A system and method for intelligent identification of optical radio frequency multi-carrier signals
By constructing training and experimental samples, using the neural network model of multi-task learning architecture for training and transfer learning, the problem of complex data processing and insufficient generalization capabilities in optically loaded radio frequency multi-carrier signal recognition is solved, and the characteristics of multi-carrier signal are realized The flexible identification and optimized allocation of spectrum resources have improved the flexibility and efficiency of the communication system.
Patent Information
- Application Number
- CN202411119503.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-15
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-08-15
AI Technical Summary
The prior art has problems in the identification of optical radio frequency multi-carrier signals such as complex data processing, simple identification objects, and weak generalization capabilities, especially in the optimization and flexible utilization of spectrum resources of multi-carrier signals.
By constructing training samples and experimental samples, a neural network model based on a multi-task learning architecture is trained and transferred learning, simplifying input feature data processing, improving adaptive learning and recognition capabilities for different signal features, and realizing spectrum resources of optical radio frequency systems Optimized allocation and flexible utilization.
The designed neural network model can simultaneously identify multiple signal characteristics of multi-carrier signals, have the ability to cope with different communication systems, accelerate the learning of new data by neural network models, and realize the flexible allocation and efficient utilization of spectrum resources of high-speed communication systems.
Smart Images

Figure CN118646483B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of wireless communications, and in particular relates to a system and method for intelligent identification of optical radio frequency multi-carrier signals. Background Art
[0002] In recent years, with the continuous development of communication and digital technology, the number and types of communication equipment have continued to increase, and emerging digital services have continued to emerge. Larger communication bandwidth and higher data rates have become the development trend of future communication systems. Due to the current shortage of spectrum resources in the low-frequency bands of the microwave area, more available frequencies are concentrated in the higher microwave bands. However, the higher the carrier frequency used in wireless communication, the more likely the transmission of wireless signals will be restricted, such as high free space path loss, severe atmospheric attenuation, limited penetration depth, etc., resulting in the wireless transmission of high-frequency signals being greatly restricted in distance and rate.
[0003] As a relatively mature technology, RF over fiber technology can transmit ultra-wideband wireless signals at high speed over long distances and is widely used in optical access networks (OANs). RF over fiber technology combines the advantages of low optical communication loss, strong anti-interference ability, and large transmission capacity with wireless communication. The transmitter modulates the electrical signal onto an optical carrier, stably transmits the modulated optical signal in the optical fiber, and the receiver restores the optical signal to an electrical signal, thus achieving large-scale, high-speed, and wide-range transmission of wireless signals.
[0004] Although the optical radio frequency technology has improved the transmission distance and rate of wireless signals, the development and application of multi-carrier multiplexing technologies such as wavelength division multiplexing (WDM) and frequency division multiplexing (FDM) have brought new challenges to the optimization and allocation of spectrum resources for OAN. In addition, with the continuous growth in the types and number of OAN terminals, the next generation of OAN is bound to develop in the direction of adaptive dynamic adjustment of network characteristics. Therefore, in order to enhance the utilization of spectrum resources and improve the flexibility and reliability of optical networks, it is necessary to carry out research on signal feature recognition. Signal modulation format and signal-to-noise ratio are two of the most important characteristics of the signal. The changing modulation format affects the receiver's flexibility in adopting the most appropriate signal processing algorithm, while the signal-to-noise ratio is the decisive factor for adaptive transmission according to channel conditions. At present, most related studies use deep learning algorithms to assist signal recognition.
[0005] Takahito Tanimura et al. published an article titled "OSNR Monitoring by Deep Neural Networks Trained with Asynchronously Sampled Data" at the OECC conference in 2016. The article proposed a 5-layer deep neural network (DNN) structure. After 400,000 sample training, it realized the recognition of OSNR of 16GBaud dual-polarization quadrature phase shift keying (DP-QPSK) coherent optical communication system. DNN learned feature vectors from data asynchronously sampled by coherent receivers. This method achieved a measurement error of 1.6dB in the OSNR range of 7.5dB to 31dB. However, this method requires a large amount of experimental data as training samples, which is large in size and difficult to obtain. In addition, this method only recognizes single-carrier low-order modulation format signals.
