measurement precision of the polarization mode dispersion of the optical fiber is improved.

Description

Technical Field

[0001] This application relates to the field of polarization measurement technology, and in particular to polarization mode dispersion measurement methods, apparatus, equipment, storage media and products thereof. Background Technology

[0002] Optical fiber communication systems are widely used in various communication fields due to their high speed, low latency, high capacity, and anti-interference capabilities. However, with the explosive growth of network traffic, the requirements for optical fiber communication systems are also increasing. Among these requirements, the polarization mode dispersion effect in optical fibers has become one of the main factors affecting the performance of optical fiber communication.

[0003] Currently, polarization mode dispersion (PMD) is mainly measured using a fixed analyzer method, which calculates the PMD by measuring and analyzing the number of extreme points in the spectrum. However, the spectral signal obtained by the fixed analyzer method is highly susceptible to noise interference in the communication link, leading to many spurious extreme points and thus reducing the accuracy of PMD measurement.

[0004] The above content is only used to help understand the technical solution of this application and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this application is to provide a method, apparatus, device, storage medium, and product for measuring polarization mode dispersion, aiming to solve the technical problem that the measurement of polarization mode dispersion is prone to many spurious extreme points, thereby reducing the measurement accuracy of polarization mode dispersion of optical fibers.

[0006] To achieve the above objectives, this application proposes a method for measuring polarization mode dispersion, the method comprising: Acquire interferometric spectral data of the optical fiber under test based on the data captured by a fixed analyzer; The interference spectral data is subjected to dimensionality reduction processing to obtain initial feature data containing the feature vector of the optical fiber under test, and a confidence score characterizing the physical consistency of the spectrum is determined based on the interference spectral data. Based on the confidence score, the initial feature data is adaptively reweighted to generate spectral feature data; The spectral feature data is input into a preset backpropagation neural network to obtain the polarization mode dispersion value of the optical fiber under test. The backpropagation neural network is trained based on the nonlinear mapping relationship between spectral features and polarization mode dispersion value.

[0007] In one embodiment, the step of performing dimensionality reduction processing on the interferometric spectral data to obtain initial feature data containing the feature vector of the optical fiber under test includes: Based on the interference spectral data, the sensitivity weight of each wavelength point in the optical fiber under test to polarization mode dispersion is determined. Based on the aforementioned sensitivity weights, a weighted covariance matrix is ​​constructed; The weighted covariance matrix is ​​subjected to eigenvalue decomposition to obtain the feature vector with the ranking of the sensitivity weight as the preset ranking; The interference spectral data is projected onto the subspace spanned by the feature vectors to obtain initial feature data containing the feature vectors of the optical fiber under test.

[0008] In one embodiment, the step of determining the sensitivity weight of each wavelength point in the optical fiber under test to polarization mode dispersion based on the interference spectral data includes: Perform a Fourier transform on the interference spectral data to obtain a complex spectrum; The phase information of the optical fiber under test is extracted from the complex spectrum; Based on the phase information, the second derivative of the phase with respect to frequency of the optical fiber under test is determined; Based on the second derivative, the wavelength regions in the interference spectral data that do not conform to the physical laws of polarization interference are attenuated to obtain the corrected target spectral data. Based on the target spectral data, the sensitivity weights of each wavelength point in the optical fiber under test to polarization mode dispersion are determined.

[0009] In one embodiment, the step of adaptively reweighting the initial feature data based on the confidence score to generate spectral feature data includes: Dynamic weighting factors are generated based on the credibility score; Based on the dynamic weighting factor, the feature components in each dimension of the initial feature data are scaled dimension by dimension to generate reweighted spectral feature data.

[0010] In one embodiment, prior to the step of acquiring the interferometric spectral data of the optical fiber under test captured by the fixed analyzer, the method further includes: The network structure parameters and training parameters of the required backpropagation neural network are automatically determined based on the Bayesian optimization algorithm, and an initial neural network is constructed based on the network structure parameters and the training parameters. The network structure parameters include the number of hidden layer neurons, and the training parameters include the learning rate and the dropout rate. With the goal of minimizing the polarization mode dispersion prediction error, the initial neural network is iteratively trained to obtain the backpropagation neural network.

[0011] In one embodiment, the step of iteratively training the initial neural network to obtain a backpropagation neural network with the objective of minimizing the polarization mode dispersion prediction error includes: The goal is to minimize the polarization mode dispersion prediction error, and the confidence value of the corresponding physical consistency is calculated based on the interference spectral data in the preset training samples. The prediction errors of each preset training sample are weighted according to the confidence value to obtain the weighted total loss; The network structure parameters and training parameters of the initial neural network are updated based on the weighted total loss to obtain the backpropagation neural network.

