FTTR gateway automatic modulation identification method, system, device and storage medium

Abstract

The present application relates to a method, system, device and storage medium for automatic modulation identification of an FTTR gateway. The method for automatic modulation identification of an FTTR gateway includes: collecting historical signal data from an FTTR network; preprocessing the historical signal data; extracting features from the preprocessed historical signal data; training a deep learning model based on historical signal features; integrating the trained deep learning model into the signal processing system of the FTTR network; obtaining real-time signal data in the FTTR network and preprocessing the real-time signal data; inputting the preprocessed real-time signal data into the deep learning model for inference to obtain an identification result; and selecting an appropriate demodulation strategy for demodulation based on the identification result. The present application can not only improve the accuracy and stability of signal transmission, reduce the bit error rate caused by signal error processing, but also indirectly reduce the delay caused by signal error processing.

Description

Technical Field

[0001] The present application relates to the technical field of FTTR gateways, and in particular to a method, system, device, and storage medium for automatic modulation identification of FTTR gateways. Background Art

[0002] In FTTR networks, the conversion of optical signals transmitted via optical fibers into electrical signals is a critical step. During this process, the signals may be subject to various interferences, such as electromagnetic interference and thermal noise. For example, electromagnetic interference (EMI) can occur when electromagnetic waves generated by surrounding electronic devices interact with the transmitted signal, causing changes in signal characteristics such as amplitude and phase, thereby degrading signal quality. Traditional modulation recognition techniques face numerous challenges in FTTR network technology. Currently, existing technologies in this field primarily rely on traditional signal processing and machine learning methods for modulation recognition. For example, some methods rely on fixed feature extraction algorithms and simple classifiers to identify modulation types. However, these methods suffer from severe inaccuracies in low signal-to-noise ratio environments. Summary of the Invention

[0003] Based on this, it is necessary to provide an FTTR gateway automatic modulation identification method, system, device and storage medium to address the problems in the related technology.

[0004] In order to achieve the above objectives, in a first aspect, the present application provides a method for automatic modulation identification of an FTTR gateway, the method comprising:

[0005] Collecting historical signal data from the FTTR network, the historical signal data including historical optical signal data and historical electrical signal data of different modulation modes;

[0006] Preprocessing the historical signal data;

[0007] Perform feature extraction on the pre-processed historical signal data to obtain historical signal features;

[0008] Training a deep learning model based on the historical signal features;

[0009] Integrate the trained deep learning model into the signal processing system of the FTTR network;

[0010] Acquiring real-time signal data in the FTTR network and preprocessing the real-time signal data;

[0011] Inputting the preprocessed real-time signal data into a deep learning model integrated into the signal processing system for reasoning to obtain a recognition result;

[0012] An appropriate demodulation strategy is selected for demodulation based on the recognition result.

[0013] In some embodiments, the historical signal data is annotated;

[0014] Perform data enhancement on the annotated historical data signals.

[0015] In some embodiments, feature extraction is performed on the pre-processed historical signal data to obtain historical signal features, including:

[0016] Use Fourier transform to convert the pre-processed historical signal data from time domain signal to frequency domain signal;

[0017] Extracting time-frequency features of the frequency domain signal using adaptive wavelet transform;

[0018] performing unsupervised feature learning using an autoencoder to reconstruct an original signal based on the frequency domain signal;

[0019] Converting the original signal into a time-frequency graph;

[0020] The time-frequency graph is subjected to image enhancement processing.

[0021] In some embodiments, training a deep learning model based on the historical signal features includes:

[0022] Build a deep learning model, including: selecting a model architecture; setting the input and output layers to obtain a deep learning model;

[0023] Training the deep learning model using the historical signal features;

[0024] Evaluate and optimize trained deep learning models.

[0025] In some embodiments, selecting an appropriate demodulation strategy for demodulation based on the recognition result includes:

[0026] A dynamic demodulation strategy or an adaptive demodulation strategy is selected for demodulation based on the identification result.

[0027] In some embodiments, after selecting an appropriate demodulation strategy for demodulation based on the recognition result, the method further includes:

[0028] Demodulation parameters are dynamically adjusted based on characteristics of the real-time signal data.

[0029] In some embodiments, the identification result includes a demodulation type; after dynamically adjusting the demodulation parameters based on the characteristics of the real-time signal data, the method further includes:

[0030] Optimizing the deep learning model based on the recognition results and demodulated data;

[0031] The demodulation strategy is adjusted based on the recognition results obtained from the optimized deep learning model.

