Adaptive equalization method for PS-OFDM signal based on transfer learning

Abstract

The application relates to a PS-OFDM signal adaptive equalization method based on transfer learning. The method comprises the following steps: taking a PS-OFDM signal with a first shaping rate generated by a transmitting end as a training sample, training a pre-constructed neural network model by using the training sample, and obtaining a basic equalization model; extracting unique features of PS-OFDM signals with other target shaping rates, fine-tuning the parameters of the basic equalization model by using the unique features, and obtaining target equalization models corresponding to the target shaping rates; a receiving end receives PS-OFDM signals, and performs pretreatment; according to the shaping rate of the signals, a corresponding equalization model is called to correct the nonlinear distortion of the pretreated PS-OFDM signals, and then the original bit sequence is recovered through PS inverse transformation. By adopting the method, compatible equalization of multi-rate PS-OFDM signals can be realized in a hybrid RoF-UOWC system.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to an adaptive equalization method for PS-OFDM signals based on transfer learning. Background Technology

[0002] As the communications field evolves towards ubiquitous connectivity across air, land, and sea, two key technologies are gradually being applied to meet the high-speed transmission demands across various scenarios: First, RoF (Road-of-Flight) technology connects central offices with remote equipment, relying on the low attenuation of optical fibers to support long-distance transmission; second, UOWC (Underwater Optical Wireless Communication) technology, designed for marine environments, utilizes the low attenuation characteristics of blue-green light in water to construct high-speed links. The hybrid RoF-UOWC system, formed by the fusion of these two technologies, has become an important direction for achieving ubiquitous connectivity across all domains. Because the hybrid system spans multiple links—fiber-to-wireless-underwater—with significant differences in attenuation and bandwidth among these links, rate-adaptive modulation signals are required based on channel conditions to ensure transmission efficiency and reliability. Furthermore, multi-rate signal transmission is susceptible to nonlinear distortion introduced by the channel and optoelectronic devices, necessitating a universal equalizer to reduce system complexity.

[0003] Current mainstream solutions are based on neural network (NN) equalizers. Compared to traditional equalizers, these equalizers have self-learning capabilities, enabling them to autonomously estimate channel states and approximate distortion patterns. For example, artificial NNs in fiber optic systems can improve the signal-to-noise ratio, bidirectional long short-term memory (LSTM) recurrent NNs can reduce inter-symbol interference (ISI), and two-dimensional convolutional NNs in RoF systems can achieve high-speed transmission and superior receiver sensitivity. However, traditional NN equalizers are designed for specific rates / types of modulated signals. Once the rate or type of the received signal changes in the system, the model needs to be retrained based on new data. This retraining efficiency is extremely low and compatibility is poor, hindering the practical application of hybrid RoF-UOWC systems. Summary of the Invention

[0004] Therefore, it is necessary to provide a PS-OFDM signal adaptive equalization method based on transfer learning to address the above-mentioned technical problems.

[0005] An adaptive equalization method for PS-OFDM signals based on transfer learning, the method comprising:

[0006] The PS-OFDM signal with a first shaping rate generated at the transmitter is used as a training sample. The pre-built neural network model is trained using the training sample to learn the general distortion characteristics of the training sample when it is transmitted in the hybrid RoF-UOWC system, and a basic equalization model is obtained.

[0007] The unique features of the PS-OFDM signals of other target forming rates are extracted, and the parameters of the basic equalization model are fine-tuned using the unique features to obtain the target equalization model corresponding to each target forming rate; the unique features include the amplitude distribution features corresponding to the target forming rate and the link distortion features in the hybrid RoF-UOWC system that are adapted to the target forming rate.

[0008] The receiving end receives the PS-OFDM signal transmitted through the hybrid RoF-UOWC system link, performs preprocessing, calls the corresponding equalization model according to the signal shaping rate to perform nonlinear distortion correction on the preprocessed PS-OFDM signal, and then recovers the original bit sequence through PS inverse transform.

[0009] The aforementioned adaptive equalization method for PS-OFDM signals based on transfer learning uses the PS-OFDM signal at the first shaping rate in a hybrid RoF-UOWC system as training samples to train a neural network model to learn the common distortion characteristics of the signal during transmission across hybrid fiber, wireless, and underwater links. The constructed basic equalization model can capture common distortion patterns such as optoelectronic device nonlinearity and fiber dispersion, avoiding repeated learning of the same features. By extracting the amplitude distribution characteristics of other target shaping rate signals and the unique distortion characteristics adapted to strong underwater scattering or high-attenuation fiber links, the parameters of the basic model are fine-tuned to obtain the corresponding target equalization model, which can specifically adapt to the unique transmission characteristics of different rate signals in the hybrid system. The receiver calls the corresponding model for equalization according to the signal shaping rate, which can efficiently correct multi-link composite distortion and recover the original bit sequence. This achieves compatible equalization of multi-rate PS-OFDM signals in the hybrid RoF-UOWC system without requiring full model retraining, significantly improving equalization efficiency and system adaptability. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating a PS-OFDM signal adaptive equalization method based on transfer learning in one embodiment.