[0012] Furthermore, to achieve the above objectives, this application also proposes a polarization mode dispersion measurement device, which includes: The acquisition module is used to acquire interference spectral data of the optical fiber under test based on the data captured by the fixed analyzer. The processing module is used to perform dimensionality reduction processing on the interference spectral data to obtain initial feature data containing the feature vector of the optical fiber under test, and to determine a confidence score characterizing the physical consistency of the spectrum based on the interference spectral data. The weighting module is used to adaptively reweight the initial feature data based on the confidence score to generate spectral feature data; The measurement module is used to input the spectral feature data into a preset backpropagation neural network to obtain the polarization mode dispersion value of the optical fiber under test. The backpropagation neural network is trained based on the nonlinear mapping relationship between spectral features and polarization mode dispersion value.

[0013] In addition, to achieve the above objectives, this application also proposes a polarization mode dispersion measurement device, the device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the polarization mode dispersion measurement method as described above.

[0014] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the polarization mode dispersion measurement method described above.

[0015] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the polarization mode dispersion measurement method described above.

[0016] One or more technical solutions proposed in this application have at least the following technical effects: Because highly linearly correlated redundant features often exist between adjacent sampling points in spectral data, dimensionality reduction is performed on the interferometric spectral data captured by a fixed analyzer when measuring fiber polarization mode dispersion (PSD). This extracts initial feature vectors characterizing fiber properties, preventing redundant features from obscuring key features that determine PSD values. A confidence score reflecting spectral physical consistency is calculated, and adaptive reweighting is applied to the initial feature data based on this score to suppress the contribution of noise interference regions, generating more robust spectral feature data. This spectral feature data is then input into a fully trained backpropagation neural network to output PSD values, avoiding sensitivity to noise and measurement instability caused by directly relying on the number of original spectral extrema. In other words, a data-driven approach learns the nonlinear mapping relationship between the overall spectral morphology and PSD, while introducing a confidence-guided feature reweighting mechanism. This automatically weakens the pseudo-structure effects caused by link noise while preserving the true physical signal characteristics. Furthermore, the combination of dimensionality reduction and neural networks enhances the resolution of weak dispersion signals under complex interference environments, thereby improving the measurement accuracy of fiber PSD. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating an embodiment of the polarization mode dispersion measurement method of this application. Figure 2 This is a flowchart illustrating Embodiment 2 of the polarization mode dispersion measurement method of this application. Figure 3 This is a flowchart illustrating Embodiment 3 of the polarization mode dispersion measurement method of this application. Figure 4 This is a schematic diagram of the module structure of the polarization mode dispersion measurement device according to an embodiment of this application; Figure 5 This is a schematic diagram of the device structure of the hardware operating environment involved in the polarization mode dispersion measurement method in the embodiments of this application.

[0019] Explanation of reference numerals in the attached figures: Processing device 1001, ROM 1002, storage device 1003, RAM 1004, bus 1005, I / O interface 1006, input device 1007, output device 1008, communication device 1009.

[0020] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0021] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0022] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0023] It should be noted that the executing entity in this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or polarization mode dispersion measurement platform capable of performing the above functions. The following description uses an optical fiber analysis platform as an example to illustrate this embodiment and the subsequent embodiments.

[0024] Based on this, embodiments of this application provide a method for measuring polarization mode dispersion, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the polarization mode dispersion measurement method of this application.

[0025] In this embodiment, the polarization mode dispersion measurement method includes steps S10~S40: Step S10: Obtain the interference spectral data of the optical fiber under test captured by the fixed analyzer; It should be noted that polarization mode dispersion (PMD) is a phenomenon in single-mode optical fibers where two orthogonal polarization modes propagate at different group velocities due to factors such as non-ideal core geometry or stress asymmetry, resulting in pulse broadening. Interference spectral data is a discrete or continuous sequence of light intensity changes with wavelength recorded in the wavelength domain (or frequency domain) after interferometry measurements of the output light from the fiber under test using a fixed analyzer. This data can include information on polarization interference fringes caused by fiber birefringence, which can be used to invert polarization mode dispersion characteristics.

[0026] Understandably, since the optical path parameters of a fixed analyzer remain constant during the measurement process, it is less sensitive to external interferences such as environmental vibration and temperature drift, which can effectively improve the stability and repeatability of a single measurement. Therefore, a fixed analyzer is preferred for obtaining the interference spectrum data of the optical fiber under test during polarization mode dispersion measurement.

[0027] Step S20: Perform dimensionality reduction processing on the interference spectral data to obtain initial feature data containing the feature vector of the optical fiber under test, and determine the confidence score characterizing the physical consistency of the spectrum based on the interference spectral data. It should be noted that the process of mapping high-dimensional interferometric spectral data (such as thousands of wavelength points) to a low-dimensional feature space through mathematical transformation aims to retain key information related to polarization mode dispersion while suppressing redundancy and noise. Eigenvectors are orthogonal basis vectors obtained by eigenvalue decomposition of the weighted covariance matrix. The subspace spanned by these eigenvectors effectively characterizes the spectral variation patterns related to polarization mode dispersion in the fiber under test, and each eigenvector corresponds to a principal component direction. Initial feature data includes low-dimensional vectors obtained by projecting the original interferometric spectral data onto the aforementioned eigenvector subspace. Spectral physical consistency refers to whether the interferometric spectrum conforms to the fundamental physical laws of polarization interference; for example, the phase should be a smooth function of frequency, the group delay dispersion should satisfy causality, or the interference fringes should have reasonable contrast. A confidence score is used to quantify the reliability of the interferometric spectral data in terms of physical consistency; a higher score indicates that the spectrum conforms more closely to the theoretical model of polarization interference and is more suitable as a basis for polarization mode dispersion estimation.