[0032] In a second aspect, the present application also provides an FTTR gateway automatic modulation identification system, comprising:

[0033] Data collection module, used to collect historical signal data and real-time signal data in the FTTR network;

[0034] A preprocessing module, configured to preprocess the historical signal data and the real-time signal data;

[0035] A feature extraction module is used to extract features from the pre-processed historical signal data to obtain historical signal features;

[0036] A model training module, configured to train a deep learning model based on the historical signal features;

[0037] The signal processing system is used to integrate the trained deep learning model and perform inference based on the input pre-processed real-time signal data to obtain recognition results;

[0038] The demodulation module is used to select an appropriate demodulation strategy for demodulation based on the recognition result.

[0039] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the FTTR gateway automatic modulation identification method as described in the first aspect are implemented.

[0040] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the FTTR gateway automatic modulation identification method as described in the first aspect.

[0041] The above-mentioned FTTR gateway automatic modulation identification method, system, device and storage medium, based on the deep learning model, can accurately identify and classify the interference-received signals. By accurately identifying the modulation type of the signal, the received signal can be better understood, so that the appropriate demodulation strategy can be adopted for demodulation; this can not only improve the accuracy and stability of signal transmission, reduce the bit error rate caused by signal error processing, but also indirectly reduce the delay caused by signal error processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the conventional technology, the following briefly introduces the drawings required for use in the embodiments or the conventional technology descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 This is a flow chart of the FTTR gateway automatic modulation identification method provided in one embodiment of the present application;

[0044] Figure 2 This is a structural block diagram of an FTTR gateway automatic modulation identification system provided in another embodiment of the present application;

[0045] Figure 3 This is a diagram of the internal structure of a computer device provided in another embodiment of the present application.

[0046] Explanation of the accompanying drawings: 10. Data collection module; 20. Preprocessing module; 30. Feature extraction module; 40. Model training module; 50. Signal processing system; 60. Demodulation module. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0048] In one embodiment, see Figure 1 The present application provides an FTTR gateway automatic modulation identification method, which includes the following steps: S10~S80.

[0049] S10: Collecting historical signal data from the FTTR network, where the historical signal data includes historical optical signal data and historical electrical signal data in different modulation modes.

[0050] S20: Preprocessing the historical signal data.

[0051] S30: Perform feature extraction on the pre-processed historical signal data to obtain historical signal features.

[0052] S40: Training a deep learning model based on the historical signal features.

[0053] S50: Integrate the trained deep learning model into the signal processing system of the FTTR network.

[0054] S60: Acquire real-time signal data in the FTTR network and pre-process the real-time signal data.

[0055] S70: Inputting the pre-processed real-time signal data into a deep learning model integrated into the signal processing system for reasoning to obtain a recognition result.

[0056] S80: Selecting an appropriate demodulation strategy for demodulation based on the recognition result.

[0057] In the FTTR gateway automatic modulation identification method of the present application, based on the deep learning model, it is possible to accurately identify and classify the interference-received signals. By accurately identifying the modulation type of the signal, the received signal can be better understood, so that an appropriate demodulation strategy can be adopted for demodulation. This can not only improve the accuracy and stability of signal transmission, reduce the bit error rate caused by signal error processing, but also indirectly reduce the delay caused by signal error processing.

[0058] In step S10, refer to Figure 1 In step S10, historical signal data is collected from the FTTR network, where the historical signal data includes historical optical signal data and historical electrical signal data of different modulation modes.

[0059] As an example, the collection and preparation of historical signal data is a basic step in the subsequent deep learning model training. Its purpose is to enable the subsequent deep learning model to be exposed to diverse data, thereby improving the generalization ability and robustness of the deep learning model.

[0060] As an example, historical optical and electrical signal data using different modulation schemes is first collected from the FTTR network as historical signal data. This historical optical and electrical signal data should cover a variety of signal-to-noise ratios (SNRs) and interference types to ensure that the subsequent deep learning model can effectively learn and identify signals in different signal environments. The signal-to-noise ratio (SNR) is an important indicator of signal quality and is defined as the ratio of signal power to noise power. The specific formula can be as follows:

[0061]

[0062] Among them, P signal is the signal power, By collecting historical optical and electrical signal data with different signal-to-noise ratios, various signal conditions in actual FTTR networks can be simulated.