[0006] In 2019, Zhiquan Wan et al. published an article titled "Intelligent Optical Performance Monitor Using Multi-Task Learning based Artificial Neural Network" in the journal Optics Express. The article proposed an artificial neural network (ANN) method based on multi-task learning (MTL) to simultaneously identify OSNR and signal modulation formats. The signal amplitude histograms (Ahs) after channel equalization were used as training samples for MTL-ANN. After a multi-task learning network structure with a shared hidden layer and combined with loss weight allocation, the modulation format and OSNR recognition of three signals, NRZ-OOK, PAM4, and PAM8, in a 20GBaud intensity modulation direct detection system were realized. The modulation format recognition accuracy and OSNR recognition mean square error (MSE) reached 100% and 0.11dB, respectively. This scheme realizes the simultaneous recognition of the modulation format and optical signal-to-noise ratio of a single carrier signal. However, the data processing for obtaining training samples is relatively complicated, and the loss weight allocation requires additional mathematical calculations. In addition, when this solution is applied to different channels, it is necessary to re-collect training samples for training, which is time-consuming. Summary of the invention
[0007] In view of the above, the purpose of the present invention is to provide a system and method for intelligent identification of optical radio frequency multi-carrier signals, construct training samples and experimental samples based on an optical radio frequency simulation system and an optical radio frequency experimental system respectively, and perform training and transfer learning based on a neural network model with a multi-task learning architecture, thereby simplifying the input feature data processing steps, improving the adaptive learning and recognition capabilities of different signal features, and helping to achieve optimal allocation and flexible utilization of spectrum resources by the optical radio frequency system.
[0008] In order to achieve the above-mentioned invention object, the technical solution provided by the present invention is as follows:
[0009] A system for intelligent identification of radio frequency over light multi-carrier signals provided by an embodiment of the present invention comprises: a training sample construction module, a neural network model training module, an experimental sample construction module, a neural network model transfer learning module and a radio frequency over light multi-carrier signal identification module;
[0010] The training sample construction module is used to generate a simulated multi-carrier signal through an optical radio frequency simulation system, and extract the in-phase / orthogonal components of the simulated multi-carrier signal to construct a training sample;
[0011] The neural network model training module is used to train the constructed neural network model for various signal feature extraction tasks based on the training samples to obtain a trained neural network model;
[0012] The experimental sample construction module is used to generate an optical radio frequency multi-carrier signal by building an optical radio frequency experimental system, and extract the in-phase / orthogonal components of the optical radio frequency multi-carrier signal to construct an experimental sample;
[0013] The neural network model transfer learning module is used to perform transfer learning-based fine-tuning on the trained neural network model based on experimental samples to obtain a fine-tuned neural network model;
[0014] The radio frequency over light multi-carrier signal recognition module is used to recognize various signal features of the radio frequency over light multi-carrier signal using a fine-tuned neural network model.
[0015] Preferably, the training sample construction module includes: an orthogonal frequency division multiplexing multi-carrier signal generating unit, an optical radio frequency simulation system transmission unit and an in-phase / orthogonal component generating unit;
[0016] The orthogonal frequency division multiplexing multi-carrier signal generating unit is used to convert a serial bit sequence into a parallel sequence, perform M-order orthogonal amplitude modulation mapping on the parallel sequence to obtain mapping symbols, and then generate a simulated orthogonal frequency division multiplexing multi-carrier signal based on an inverse fast Fourier transform method and insert a cyclic prefix, and convert the parallel sequence with the cyclic prefix into a serial sequence and input it into the optical radio frequency simulation system;
[0017] The transmission unit of the radio frequency over light simulation system is used to perform noise processing on the received serial sequence by simulating the working principle of the radio frequency over light system device and simulate the transmission of the orthogonal frequency division multiplexing multi-carrier signals with different signal-to-noise ratios of the serial sequence;
[0018] The in-phase / quadrature component generating unit is used to convert serial OFDM multi-carrier signals with different signal-to-noise ratios into parallel sequences, remove cyclic prefixes and perform fast Fourier transform in sequence, obtain M-order orthogonal amplitude modulation symbols and use their in-phase / quadrature components as training samples.
[0019] Preferably, the value of M of the M-order orthogonal amplitude modulation symbol obtained in the in-phase / quadrature component generating unit is 2 to the mth power.