[0028] Understandably, since interferometric spectral data has high dimensionality (e.g., thousands of wavelength sampling points), directly inputting it into a neural network would cause huge computational overhead and could easily lead to overfitting of the neural network model due to redundant information and noise. Therefore, by performing dimensionality reduction on the interferometric spectral data, especially by constructing a weighted covariance matrix based on the sensitivity weights of each wavelength point to polarization mode dispersion and performing eigenvalue decomposition, the interferometric spectral data can be projected into a low-dimensional subspace spanned by the most discriminative physical features. This can both retain the effective information strongly correlated with polarization mode dispersion and significantly compress the amount of data, thereby improving the processing efficiency of the backpropagation neural network.

[0029] Understandably, in actual measurement environments, the acquired interference spectra may deviate from the ideal physical laws of polarization interference due to light source fluctuations, detector noise, or fiber micro-perturbations. Therefore, by analyzing the phase characteristics (such as the second derivative) of the interference spectrum to evaluate its physical consistency and generating a quantitative index (confidence score), the reliability of the current measurement data can be objectively assessed. When the spectral data is distorted due to external interference, a low confidence score can warn the fiber optic analysis platform that the measurement result may have a large deviation, thereby avoiding the use of unreliable data for the final calculation of polarization mode dispersion values ​​and effectively improving the accuracy of the entire measurement scheme under non-ideal conditions.

[0030] Optionally, if the measuring device integrates a temperature sensor and a vibration sensor, the final confidence score can be dynamically adjusted when a drastic change in ambient temperature or significant mechanical vibration is detected. This ensures that the confidence score not only reflects the spectral quality but also takes into account external environmental interference factors, thereby maintaining measurement stability in more complex field environments.

[0031] In practical implementation, the data can first undergo a Fourier transform to extract the complex spectrum and calculate the phase information from it. Then, based on the phase information, its second derivative with respect to frequency is calculated, and according to the magnitude of the second derivative, wavelength regions in the original interference spectrum that do not conform to the physical laws of polarization interference are attenuated to obtain the corrected target spectral data. On this basis, the sensitivity weight of each wavelength point to polarization mode dispersion is determined, and a weighted covariance matrix is ​​constructed using this weight. By performing eigenvalue decomposition on this matrix, the top-ranked preset number of eigenvectors are selected, and the original interference spectral data is projected onto the subspace spanned by the preset number of eigenvectors, thereby obtaining an initial feature data of a preset number of dimensions. At the same time, based on the smoothness and interference contrast of the corrected target spectral data, a confidence score characterizing its physical consistency is calculated.

[0032] Step S30: Adaptively reweight the initial feature data based on the confidence score to generate spectral feature data; It should be noted that adaptive reweighting dynamically adjusts the weights of each dimension in the initial feature data based on the confidence score, thereby enhancing the feature components corresponding to high-confidence initial feature data and suppressing low-confidence initial feature data, thus improving the quality of neural network input.

[0033] Understandably, the adaptive reweighting operation is equivalent to pre-screening and enhancing the features of the input backpropagation neural network based on physical prior knowledge. This can effectively suppress the negative impact of feature components caused by abnormal spectra on the final prediction results, while strengthening the feature contributions from spectral regions with high reliability and clear physical laws. This makes the generated spectral feature data more accurately reflect the inherent polarization mode dispersion characteristics of the fiber under test, thereby improving the measurement accuracy of the polarization mode dispersion of the fiber.

[0034] In a practical implementation, a dynamic weighting factor can be generated based on the confidence score (e.g., taking the square of the confidence score), and this factor can be used to scale each dimension of the initial feature data one dimension at a time, thereby generating a reweighted spectral feature data that better reflects the true characteristics of the optical fiber.

[0035] Furthermore, step S30 also includes: Dynamic weighting factors are generated based on the credibility score; Based on the dynamic weighting factor, the feature components in each dimension of the initial feature data are scaled dimension by dimension to generate reweighted spectral feature data.

[0036] It should be noted that the dynamic weighting factor is a scaling coefficient calculated in real time based on the confidence score. It is used to adjust the intensity of subsequent feature processing, and the factor is not a fixed constant but adaptively adjusts according to the changes in the spectral quality obtained from each measurement, reflecting the system's responsiveness to data reliability. Dimensional scaling independently multiplies each dimension (i.e., each feature component) in the initial feature data by the same dynamic weighting factor, thereby achieving a uniform amplitude adjustment of the entire feature vector, rather than applying differentiated weights to different dimensions.