[0063] In step S20, refer to Figure 1 In step S20, the historical signal data is preprocessed.

[0064] As an example, in step S20 , preprocessing the historical signal data may include the following steps: S201 - S202 .

[0065] S201: Marking the historical signal data.

[0066] S202: Perform data enhancement on the annotated historical data signal.

[0067] As an example, in step S201, the collected historical signal data can be annotated to ensure that each historical signal data corresponds to the correct modulation type. This annotation process can be performed by professionals, who can determine the modulation type based on the modulation characteristics of the historical signal data (such as amplitude, frequency, and phase). Common modulation types include amplitude shift keying (ASK), frequency shift keying (FSK), and phase shift keying (PSK). The annotated historical signal data will serve as labels for supervised learning and be used to train subsequent deep learning models.

[0068] As an example, in step S202, in order to improve the robustness of the deep learning model subsequently trained, the annotated historical signal data can be enhanced. Data enhancement techniques can include adding noise, changing signal amplitude and time offset, etc. For example, Gaussian noise can be added to the historical signal data to simulate the noise interference in the actual FTTR network. The specific formula can be as follows:

[0069]

[0070] in, is the historical signal data after data enhancement (adding Gaussian noise), is the original historical signal data, is Gaussian noise with a mean of 0 and a variance of .

[0071] Specifically, by changing the amplitude of the original signal data, we can simulate the attenuation or amplification that the signal may encounter during transmission. Time offset can be achieved by shifting the historical signal data in time to simulate the delay of the signal during transmission.

[0072] In a specific example, suppose we need to collect and prepare historical signal data for training deep learning models. First, we collected 1,000 historical signal data of different modulation modes from the FTTR network, including ASK, FSK, and PSK modulation. The signal-to-noise ratio of each historical signal data ranges from 0 dB to 20 dB, covering low and high signal-to-noise ratio situations. Next, these historical signal data are labeled, and professionals determine the modulation type of each historical signal data based on the modulation characteristics of the signal. Then, we perform data enhancement on the labeled historical signal data, such as adding Gaussian noise to the historical signal data, and the variance of the noise. We set the value of 0.1 to simulate noise interference in real networks. Furthermore, we time-shifted the historical signal data to simulate signal transmission delays. Ultimately, we obtained a dataset containing 1,000 samples (i.e., historical signal data). Each sample was annotated and augmented for subsequent deep learning model training.

[0073] In step S30, refer to Figure 1 In step S30, feature extraction is performed on the preprocessed historical signal data to obtain historical signal features.

[0074] As an example, in step S30 , feature extraction is performed on the pre-processed historical signal data to obtain historical signal features, which may include the following steps: S301 to S305 .

[0075] S301: Convert the pre-processed historical signal data from a time domain signal to a frequency domain signal using Fourier transform.

[0076] S302: Extracting time-frequency features of the frequency domain signal using adaptive wavelet transform.

[0077] S303: Perform unsupervised feature learning using an autoencoder to reconstruct the original signal based on the frequency domain signal.

[0078] S304: Convert the original signal into a time-frequency graph.

[0079] S305: Perform image enhancement processing on the time-frequency graph.

[0080] As an example, feature extraction is a key step in modulation recognition, and its purpose is to extract features from historical signal data that can effectively distinguish different modulation types. Methods such as Fourier transform, adaptive wavelet transform, and autoencoder can be used to extract the time domain and frequency domain features of historical signal data, and convert the signal into an image form, such as a time-frequency diagram, to enhance the identifiability of the features.

[0081] As an example, in step S301, the historical signal data is converted from the time domain to the frequency domain using Fourier transform to extract the frequency domain features of the historical signal data. The Fourier transform formula can be as follows:

[0082]

[0083] Among them, x(t) is the time domain signal, X(f) is the frequency domain signal, e -j2πft It is the kernel function of Fourier transform, which converts the time domain signal into frequency domain representation. Through Fourier transform, we can obtain the spectrum information of historical signal data, and thus analyze the energy distribution of historical signal data at different frequencies.