[0020] Preferably, the neural network model includes multiple output layers, and signal recognition is performed according to the in-phase / orthogonal components of the input, and each output layer outputs a corresponding signal feature.
[0021] Preferably, the radio frequency over light experiment system comprises: a signal generating unit, a signal transmitting unit, a signal receiving unit and a signal processing unit;
[0022] The signal generating unit is used to generate an orthogonal frequency division multiplexing multi-carrier signal of M-order orthogonal amplitude modulation;
[0023] The signal transmitting unit is used to modulate the orthogonal frequency division multiplexing multi-carrier signal output by the signal generating unit onto an optical carrier to form an optical carrier signal;
[0024] The signal receiving unit is used to receive the optical carrier signal output by the signal transmitting unit and convert it into an analog electrical signal;
[0025] The signal processing unit is used to receive the analog electrical signal output by the signal receiving unit, convert it into a digital signal, and then restore it to obtain a multi-carrier signal, and use the in-phase / orthogonal components of the multi-carrier signal as experimental samples.
[0026] Preferably, the signal generating unit comprises: an orthogonal frequency division multiplexing signal generator and an arbitrary signal generator;
[0027] The OFDM signal generator is used to generate an M-order orthogonal amplitude modulated OFDM multi-carrier signal;
[0028] The arbitrary signal generator is used to convert the orthogonal frequency division multiplexing multi-carrier signal into an analog signal and transmit it to the signal transmitting unit.
[0029] Preferably, the signal transmitting unit comprises: a laser, a polarization controller and an optical modulator;
[0030] The laser is used to generate and output an optical signal;
[0031] The polarization controller is used to change the polarization state of the optical signal output by the laser to maximize the output power;
[0032] The optical modulator is used to modulate the orthogonal frequency division multiplexing multi-carrier signal output by the signal generating unit onto the optical signal with the largest output power to form an optical carrier signal and output it.
[0033] Preferably, the optical modulator is a Mach-Zehnder modulator.
[0034] Preferably, the signal receiving unit comprises: an adjustable optical attenuator and a photodetector;
[0035] The adjustable optical attenuator is used to adjust the optical power of the optical carrier signal output by the signal transmitting unit;
[0036] The photoelectric detector is used to convert the optical carrier signal after the optical power is adjusted and output by the adjustable optical attenuator into an analog electrical signal and input the analog electrical signal into the signal processing unit.
[0037] To achieve the above-mentioned purpose of the invention, an embodiment of the present invention further provides a method for intelligent identification of optical radio frequency multi-carrier signals, which uses the above-mentioned system for intelligent identification of optical radio frequency multi-carrier signals to perform intelligent identification of optical radio frequency multi-carrier signals, and is characterized in that it includes the following steps:
[0038] Generate a simulated multi-carrier signal by using a training sample construction module through an optical radio frequency simulation system, and extract the in-phase / quadrature components of the simulated multi-carrier signal to construct a training sample;
[0039] Using the neural network model training module to train the constructed neural network model for various signal feature extraction tasks based on the training samples, and obtaining the trained neural network model;
[0040] The experimental sample construction module is used to build an optical RF experimental system to generate an optical RF multi-carrier signal, and the in-phase / quadrature components of the optical RF multi-carrier signal are extracted to construct the experimental sample;
[0041] Using the neural network model transfer learning module to fine-tune the trained neural network model based on transfer learning based on experimental samples, and obtain a fine-tuned neural network model;
[0042] The optical radio frequency multi-carrier signal recognition module uses a fine-tuned neural network model to identify various signal characteristics of the optical radio frequency multi-carrier signal.