[0037] Understandably, by generating dynamic weighting factors based on the confidence score and scaling the initial feature data dimension by dimension, the quality of the input features is explicitly encoded. That is, high-quality spectra correspond to high-amplitude features, and low-quality spectra correspond to low-amplitude features. This allows the subsequent backpropagation neural network to reasonably perceive the reliability of the input data and make more robust predictions.

[0038] Optionally, to enhance the flexibility of the reweighting mechanism, the single dynamic weight factor can be extended into a dimension-dependent weight vector. Specifically, the weight vector is constructed by combining the sensitivity of each feature component to polarization mode dispersion prediction in the training set (e.g., evaluated by gradient magnitude or SHAP value), thereby accurately distinguishing the reliability of different feature dimensions.

[0039] Step S40: Input the spectral feature data into a preset backpropagation neural network to obtain the polarization mode dispersion value of the optical fiber under test. The backpropagation neural network is trained based on the nonlinear mapping relationship between spectral features and polarization mode dispersion value.

[0040] It should be noted that the structure and hyperparameters of the backpropagation neural network are automatically determined by Bayesian optimization. The nonlinear mapping relationship can be understood as the absence of a simple linear proportional relationship between the interference spectral characteristics and the polarization mode dispersion value, instead being influenced by a combination of factors such as fiber length, temperature, and stress distribution.

[0041] It is understandable that there is a complex and highly nonlinear physical coupling relationship between the polarization mode dispersion of optical fiber and the interference spectral characteristics of optical fiber. The backpropagation neural network can learn a large amount of sample data, automatically mine and fit this complex mapping relationship, thereby achieving high-precision end-to-end prediction from spectral features to polarization mode dispersion values. Therefore, by inputting spectral feature data into a pre-trained neural network, accurate polarization mode dispersion values ​​can be output in a very short time.

[0042] Understandably, inputting the more representative and reliable spectral feature data generated after physical consistency evaluation and adaptive reweighting into the backpropagation neural network can efficiently and accurately achieve end-to-end mapping from high-quality spectral features to polarization mode dispersion values, thereby avoiding cumbersome iterative calculations or approximate assumptions and improving the measurement accuracy of polarization mode dispersion of optical fibers.

[0043] In the specific implementation, a backpropagation neural network is used to learn the nonlinear mapping relationship between the dimensionality-reduced spectral features and the polarization mode dispersion value. The number of neurons in the input layer of the backpropagation neural network is equal to the dimension of the spectral data after dimensionality reduction by principal component analysis. The backpropagation neural network contains multiple hidden layers, each containing several neurons, and uses activation functions for nonlinear transformation. To prevent overfitting, dropout rate regularization is applied after the output of the hidden layers. The output layer of the backpropagation neural network contains one neuron to output the predicted polarization mode dispersion value.

[0044] Alternatively, by linking the design of dynamic weight factors with the design of the neural network's loss function, physical confidence can be introduced as a priori for sample importance during the offline training phase. Higher penalty weights are applied to low-confidence samples, forcing the neural network to prioritize the prediction accuracy of high-confidence samples during optimization. This allows the network to learn a robust nonlinear mapping function that is closer to real physical laws. During the online inference phase, the low-amplitude features corresponding to low confidence naturally guide the network to output more uncertain results. This forms a closed-loop physical constraint internalization mechanism, further strengthening the internalization of physical constraints in the backpropagation neural network.

[0045] This embodiment provides a method for measuring polarization mode dispersion. Since there are usually highly linearly correlated redundant features between adjacent sampling points in spectral data, when measuring the polarization mode dispersion of an optical fiber, the interferometric spectral data captured by a fixed analyzer is dimensionality-reduced to extract an initial feature vector characterizing the fiber properties. This avoids redundant features from obscuring key features that play a decisive role in the polarization mode dispersion value. A confidence score reflecting the physical consistency of the spectrum is calculated, and then the initial feature data is adaptively reweighted based on this confidence score to suppress the contribution of noise interference regions, generating more robust spectral feature data. This spectral feature data is then input into a fully trained backpropagation neural network to output the polarization mode dispersion value, avoiding the sensitivity to noise and unstable measurement results caused by directly relying on the number of extreme points in the original spectrum. That is, by learning the nonlinear mapping relationship between the overall spectral morphology and polarization mode dispersion through a data-driven approach, and introducing a credibility-guided feature reweighting mechanism, the pseudo-structure influence caused by link noise is automatically weakened while preserving the true physical signal characteristics. Furthermore, the combination of dimensionality reduction and neural network enhances the ability to resolve weak dispersion signals under complex interference environments, thereby improving the measurement accuracy of polarization mode dispersion of optical fibers.