[0084] As an example, in step S302, in order to more precisely capture the instantaneous changes and local features of the signal, an adaptive wavelet transform can be used. The adaptive wavelet transform can dynamically select the wavelet basis function according to the local characteristics of the signal, thereby extracting more representative time-frequency features. The formula of the adaptive wavelet transform can be as follows:

[0085]

[0086] in, is the wavelet coefficient of the frequency domain signal X(t), is the scale parameter, b is the translation parameter, is the wavelet basis function. By adjusting the scale parameter The translation parameter b can extract the local features of the frequency domain signal at different times and frequencies.

[0087] As an example, in step S303, a self-editor can be used for unsupervised feature learning to automatically extract important features from the frequency domain signal. The self-editor is a neural network whose structure can include an encoder and a decoder. The encoder compresses the input frequency domain signal into a low-dimensional feature representation, and the decoder reconstructs the original signal from these features. The goal of the self-editor is to minimize the reconstruction error L. The formula for the reconstruction error L can be as follows:

[0088]

[0089] in, is the original frequency domain signal, To reconstruct the original signal (which is still a frequency domain signal), by training the self-editor, the important features of the frequency domain signal can be learned and used for subsequent classification tasks.

[0090] As an example, step S304 is a signal visualization step. To enhance feature recognizability, the reconstructed original signal is converted into an image format, such as a time-frequency diagram. A time-frequency diagram can simultaneously display the signal's changes in time and frequency, providing a joint representation of the signal. Common time-frequency diagram generation methods include short-time Fourier transform (STFT) and continuous wavelet transform (CWT); the formula for STFT can be as follows:

[0091]

[0092] in, To reconstruct the original signal, is a sliding window function that limits the time range of analysis. It is the kernel function of Fourier transform, which converts the time domain signal into frequency domain representation. Through STFT, we can get the energy distribution of the signal at different time and frequency.

[0093] As an example, in step S305, image enhancement techniques such as contrast enhancement and edge detection are applied to the generated time-frequency graph to improve the recognizability of the time-frequency graph. The formula for contrast enhancement can be as follows:

[0094]

[0095] Among them, I is the time-frequency image before image enhancement processing, and To enhance the parameters; by adjusting and , which can enhance the contrast of the time-frequency graph and make the features more obvious.

[0096] In one example, feature extraction of preprocessed historical signal data may include the following: First, the preprocessed historical signal data is converted from the time domain to the frequency domain using Fourier transform to obtain the spectral information of the frequency domain signal. Then, the time-frequency features of the frequency domain signal are extracted using adaptive wavelet transform, and the wavelet basis function is dynamically selected to capture the instantaneous changes of the signal. Next, unsupervised feature learning is performed using an autoencoder to compress the frequency domain signal into a low-dimensional feature representation and reconstruct the original signal. Finally, we convert the reconstructed original signal into a time-frequency image and apply contrast enhancement technology to improve the image recognizability. Through these steps, we obtain features that can effectively distinguish different modulation types for subsequent deep learning model training.

[0097] In step S40, refer to Figure 1 In step S40, the deep learning model is trained based on the historical signal features.

[0098] As an example, in step S40, training the deep learning model based on the historical signal features may include the following steps: S401 to S403.

[0099] S401: Constructing a deep learning model, including: selecting a model architecture; setting input layers and output layers to obtain a deep learning model.

[0100] S402: Use the historical signal features to train the deep learning model.

[0101] S403: Evaluate and optimize the trained deep learning model.

[0102] As an example, in step S401, an appropriate deep learning architecture can be selected based on the characteristics of the signal features. For image-based signal features (e.g., time-frequency plots), a convolutional neural network (CNN) can be used for feature extraction and classification. For time series signal features, a long short-term memory network (LSTM) or its variants (e.g., bidirectional LSTM) can be used to capture the signal's temporal dependencies. For signals containing both image-based and time series features, a hybrid model, such as CNN-LSTM, can be used, where CNN is used to extract image features and LSTM is used to process time series information.

[0103] As an example, the historical signal features are used as the input of the deep learning model and input into the input layer of the deep learning model; for CNN, the input is a time-frequency graph, usually a two-dimensional matrix; for LSTM, the input is a time series signal feature, usually a vector sequence.

[0104] As an example, the design outputs the classification result of the modulation type, that is, the output layer outputs the classification result of the modulation type. Specifically, the softmax function can be used for multi-class classification. The formula of the softmax function can be as follows:

[0105]

[0106] in, is the output value of the kth category, is the output value of the jth category, and K is the total number of categories. Through the softmax function, the probability distribution of each category can be obtained, and the category with the highest probability is finally selected as the prediction result, that is, the category with the highest probability is selected as the output result of the output layer.