[0043] Compared with the prior art, the present invention has the following beneficial effects:
[0044] (1) The present invention designs a neural network model that can simultaneously identify multiple signal features of multi-carrier signals based on the in-phase / quadrature components of the demodulated symbols;
[0045] (2) The present invention combines the neural network model with transfer learning, so that the neural network model has the ability to cope with different communication systems and accelerates the learning of new data by the neural network model;
[0046] (3) The present invention applies the neural network model to the optical radio frequency system, which helps to achieve flexible allocation and efficient utilization of spectrum resources in high-speed communication systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0048] Figure 1 It is a structural diagram of a system for intelligent identification of optical radio frequency multi-carrier signals provided by an embodiment of the present invention;
[0049] Figure 2 is a schematic diagram of the structure of a training sample construction module provided by an embodiment of the present invention;
[0050] Figure 3 is a schematic diagram of the structure of a neural network model provided by an embodiment of the present invention;
[0051] Figure 4 It is a structural schematic diagram of the radio frequency over light experimental system provided by an embodiment of the present invention;
[0052] Figure 5 The present invention is a flowchart of a method for intelligently identifying optical radio frequency multi-carrier signals provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0053] To make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation methods described herein are only used to explain the present invention and do not limit the scope of protection of the present invention.
[0054] The inventive concept of the present invention is: in view of the technical problems of existing signal recognition methods, such as complex data processing, simple recognition objects and weak generalization ability, the embodiments of the present invention provide a system and method for intelligent recognition of optical radio frequency multi-carrier signals. By building a neural network model based on demodulated symbol data and combining transfer learning, it can simultaneously recognize multiple signal features of optical radio frequency multi-carrier signals, simplify the data processing process, and improve the system's flexible allocation and efficient utilization of spectrum resources.
[0055] Figure 1 FIG. 1 is a schematic diagram of a system for intelligent identification of optical radio frequency multi-carrier signals provided by an embodiment of the present invention. Figure 1 As shown, the embodiment provides a system 100 for intelligent identification of optical radio frequency multi-carrier signals, including: a training sample construction module 110, a neural network model training module 120, an experimental sample construction module 130, a neural network model transfer learning module 140 and an optical radio frequency multi-carrier signal identification module 150.
[0056] Specifically, the training sample construction module 110 is used to generate a simulated multi-carrier signal through an RF over fiber simulation system, and extract the in-phase / orthogonal components of the simulated multi-carrier signal to construct a training sample. Figure 2 As shown, the training sample construction module 110 includes: an orthogonal frequency division multiplexing multi-carrier signal generating unit 111, an optical radio frequency simulation system transmission unit 112 and an in-phase / quadrature component generating unit 113.
[0057] The orthogonal frequency division multiplexing multi-carrier signal generating unit 111 is used to convert a serial bit sequence into a parallel sequence, perform M-order orthogonal amplitude modulation mapping on the parallel sequence to obtain mapping symbols, and then generate a simulated orthogonal frequency division multiplexing multi-carrier signal based on the inverse fast Fourier transform method and insert a cyclic prefix, convert the parallel sequence with the cyclic prefix into a serial sequence and input it into the optical carrier radio frequency simulation system.
[0058] The RF over fiber simulation system transmission unit 112 is used to perform noise processing on the received serial sequence by simulating the working principle of the RF over fiber system components and simulate the transmission of serial OFDM multi-carrier signals with different signal-to-noise ratios.
[0059] The in-phase / quadrature component generating unit 113 is used to convert the serial OFDM multi-carrier signals with different signal-to-noise ratios into parallel sequences, remove the cyclic prefix and perform fast Fourier transform in sequence, obtain M-order orthogonal amplitude modulation symbols and input their in-phase / quadrature components as training samples into the neural network model. Among them, the M value of the M-order orthogonal amplitude modulation symbol is 2 to the power of m, where m is a positive integer greater than or equal to 2 and less than or equal to 8. The modulation format is represented by integer labels starting from 0 in the order of M from small to large. The modulation format of the generated training samples includes all the values of m mentioned above, and the amount of training samples generated is greater than or equal to 20,000, which further improves the generalization and robustness of the neural network model training.
[0060] Specifically, the neural network model training module 120 is used to train the constructed neural network model for various signal feature extraction tasks based on the training samples to obtain a trained neural network model.
[0061] The neural network model adopts a multi-task learning architecture, such as Figure 3 As shown, it includes an input layer, a shared hidden layer, a specific hidden layer and an output layer, and the in-phase / orthogonal components are respectively regarded as one-dimensional input features, and different features are extracted and identified by the shared hidden layer and the specific hidden layer to obtain a variety of signal features as outputs, and each output layer outputs a signal feature. In the embodiment, the neural network model is used to identify a variety of signal features including modulation formats and signal-to-noise ratios. For this purpose, the number of shared hidden layers of the neural network model is designed to be 8, the number of specific hidden layers for modulation format identification is 3, and the number of specific hidden layers for signal-to-noise ratio identification is 0. The number of neurons in the output layer of the modulation format identification of the network model is the number of modulation format types, and the number of neurons in the output layer of the signal-to-noise ratio identification is 1. The minimum width of the convolution kernel in the neural network model is 3, which is used to fully extract signal features.