[0046] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 Step S20 also includes steps S01 to S04: Step S01: Based on the interference spectral data, determine the sensitivity weight of each wavelength point in the optical fiber under test to polarization mode dispersion. Step S02: Construct a weighted covariance matrix based on the sensitivity weights; Step S03: Perform eigenvalue decomposition on the weighted covariance matrix to obtain the feature vector of the sensitivity weight ranking as a preset ranking; Step S04: Project the interference spectral data onto the subspace spanned by the feature vector to obtain initial feature data containing the feature vector of the optical fiber under test.

[0047] It should be noted that the sensitivity weight is a weight vector with a one-to-one correspondence to wavelength, used to quantify the sensitivity of the spectral response at a specific wavelength to changes in polarization mode dispersion. A larger weight value indicates that the spectral information in that wavelength region is more discriminative for estimating polarization mode dispersion. The weighted covariance matrix is ​​a modified covariance matrix constructed by introducing the aforementioned sensitivity weight as a priori when calculating the covariance matrix of the interference spectral data, assigning different importance to different wavelength dimensions. Eigenvalue decomposition involves performing eigenvalue decomposition on the weighted covariance matrix to obtain a set of orthogonal eigenvectors. The subspace can approximate the original high-dimensional spectral data with minimal information loss.

[0048] Understandably, since different spectral bands contribute differently to polarization mode dispersion, by introducing a weighting mechanism based on physical sensitivity, the construction process of the covariance matrix can incorporate prior knowledge related to polarization mode dispersion. This ensures that the extracted principal component directions are more focused on the spectral regions most valuable for polarization mode dispersion discrimination, and selects high-ranking feature vectors. This significantly reduces data dimensionality while maximizing the retention of effective information strongly correlated with polarization mode dispersion, thereby reducing the input complexity and overfitting risk of subsequent neural networks and improving overall measurement efficiency and accuracy.

[0049] Understandably, projecting interferometric spectral data onto a sensitivity-guided feature subspace allows each feature component to correspond to a typical spectral variation pattern sensitive to polarization mode dispersion. This can be represented by physically interpretable features, providing highly relevant input data for confidence assessment, reweighting, and neural network inference, thereby improving the measurement accuracy of polarization mode dispersion in optical fibers.

[0050] Furthermore, step S01 also includes: Perform a Fourier transform on the interference spectral data to obtain a complex spectrum; The phase information of the optical fiber under test is extracted from the complex spectrum; Based on the phase information, the second derivative of the phase with respect to frequency of the optical fiber under test is determined; Based on the second derivative, the wavelength regions in the interference spectral data that do not conform to the physical laws of polarization interference are attenuated to obtain the corrected target spectral data. Based on the target spectral data, the sensitivity weights of each wavelength point in the optical fiber under test to polarization mode dispersion are determined.

[0051] It should be noted that the complex spectrum is a complex numerical frequency domain signal obtained after Fourier transform, containing both amplitude and phase information. Phase information, extracted from the complex spectrum, reflects the propagation delay characteristics of different frequency components in the fiber under test and is fundamental for calculating higher-order parameters such as group delay and dispersion. The second derivative of phase with respect to frequency is a physical quantity obtained by taking the second derivative of the phase spectrum with respect to frequency; in optics, it is proportional to group delay and dispersion. The wavelength region refers to areas in actual measurements where, due to factors such as noise, detector nonlinearity, light source fluctuations, or environmental disturbances, the second derivative of phase at certain wavelengths exhibits severe oscillations, singular values, or exceeds the theoretically reasonable range. Attenuation processing involves applying a weighting factor less than 1 to the spectral intensity of these unreliable wavelength regions to reduce their influence in subsequent processing, thereby improving overall data quality.

[0052] Understandably, since feature extraction using the raw interference spectrum is susceptible to noise and non-physical disturbances, it can easily lead to deviations in subsequent polarization mode dispersion estimation. Therefore, the physical signal processing chain of Fourier transform-phase extraction-second derivative analysis can accurately identify and locate abnormal wavelength regions that do not conform to the basic laws of polarization interference, providing a scientific basis for data cleaning and thus improving the accuracy of the final polarization mode dispersion measurement.

[0053] Understandably, attenuation processing of anomalous regions can reasonably preserve the overall structural integrity of the spectrum, effectively suppress the influence of local noise on global features, and avoid introducing new spectral leakage or artifacts. This prevents spectral leakage or artifacts from affecting the final measurement, thereby improving the accuracy of the final polarization mode dispersion measurement.

[0054] Understandably, calculating sensitivity weights based on the corrected target spectral data ensures that the weights themselves are built on high-quality, physically consistent data. This allows the subsequent weighted dimensionality reduction process to truly focus on reliable and sensitive spectral regions, avoiding misclassification of noisy regions as high-information regions, thereby improving the accuracy and robustness of feature extraction.

[0055] Optionally, the attenuation processing can be linked with the generation of the confidence score. That is, the higher the proportion of the abnormal area, the lower the final confidence score, forming a closed loop from physical anomaly detection to data quality scoring, thereby enhancing the overall consistency of the polarization mode dispersion measurement platform.