[0107] As an example, step S402 may include data set division, loss function selection, optimizer selection, and training process monitoring.

[0108] Specifically, regarding data set partitioning, the data set consisting of the historical signal features can be divided into a training set, a validation set, and a test set. Typically, 70% of the data set is used as the training set for training, 15% as the validation set for verification, and 15% as the test set for testing. The validation set is used to monitor the training process of the deep learning model and prevent overfitting.

[0109] Specifically, regarding the selection of loss function, the cross entropy loss function can be selected as the loss function of the deep learning model. The formula can be as follows:

[0110]

[0111] in, is the true label of the i-th sample, The probability that the i-th sample belongs to the k-th category predicted by the deep learning model, N is the total number of samples, and K is the total number of categories. By minimizing the cross entropy loss, the classification performance of the deep learning model can be optimized.

[0112] Specifically, when choosing an optimizer, you can choose the Adam optimizer for training deep learning models. Adam is an adaptive learning rate optimization algorithm that combines the advantages of momentum and RMSProp. It can automatically adjust the learning rate during training to improve training efficiency.

[0113] Specifically, regarding training process monitoring, during training, you can monitor the loss and accuracy of the training set and validation set, use the early stopping strategy, and terminate training early when the validation set loss no longer decreases to prevent overfitting.

[0114] As an example, in step S403, evaluating and optimizing the trained deep learning model may include: optimizing evaluation indicators and hyperparameters.

[0115] Specifically, regarding evaluation indicators, the recognition accuracy, recall rate, and F1-score of the deep learning model can be evaluated on the test set.

[0116] Precision refers to the proportion of samples that are actually positive among those predicted by the deep learning model. The formula can be as follows:

[0117]

[0118] Among them, TP represents true positive examples and FP represents false positive examples.

[0119] Recall refers to the proportion of samples that are actually positive that are predicted as positive by the deep learning model. The formula can be as follows:

[0120]

[0121] Among them, TP represents true positive examples and FN represents false negative examples.

[0122] F1-score is the harmonic mean of precision and recall, and the formula can be as follows:

[0123]

[0124] Among them, Precision is the accuracy and Recall is the recall rate.

[0125] Specifically, regarding hyperparameter optimization, you can adjust the hyperparameters of the deep learning model based on the evaluation results, such as the learning rate, batch size, and number of network layers. You can use grid search or random search methods to find the optimal hyperparameter combination.

[0126] In step S50, refer to Figure 1 In step S50, the trained deep learning model is integrated into the signal processing system of the FTTR network.

[0127] As an example, a specific implementation method of integrating the trained deep learning model into the signal processing system of the FTTR network can be: deploying the trained deep learning model at the front end of the signal processor in the signal processing system of the FTTR network.

[0128] In step S60, refer to Figure 1 In step S60, real-time signal data in the FTTR network is acquired and pre-processed.

[0129] As an example, real-time signals transmitted over optical fibers in an FTTR network can be received in real time. The acquisition rate of the real-time signal needs to match the input requirements of the deep learning model, and signal consistency is ensured by using a fixed sampling rate (e.g., 1 MHz).

[0130] As an example, after acquiring the real-time signal data, the received real-time signal data needs to be preprocessed, which may include normalization, denoising, and feature extraction.

[0131] Specifically, the normalization formula can be as follows:

[0132]

[0133] in, For real-time signal data, is the mean of the real-time signal data, is the standard deviation of the real-time signal data.

[0134] Specifically, a wavelet threshold denoising algorithm may be used to perform denoising on the normalized real-time signal data.

[0135] As an example, the feature extraction method of historical signal data may be used to extract features from real-time signal data, or any existing suitable feature extraction method may be used to extract features from real-time signal data.

[0136] In step S70, refer to Figure 1 In step S70, the preprocessed real-time signal data is input into a deep learning model integrated into the signal processing system for reasoning to obtain a recognition result.

[0137] As an example, in step S70, the preprocessed real-time signal data can be input into a deep learning model integrated into the signal processing system for real-time inference. During the inference process, the deep learning model outputs the modulation type probability distribution of the real-time signal data, and the category with the highest probability can be selected as the final recognition result.