[0062] In the neural network model, the rectified linear unit is used as the activation function of the hidden layer, the normalized exponential function is used as the activation function of the modulation format recognition output layer, and the linear function is used as the activation function of the signal-to-noise ratio recognition output layer. At the same time, the cross entropy function is used as the loss function of the modulation format recognition, and the Smooth-L1 function is used as the loss function of the signal-to-noise ratio recognition. The total loss function of the multi-task recognition of the neural network model is the sum of the two.
[0063] When training the neural network model, the training samples are divided into a training set and a validation set at a ratio of 8 to 2 to train the neural network model's ability to recognize modulation formats and signal-to-noise ratios. At the same time, the appropriate learning rate is adjusted to be less than 0.1, and the Adam optimization algorithm is used to update the model's hyperparameters on the training set to reduce the model's loss on the validation set. At the same time, the L2 normalization strategy is used to avoid model overfitting, obtain the optimal model hyperparameters, and improve the ability to recognize multi-carrier signal features.
[0064] Specifically, the experimental sample construction module 130 is used to generate an RF-over-light multi-carrier signal by constructing an RF-over-light experimental system, and extract the in-phase / quadrature components of the RF-over-light multi-carrier signal to construct an experimental sample. Figure 4 As shown, the radio frequency over light experiment system includes: a signal generating unit 131 , a signal transmitting unit 132 , a signal receiving unit 133 and a signal processing unit 134 .
[0065] The signal generating unit 131 is used to generate an M-order orthogonal frequency division multiplexing multi-carrier signal of orthogonal amplitude modulation. The M-order orthogonal frequency division multiplexing multi-carrier signal of orthogonal amplitude modulation is generated by an orthogonal frequency division multiplexing signal generator, and then the orthogonal frequency division multiplexing multi-carrier signal is converted into an analog signal by an arbitrary signal generator and transmitted to the signal transmitting unit 132.
[0066] The signal transmitting unit 132 is used to modulate the OFDM multi-carrier signal output by the signal generating unit 131 onto an optical carrier to form an optical carrier signal. First, the optical signal is generated and output by a laser, and then the polarization state of the optical signal output by the laser is changed by a polarization controller to maximize the output power, and then the OFDM multi-carrier signal output by the signal generating unit 131 is modulated onto an optical signal with the maximum output power by an optical modulator to form an optical carrier signal and output it, wherein the optical modulator adopts a Mach-Zehnder modulator.
[0067] The signal receiving unit 133 is used to receive the optical carrier signal output by the signal transmitting unit 132 and convert it into an analog electrical signal. The optical power of the optical carrier signal output by the signal transmitting unit 132 is first adjusted by the adjustable optical attenuator, and then the optical carrier signal after the optical power adjustment output by the adjustable optical attenuator is converted into an analog electrical signal by the photodetector and input into the signal processing unit 134.
[0068] The signal processing unit 134 is used to receive the analog electrical signal output by the signal receiving unit 133, convert it into a digital signal, and then restore it to obtain a multi-carrier signal, and use the in-phase / quadrature components of the multi-carrier signal as experimental samples.
[0069] Specifically, the neural network model transfer learning module 140 is used to perform transfer learning-based fine-tuning on the trained neural network model based on experimental samples to obtain a fine-tuned neural network model.
[0070] The migration of neural network model parameters is based on the optimal network model parameters by deleting some hidden layers and freezing some parameters to fine-tune the neural network model. It is trained based on experimental samples of optical radio frequency orthogonal frequency division multiplexing signals to identify various signal characteristics of experimental signals.
[0071] Specifically, the RF-over-optical multi-carrier signal identification module 150 is used to identify various signal features of the RF-over-optical multi-carrier signal using a fine-tuned neural network model.
[0072] The fine-tuned neural network model is more suitable for the new task of optical RF multi-carrier signal recognition, and will be used to identify various signal features of newly input optical RF multi-carrier signals.