[0056] Based on the first and second embodiments of this application, the same or similar content as the above embodiments in the third embodiment of this application can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 3 Before step S10, the polarization mode dispersion measurement method further includes steps S1~S2: Step S1: Automatically determine the network structure parameters and training parameters of the required backpropagation neural network based on the Bayesian optimization algorithm, and construct the initial neural network based on the network structure parameters and the training parameters; The network structure parameters include the number of hidden layer neurons, and the training parameters include the learning rate and the dropout rate. Step S2: With the goal of minimizing the polarization mode dispersion prediction error, the initial neural network is iteratively trained to obtain the backpropagation neural network.

[0057] It's important to note that network structure parameters are configurable variables that determine the topology of a neural network, directly impacting the model's expressive power and complexity. Training parameters are configurable items that control the dynamic behavior of the neural network during training. They are not updated with forward / backward propagation but affect convergence speed and generalization performance, and can include the learning rate and dropout rate. The learning rate is a positive scalar used to control the step size of each parameter update during gradient descent optimization. The dropout rate is the probability of randomly dropping a certain percentage of neurons in the neural network to zero during training, used to prevent overfitting and improve generalization ability. Polarization mode dispersion prediction error is a measure of the difference between the predicted polarization mode dispersion value output by the neural network and the corresponding sample's true polarization mode dispersion label.

[0058] Understandably, since the performance of neural networks is highly dependent on the selection of hyperparameters, by introducing Bayesian optimization algorithms, it is possible to automatically and intelligently search for near-optimal combinations of network structure parameters and training parameters with fewer training trials, thereby improving the efficiency of model development and the ultimate performance ceiling.

[0059] Understandably, minimizing the polarization mode dispersion prediction error as a clear optimization objective allows the entire neural network construction and training process to be closely centered around the core measurement task, enabling customized modeling for specific physical quantities and effectively improving the accuracy and reliability of polarization mode dispersion measurements.

[0060] Furthermore, step S2 also includes: The goal is to minimize the polarization mode dispersion prediction error, and the confidence value of the corresponding physical consistency is calculated based on the interference spectral data in the preset training samples. The prediction errors of each preset training sample are weighted according to the confidence value to obtain the weighted total loss; The network structure parameters and training parameters of the initial neural network are updated based on the weighted total loss to obtain the backpropagation neural network.

[0061] It should be noted that the preset training samples are a set of labeled datasets prepared offline. Each sample contains interferometric spectral data acquired or simulated using a fixed analyzer, along with the corresponding true polarization mode dispersion values. The confidence value is a quantitative score calculated for the interferometric spectral data in each preset training sample, based on whether it conforms to the basic physical laws of polarization interference, and is used to characterize the reliability of the sample data. The weighted total loss is the overall loss function obtained by weighting and summing the prediction errors of each training sample according to their corresponding confidence values ​​during the neural network training process.

[0062] Understandably, since high-confidence samples are usually closer to the ideal measurement conditions in real-world application scenarios, focusing on optimizing high-confidence samples can directly improve the measurement accuracy of the backpropagation neural network under normal operating conditions. At the same time, because low-confidence samples have lower weights, the impact of anomalous modes on the model can be effectively suppressed, thereby enhancing the robustness of the backpropagation neural network in measuring polarization mode dispersion in complex or perturbed environments.

[0063] It should be noted that the above examples are only for understanding this application and do not constitute a limitation on the polarization mode dispersion measurement method of this application. Any simple transformations based on this technical concept are within the protection scope of this application.

[0064] This application also provides a polarization mode dispersion measurement device, please refer to... Figure 4 The polarization mode dispersion measurement device includes: The acquisition module 10 is used to acquire interference spectral data of the optical fiber under test captured by the fixed analyzer. The processing module 20 is used to perform dimensionality reduction processing on the interference spectral data to obtain initial feature data containing the feature vector of the optical fiber under test, and to determine a confidence score characterizing the physical consistency of the spectrum based on the interference spectral data. Weighting module 30 is used to adaptively reweight the initial feature data based on the confidence score to generate spectral feature data; The measurement module 40 is used to input the spectral feature data into a preset backpropagation neural network to obtain the polarization mode dispersion value of the optical fiber under test. The backpropagation neural network is trained based on the nonlinear mapping relationship between spectral features and polarization mode dispersion value.

[0065] Optionally, the processing module 20 is further configured to: determine the sensitivity weights of each wavelength point in the optical fiber under test to polarization mode dispersion based on the interference spectral data; construct a weighted covariance matrix based on the sensitivity weights; perform eigenvalue decomposition on the weighted covariance matrix to obtain eigenvectors whose sensitivity weights are ranked according to a preset ranking; and project the interference spectral data onto the subspace spanned by the eigenvectors to obtain initial feature data containing the eigenvectors of the optical fiber under test.