[0138] In step S80, refer to Figure 1 In step S80, a suitable demodulation strategy is selected for demodulation based on the recognition result.

[0139] As an example, a dynamic demodulation strategy or an adaptive demodulation strategy may be selected for demodulation based on the identification result.

[0140] In one example, a dynamic adjustment strategy can be used to demodulate real-time signal data. Specifically, different demodulation strategies can be dynamically selected based on the recognition results of the deep learning model. For example, if the recognition result is QPSK (quadrature phase shift keying), the QPSK demodulation algorithm is used; if the recognition result is 16-QAM (hexadecimal amplitude modulation), the 16-QAM demodulation algorithm is used. The demodulation algorithm can be selected based on the following formula:

[0141]

[0142] By using the above demodulation algorithm selection formula, the most appropriate demodulation algorithm can be selected for demodulation according to the recognition result.

[0143] In another example, an adaptive demodulation strategy can be used to demodulate real-time signal data. Specifically, in situations where the signal-to-noise ratio is low or interference is severe, a more robust demodulation algorithm can be selected for demodulation. For example, in low signal-to-noise ratio environments, a demodulation algorithm based on maximum likelihood estimation (MLE) can be used.

[0144] As an example, demodulation may be performed based on a dynamic demodulation strategy or an adaptive demodulation strategy adopted by the demodulator.

[0145] As an example, after step S80, the following step may be further included: dynamically adjusting demodulation parameters based on the characteristics of the real-time signal data.

[0146] As an example, the demodulation parameters can be dynamically adjusted according to the characteristics of the real-time signal data, such as the signal-to-noise ratio and the interference type; for example, when the signal-to-noise ratio is high, the decision threshold of the demodulator can be increased to improve the demodulation efficiency; when the signal-to-noise ratio is low, the decision threshold of the demodulator can be lowered to improve the robustness of the demodulation.

[0147] As an example, after dynamically adjusting the demodulation parameters based on the characteristics of the real-time signal data, the following steps may also be included: optimizing the deep learning model based on the recognition results and demodulation data; and adjusting the demodulation strategy based on the recognition results obtained by the optimized deep learning model.

[0148] As an example, comparison data of the recognition results of the deep learning model and the actual demodulation type can be collected. The collected comparison data may include comparison data of signal characteristics, recognition results, actual demodulation type, signal-to-noise ratio, interference type, etc.

[0149] As a result, the deep learning model is regularly optimized based on the collected comparison data. The optimization methods may include incremental training, model fine-tuning, and hyperparameter adjustment. The formula of the loss function of incremental training can be as follows:

[0150]

[0151] in, is the loss of the original model, is the loss of new data, is the weight coefficient.

[0152] Alternatively, the demodulation strategy can be updated and adjusted based on the recognition results of the optimized deep learning model. For example, if the recognition accuracy of the deep learning model under a certain interference type is significantly improved, the demodulation algorithm parameters under that interference type can be adjusted to improve the accuracy and efficiency of demodulation.

[0153] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. In addition, Figure 1At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0154] In another embodiment, see Figure 2 The present application also provides an FTTR gateway automatic modulation identification system, which may include: a data collection module 10, a preprocessing module 20, a feature extraction module 30, a model training module 40, a signal processing system 50 and a demodulation module 60; wherein, the data collection module 10 is used to collect historical signal data and real-time signal data in the FTTR network; the preprocessing module 20 is used to preprocess the historical signal data and the real-time signal data; the feature extraction module 30 is used to extract features from the preprocessed historical signal data to obtain historical signal features; the model training module 40 is used to train a deep learning model based on the historical signal features; the signal processing system 50 is used to integrate the trained deep learning model, and perform inference based on the input preprocessed real-time signal data to obtain an identification result; the demodulation module 60 is used to select an appropriate demodulation strategy for demodulation based on the identification result.

[0155] As an example, the FTTR gateway automatic modulation identification system in this embodiment can be used to perform the following Figure 1 And the FTTR gateway automatic modulation identification method described in related embodiments.

[0156] It should be noted that the feature extraction module 30 can also be used to extract features from pre-processed real-time signal data.

[0157] The specific limitations of the FTTR gateway automatic modulation identification system can be found in the limitations of the frequency resource reuse allocation method described above and will not be further elaborated here. Each module in the FTTR gateway automatic modulation identification system can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or stored in a computer device memory in software form, allowing the processor to call and execute the corresponding operations of each module.