[0073] In summary, the embodiment of the present invention provides a system for intelligent identification of optical radio frequency multi-carrier signals, designs a neural network model, and can simultaneously identify multiple signal characteristics of multi-carrier signals based on the in-phase / quadrature components of the demodulation symbols; combines the neural network model with transfer learning, so that the neural network model has the ability to cope with different communication systems, and accelerates the learning of new data by the neural network model; applies the neural network model to the optical radio frequency system, which helps to achieve flexible allocation and efficient utilization of spectrum resources in high-speed communication systems.
[0074] Based on the same inventive concept, Figure 5 As shown, the embodiment of the present invention also provides a method for intelligent identification of optical radio frequency multi-carrier signals, comprising the following steps:
[0075] S1, using the training sample construction module to generate a simulated multi-carrier signal through an optical radio frequency simulation system, and extracting the in-phase / orthogonal components of the simulated multi-carrier signal to construct a training sample.
[0076] S2, using the neural network model training module to train the constructed neural network model for various signal feature extraction tasks based on the training samples and obtain the trained neural network model.
[0077] S3, using the experimental sample construction module to generate an optical radio frequency multi-carrier signal by building an optical radio frequency experimental system, and extracting the in-phase / orthogonal components of the optical radio frequency multi-carrier signal to construct an experimental sample.
[0078] S4, using the neural network model transfer learning module to perform transfer learning-based fine-tuning on the trained neural network model based on experimental samples and obtain a fine-tuned neural network model.
[0079] S5, using the optical radio frequency multi-carrier signal recognition module to use the fine-tuned neural network model to recognize multiple signal features of the optical radio frequency multi-carrier signal.
[0080] It should be noted that the method for intelligent identification of optical radio frequency multi-carrier signals provided in the above embodiment and the system for intelligent identification of optical radio frequency multi-carrier signals belong to the same inventive concept. The specific implementation process is detailed in the embodiment of the system for intelligent identification of optical radio frequency multi-carrier signals, which will not be repeated here.
[0081] The specific implementation methods described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above is only the most preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A system for intelligent identification of optical radio frequency multi-carrier signals, It is characterized in that include: Training sample construction module, neural network model training module, experimental sample construction module, neural network model transfer learning module and optical radio frequency multi-carrier signal recognition module; The training sample construction module is used to generate a simulated multi-carrier signal through an optical radio frequency simulation system, and extract the in-phase / orthogonal components of the simulated multi-carrier signal to construct a training sample; The neural network model training module is used to train the constructed neural network model for various signal feature extraction tasks based on the training samples to obtain a trained neural network model; The experimental sample construction module is used to generate an optical radio frequency multi-carrier signal by building an optical radio frequency experimental system, and extract the in-phase / orthogonal components of the optical radio frequency multi-carrier signal to construct an experimental sample; The neural network model transfer learning module is used to perform transfer learning-based fine-tuning on the trained neural network model based on experimental samples to obtain a fine-tuned neural network model; The radio frequency over light multi-carrier signal recognition module is used to recognize various signal features of the radio frequency over light multi-carrier signal using a fine-tuned neural network model.
2. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 1, It is characterized in that The training sample construction module includes: an orthogonal frequency division multiplexing multi-carrier signal generation unit, an optical radio frequency simulation system transmission unit and an in-phase / orthogonal component generation unit; The orthogonal frequency division multiplexing multi-carrier signal generating unit is used to convert a serial bit sequence into a parallel sequence, perform M-order orthogonal amplitude modulation mapping on the parallel sequence to obtain mapping symbols, and then generate a simulated orthogonal frequency division multiplexing multi-carrier signal based on an inverse fast Fourier transform method and insert a cyclic prefix, and convert the parallel sequence with the cyclic prefix into a serial sequence and input it into the optical radio frequency simulation system; The transmission unit of the radio frequency over light simulation system is used to perform noise processing on the received serial sequence by simulating the working principle of the radio frequency over light system device and simulate the transmission of the orthogonal frequency division multiplexing multi-carrier signals with different signal-to-noise ratios of the serial sequence; The in-phase / quadrature component generating unit is used to convert serial OFDM multi-carrier signals with different signal-to-noise ratios into parallel sequences, remove cyclic prefixes and perform fast Fourier transform in sequence, obtain M-order orthogonal amplitude modulation symbols and use their in-phase / quadrature components as training samples.
3. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 2, It is characterized in that The value of M of the M-order quadrature amplitude modulation symbol obtained in the in-phase / quadrature component generating unit is 2 to the mth power.
4. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 1, It is characterized in that The neural network model contains multiple output layers, which perform signal recognition according to the in-phase / quadrature components of the input, and each output layer outputs a corresponding signal feature.
5. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 1, It is characterized in that The optical radio frequency experimental system includes: a signal generating unit, a signal transmitting unit, a signal receiving unit and a signal processing unit; The signal generating unit is used to generate an orthogonal frequency division multiplexing multi-carrier signal of M-order orthogonal amplitude modulation; The signal transmitting unit is used to modulate the orthogonal frequency division multiplexing multi-carrier signal output by the signal generating unit onto an optical carrier to form an optical carrier signal; The signal receiving unit is used to receive the optical carrier signal output by the signal transmitting unit and convert it into an analog electrical signal; The signal processing unit is used to receive the analog electrical signal output by the signal receiving unit, convert it into a digital signal, and then restore it to obtain a multi-carrier signal, and use the in-phase / orthogonal components of the multi-carrier signal as experimental samples.
6. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 5, It is characterized in that The signal generating unit comprises: an orthogonal frequency division multiplexing signal generator and an arbitrary signal generator; The OFDM signal generator is used to generate an M-order orthogonal amplitude modulated OFDM multi-carrier signal; The arbitrary signal generator is used to convert the orthogonal frequency division multiplexing multi-carrier signal into an analog signal and transmit it to the signal transmitting unit.
7. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 5, It is characterized in that The signal transmitting unit includes: a laser, a polarization controller and an optical modulator; The laser is used to generate and output an optical signal; The polarization controller is used to change the polarization state of the optical signal output by the laser to maximize the output power; The optical modulator is used to modulate the orthogonal frequency division multiplexing multi-carrier signal output by the signal generating unit onto the optical signal with the largest output power to form an optical carrier signal and output it.
8. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 7, It is characterized in that The optical modulator adopts Mach-Zehnder modulator.
9. The system for intelligent identification of optical radio frequency multi-carrier signals according to claim 5, It is characterized in that The signal receiving unit includes: an adjustable optical attenuator and a photoelectric detector; The adjustable optical attenuator is used to adjust the optical power of the optical carrier signal output by the signal transmitting unit; The photoelectric detector is used to convert the optical carrier signal after the optical power is adjusted and output by the adjustable optical attenuator into an analog electrical signal and input the analog electrical signal into the signal processing unit.
10. A method for intelligent identification of radio frequency over light multi-carrier signals, comprising: using the system for intelligent identification of radio frequency over light multi-carrier signals according to any one of claims 1 to 9 to perform intelligent identification of radio frequency over light multi-carrier signals, It is characterized in that The following steps are involved: Generate a simulated multi-carrier signal by using a training sample construction module through an optical radio frequency simulation system, and extract the in-phase / quadrature components of the simulated multi-carrier signal to construct a training sample; Using the neural network model training module to train the constructed neural network model for various signal feature extraction tasks based on the training samples, and obtaining the trained neural network model; The experimental sample construction module is used to build an optical RF experimental system to generate an optical RF multi-carrier signal, and the in-phase / quadrature components of the optical RF multi-carrier signal are extracted to construct the experimental sample; Using the neural network model transfer learning module to fine-tune the trained neural network model based on transfer learning based on experimental samples, and obtain a fine-tuned neural network model; The optical radio frequency multi-carrier signal recognition module uses a fine-tuned neural network model to identify various signal characteristics of the optical radio frequency multi-carrier signal.
Citation Information
Patent Citations
Power amplifier-oriented transfer learning rapid low-complexity modeling method
CN113221308A
Multi-core optical fiber channel modulation format identification method and device
CN114024810A
Training object detection models using transfer learning
CN115830486A
Nonlinear compensation method based on machine learning
CN115882959A
Signal distortion processing method for multimode photoelectric hybrid communication cable
CN117896004A