[0066] Optionally, the processing module 20 is further configured to perform a Fourier transform on the interference spectral data to obtain a complex spectrum; extract the phase information of the optical fiber under test from the complex spectrum; determine the second derivative of the phase of the optical fiber under test with respect to frequency based on the phase information; perform attenuation processing on the wavelength regions in the interference spectral data that do not conform to the physical laws of polarization interference according to the second derivative, to obtain corrected target spectral data; and determine the sensitivity weight of each wavelength point in the optical fiber under test to polarization mode dispersion based on the target spectral data.

[0067] Optionally, the weighting module 30 is further configured to generate dynamic weighting factors based on the confidence score; and to scale the feature components of each dimension in the initial feature data dimension by dimension based on the dynamic weighting factors to generate reweighted spectral feature data.

[0068] Optionally, the acquisition module 10 is further configured to automatically determine the network structure parameters and training parameters of the required backpropagation neural network based on the Bayesian optimization algorithm, and construct an initial neural network based on the network structure parameters and the training parameters; wherein the network structure parameters include the number of hidden layer neurons, and the training parameters include the learning rate and the dropout rate; with the goal of minimizing the polarization mode dispersion prediction error, the initial neural network is iteratively trained to obtain the backpropagation neural network.

[0069] Optionally, the acquisition module 10 is further configured to minimize the polarization mode dispersion prediction error, calculate the corresponding physical consistency confidence value based on the interference spectral data in the preset training samples; weight the prediction error of each preset training sample according to the confidence value to obtain the weighted total loss; and update the network structure parameters and training parameters of the initial neural network based on the weighted total loss to obtain the backpropagation neural network.

[0070] The polarization mode dispersion (PMD) measurement device provided in this application, employing the PMD measurement method described in the above embodiments, can solve the technical problem that PMD measurement is prone to producing many spurious extrema, thereby reducing the measurement accuracy of PMD in optical fibers. Compared with the prior art, the beneficial effects of the PMD measurement device provided in this application are the same as those of the PMD measurement method provided in the above embodiments, and other technical features in the PMD measurement device are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0071] This application provides a polarization mode dispersion measurement device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the polarization mode dispersion measurement method in the above embodiment 1.

[0072] The following is for reference. Figure 5 The diagram illustrates a structural schematic of a polarization mode dispersion measurement device suitable for implementing embodiments of this application. The polarization mode dispersion measurement device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, personal digital assistants, tablets, PMP portable multimedia players, and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The polarization mode dispersion measurement device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0073] like Figure 5 As shown, the polarization mode dispersion measurement device may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in ROM (Read Only Memory) 1002 or a program loaded from storage device 1003 into RAM (Random Access Memory) 1004. RAM 1004 also stores various programs and data required for the operation of the polarization mode dispersion measurement device. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. I / O (input / output) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touch screens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the polarization mode dispersion measurement device to communicate wirelessly or wiredly with other devices to exchange data. Although the figure shows polarization mode dispersion measurement devices with various systems, it should be understood that it is not required to implement or possess all of the systems shown. More or fewer systems may be implemented alternatively.

[0074] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0075] The polarization mode dispersion measurement device provided in this application, employing the polarization mode dispersion measurement method described in the above embodiments, can solve the technical problem that the measurement of polarization mode dispersion is prone to many spurious extrema, thereby reducing the measurement accuracy of polarization mode dispersion of optical fibers. Compared with the prior art, the beneficial effects of the polarization mode dispersion measurement device provided in this application are the same as those of the polarization mode dispersion measurement method provided in the above embodiments, and other technical features of this polarization mode dispersion measurement device are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0076] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0077] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0078] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, the computer-readable program instructions being used to execute the polarization mode dispersion measurement method in the above embodiments.

[0079] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0080] The aforementioned computer-readable storage medium may be included in the polarization mode dispersion measurement device; or it may exist independently and not assembled into the polarization mode dispersion measurement device.

[0081] The aforementioned computer-readable storage medium carries one or more programs that, when executed by a polarization mode dispersion measurement device, cause the polarization mode dispersion measurement device to: acquire interference spectral data of the optical fiber under test captured by a fixed analyzer; perform dimensionality reduction processing on the interference spectral data to obtain initial feature data containing the feature vector of the optical fiber under test, and determine a confidence score characterizing the physical consistency of the spectrum based on the interference spectral data; adaptively reweight the initial feature data based on the confidence score to generate spectral feature data; and input the spectral feature data into a preset backpropagation neural network to obtain the polarization mode dispersion value of the optical fiber under test, wherein the backpropagation neural network is trained based on the nonlinear mapping relationship between spectral features and polarization mode dispersion values.

[0082] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. These programming languages ​​include object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a Local Area Network (LAN) or a Wide Area Network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0083] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0084] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0085] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described polarization mode dispersion measurement method. This solves the technical problem that the measurement of polarization mode dispersion is prone to producing many spurious extrema, thereby reducing the measurement accuracy of the polarization mode dispersion of optical fibers. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the polarization mode dispersion measurement method provided in the above embodiments, and will not be repeated here.

[0086] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the polarization mode dispersion measurement method as described above.