[0158] In another embodiment, the present application also provides a computer device, which may be a server, and its internal structure diagram may be as follows: Figure 3As shown. The computer device includes a processor, memory, network interface and database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store data such as historical signal data and real-time signal data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps of a method for automatic modulation identification of an FTTR gateway are implemented.

[0159] In another embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 3 As shown. The computer device includes a processor, memory, network interface, display screen and input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps of a method for automatic modulation identification of an FTTR gateway are implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0160] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0161] In another embodiment, the present application further provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the computer device can achieve the following: Figure 1 And the steps of a FTTR gateway automatic modulation identification method described in related embodiments.

[0162] In another embodiment, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the following Figure 1 And the steps of a FTTR gateway automatic modulation identification method described in related embodiments.

[0163] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0164] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features of the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0165] The above embodiments merely illustrate several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A method for automatic modulation identification of an FTTR gateway, characterized in that: include: Collecting historical signal data from the FTTR network, the historical signal data including historical optical signal data and historical electrical signal data of different modulation modes; Preprocessing the historical signal data; Performing feature extraction on the preprocessed historical signal data to obtain historical signal features, including: using Fourier transform to convert the preprocessed historical signal data from a time domain signal to a frequency domain signal; using adaptive wavelet transform to extract time-frequency features of the frequency domain signal; using an autoencoder to perform unsupervised feature learning to reconstruct the original signal based on the frequency domain signal; converting the original signal into a time-frequency graph; and performing image enhancement processing on the time-frequency graph; Training a deep learning model based on the historical signal features, including: constructing a deep learning model, including: selecting a model architecture; setting an input layer and an output layer to obtain a deep learning model; training the deep learning model using the historical signal features; and evaluating and optimizing the trained deep learning model; Integrate the trained deep learning model into the signal processing system of the FTTR network; Acquiring real-time signal data in the FTTR network and preprocessing the real-time signal data; Inputting the preprocessed real-time signal data into a deep learning model integrated into the signal processing system for reasoning to obtain a recognition result; An appropriate demodulation strategy is selected for demodulation based on the recognition result.

2. The method according to claim 1, characterized in that Preprocessing the historical signal data includes: marking the historical signal data; Perform data enhancement on the annotated historical data signals.

3. The method according to claim 2, characterized in that After the historical signal data are marked, each historical signal data corresponds to a correct modulation type, and the modulation types include: amplitude shift keying, frequency shift keying, and phase shift keying.

4. The method according to claim 2, characterized in that The enhancement techniques for data enhancement of the annotated historical data signal include: adding noise, changing signal amplitude and time offset.

5. The method according to claim 1, characterized in that Selecting an appropriate demodulation strategy for demodulation based on the recognition result includes: A dynamic demodulation strategy or an adaptive demodulation strategy is selected for demodulation based on the identification result.

6. The method according to any one of claims 1 to 5, characterized in that After selecting a suitable demodulation strategy based on the recognition result for demodulation, the method further includes: Demodulation parameters are dynamically adjusted based on characteristics of the real-time signal data.

7. The method according to claim 6, characterized in that The identification result includes a demodulation type; after dynamically adjusting the demodulation parameters based on the characteristics of the real-time signal data, the method further includes: Optimizing the deep learning model based on the recognition results and demodulated data; The demodulation strategy is adjusted based on the recognition results obtained from the optimized deep learning model.

8. An FTTR gateway automatic modulation identification system, characterized in that: The FTTR gateway automatic modulation identification system is used to perform the FTTR gateway automatic modulation identification method according to any one of claims 1 to 7; the FTTR gateway automatic modulation identification system includes: Data collection module, used to collect historical signal data and real-time signal data in the FTTR network; A preprocessing module, configured to preprocess the historical signal data and the real-time signal data; A feature extraction module is used to extract features from the pre-processed historical signal data to obtain historical signal features; A model training module, configured to train a deep learning model based on the historical signal features; The signal processing system is used to integrate the trained deep learning model and perform inference based on the input pre-processed real-time signal data to obtain recognition results; The demodulation module is used to select an appropriate demodulation strategy for demodulation based on the recognition result.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the FTTR gateway automatic modulation identification method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the FTTR gateway automatic modulation identification method according to any one of claims 1 to 7 are implemented.

Images