[0087] The computer program product provided in this application can solve the technical problem that the measurement of polarization mode dispersion is prone to many spurious extrema, thereby reducing the measurement accuracy of polarization mode dispersion of optical fibers. Compared with the prior art, the beneficial effects of the computer program product provided in this application are the same as those of the polarization mode dispersion measurement method provided in the above embodiments, and will not be repeated here.

[0088] The above description is only a part of the embodiments of this application and does not limit the scope of protection of this application. All equivalent structural transformations made under the technical concept of this application and using the content of this application specification and drawings, or direct / indirect applications in other related technical fields, are included in the scope of protection of this application.

Claims

1. A method for measuring polarization mode dispersion, characterized in that, The method includes: Acquire interferometric spectral data of the optical fiber under test based on the data captured by a fixed analyzer; The interference spectral data is subjected to dimensionality reduction processing to obtain initial feature data containing the feature vector of the optical fiber under test, and a confidence score characterizing the physical consistency of the spectrum is determined based on the interference spectral data. Based on the confidence score, the initial feature data is adaptively reweighted to generate spectral feature data; The spectral feature data is input into a preset backpropagation neural network to obtain the polarization mode dispersion value of the optical fiber under test. The backpropagation neural network is trained based on the nonlinear mapping relationship between spectral features and polarization mode dispersion value.

2. The method as described in claim 1, characterized in that, The step of performing dimensionality reduction processing on the interferometric spectral data to obtain initial feature data containing the feature vector of the optical fiber under test includes: Based on the interference spectral data, the sensitivity weight of each wavelength point in the optical fiber under test to polarization mode dispersion is determined. Based on the aforementioned sensitivity weights, a weighted covariance matrix is ​​constructed; The weighted covariance matrix is ​​subjected to eigenvalue decomposition to obtain the feature vector with the ranking of the sensitivity weight as the preset ranking; The interference spectral data is projected onto the subspace spanned by the feature vectors to obtain initial feature data containing the feature vectors of the optical fiber under test.

3. The method as described in claim 2, characterized in that, The step of determining the sensitivity weight of each wavelength point in the optical fiber under test to polarization mode dispersion based on the interference spectral data includes: Perform a Fourier transform on the interference spectral data to obtain a complex spectrum; The phase information of the optical fiber under test is extracted from the complex spectrum; Based on the phase information, the second derivative of the phase with respect to frequency of the optical fiber under test is determined; Based on the second derivative, the wavelength regions in the interference spectral data that do not conform to the physical laws of polarization interference are attenuated to obtain the corrected target spectral data. Based on the target spectral data, the sensitivity weights of each wavelength point in the optical fiber under test to polarization mode dispersion are determined.

4. The method as described in claim 1, characterized in that, The step of adaptively reweighting the initial feature data based on the confidence score to generate spectral feature data includes: Dynamic weighting factors are generated based on the credibility score; Based on the dynamic weighting factor, the feature components in each dimension of the initial feature data are scaled dimension by dimension to generate reweighted spectral feature data.

5. The method as described in claim 1, characterized in that, Before the step of acquiring the interferometric spectral data of the optical fiber under test captured by the fixed analyzer, the method further includes: The network structure parameters and training parameters of the required backpropagation neural network are automatically determined based on the Bayesian optimization algorithm, and an initial neural network is constructed based on the network structure parameters and the training parameters. The network structure parameters include the number of hidden layer neurons, and the training parameters include the learning rate and the dropout rate. With the goal of minimizing the polarization mode dispersion prediction error, the initial neural network is iteratively trained to obtain the backpropagation neural network.

6. The method as described in claim 5, characterized in that, The step of iteratively training the initial neural network to obtain the backpropagation neural network with the objective of minimizing the polarization mode dispersion prediction error includes: The goal is to minimize the polarization mode dispersion prediction error, and the confidence value of the corresponding physical consistency is calculated based on the interference spectral data in the preset training samples. The prediction errors of each preset training sample are weighted according to the confidence value to obtain the weighted total loss; The network structure parameters and training parameters of the initial neural network are updated based on the weighted total loss to obtain the backpropagation neural network.

7. A polarization mode dispersion measuring device, characterized in that, The device includes: The acquisition module is used to acquire interference spectral data of the optical fiber under test based on the data captured by the fixed analyzer. The processing module is used to perform dimensionality reduction processing on the interference spectral data to obtain initial feature data containing the feature vector of the optical fiber under test, and to determine a confidence score characterizing the physical consistency of the spectrum based on the interference spectral data. The weighting module is used to adaptively reweight the initial feature data based on the confidence score to generate spectral feature data; The measurement module is used to input the spectral feature data into a preset backpropagation neural network to obtain the polarization mode dispersion value of the optical fiber under test. The backpropagation neural network is trained based on the nonlinear mapping relationship between spectral features and polarization mode dispersion value.

8. A polarization mode dispersion measurement device, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the polarization mode dispersion measurement method as described in any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the polarization mode dispersion measurement method as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the steps of the polarization mode dispersion measurement method as described in any one of claims 1 to 6.

Images