[0011] Figure 2 This is a schematic diagram of adaptive equalization and verification in a hybrid RoF-UOWC system in one embodiment, wherein, Figure 2 (a) is a flowchart illustrating the transfer learning process. Figure 2 (b) is a schematic diagram of the signal processing flow of the hybrid RoF-UOWC system;

[0012] Figure 3 This is a schematic diagram illustrating the variation of GMI performance of different equalizers with received optical power under different shaping rate signals in one embodiment. Figure 3 (a) is a schematic diagram showing the change of GMI performance with received optical power at a forming rate of PS(R1). Figure 3(b) is a schematic diagram showing the change of GMI performance with received optical power at a forming rate of PS(R2);

[0013] Figure 4 In one embodiment, PS(R1) and PS(R2) are obtained by ANN equalization and ATL equalization according to the present invention, whereby... Figure 4 (a) is the constellation diagram after PS(R1) is balanced by ANN. Figure 4 (b) is the constellation diagram of PS(R1) after ATL equalization according to the present invention. Figure 4 (c) is the constellation diagram after PS(R2) is balanced by an ANN. Figure 4 (d) is the constellation diagram of PS(R2) after ATL equalization according to the present invention. Detailed Implementation

[0014] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0015] In one embodiment, such as Figure 1 As shown, an adaptive equalization method for PS-OFDM signals based on transfer learning is provided, including the following steps:

[0016] Step 102: The PS-OFDM signal with the first shaping rate generated by the transmitter is used as a training sample. The pre-built neural network model is trained using the training sample to learn the general distortion characteristics of the training sample when it is transmitted in the hybrid RoF-UOWC system, and the basic equalization model is obtained.

[0017] The first shaping rate refers to the initial information rate set for the PS-OFDM signal by a distributed matching unit in probabilistic shaping (PS) technology, used to adapt to the channel conditions of typical underwater links in a hybrid RoF-UOWC system. The PS-OFDM signal is a modulation signal combining PS and Orthogonal Frequency Division Multiplexing (OFDM), improving channel capacity through non-uniform symbol distribution to adapt to the bandwidth and attenuation characteristics of different links. Common distortion characteristics refer to the common distortion patterns of fiber optic, wireless, and underwater links when the PS-OFDM signal is transmitted in a hybrid RoF-UOWC system, including nonlinearity of optoelectronic devices such as modulators and detectors, fiber dispersion, and slight interference from wireless links. The basic equalization model is a neural network model trained based on the first shaping rate signal, which can initially correct the common distortions of various links in the hybrid system.

[0018] Step 102 addresses the characteristics of the hybrid RoF-UOWC system, which spans multiple links across fiber optic, wireless, and underwater networks and exhibits significant common distortions. By training a basic equalization model, the common distortion patterns such as nonlinearity and dispersion of each link are learned in advance. This lays the foundation for subsequent adaptation to signals of different rates and reduces redundant computations in model training. It is particularly suitable for adapting to common scenarios in hybrid systems where multiple links exhibit nonlinearity of optoelectronic devices.

[0019] Step 104: Extract the unique features of the PS-OFDM signals of other target forming rates, and use the unique features to fine-tune the parameters of the basic equalization model to obtain the target equalization model corresponding to each target forming rate.

[0020] The target shaping rate is an information rate other than the first shaping rate (e.g., 3.6 bits / symbol), used to adapt to links with better conditions (e.g., fiber optic links) in a hybrid RoF-UOWC system. Specific characteristics include the amplitude distribution characteristics corresponding to the target shaping rate and the link distortion characteristics adapted to the target shaping rate in a hybrid RoF-UOWC system. Specific characteristics are properties unique to the target shaping rate signal, including the amplitude distribution characteristics of the probability distribution differences of symbols corresponding to different rates and the distortion characteristics of the target rate adapted to links with strong scattering and high attenuation.

[0021] Step 104 addresses the core characteristics of the hybrid RoF-UOWC system, namely the large differences in attenuation and bandwidth among links and the need for dynamic rate adjustment. By extracting the specific distortion features of the target rate-adaptive link, the basic model is fine-tuned, significantly reducing the complexity of equalizing multi-rate signals.

[0022] Step 106: The receiving end receives the PS-OFDM signal transmitted through the hybrid RoF-UOWC system link and performs preprocessing. According to the signal shaping rate, the corresponding equalization model is called to perform nonlinear distortion correction on the preprocessed PS-OFDM signal, and then the original bit sequence is recovered by PS inverse transformation.

[0023] Nonlinear distortion correction is the process by which an equalization model corrects the composite distortion of a signal caused by fiber dispersion, radio frequency interference, and underwater laser scattering.

[0024] Step 106 addresses the nonlinear characteristics introduced by photoelectric / electro-optical conversion after multi-link transmission in the hybrid RoF-UOWC system. Accurate recovery of multi-rate signals is achieved through rate-adaptive model equalization. This ensures the stability of low-rate signals in strong scattering scenarios of underwater links and the capacity of high-rate signals in high-bandwidth scenarios of fiber optic links, ultimately supporting the hybrid RoF-UOWC system to achieve reliable communication across air, land, and sea.

[0025] In the aforementioned adaptive equalization method for PS-OFDM signals based on transfer learning, the PS-OFDM signal at the first shaping rate in the hybrid RoF-UOWC system is used as a training sample to train a neural network model to learn the common distortion characteristics of the signal during multi-link transmission in the system, including fiber optic, radio frequency wireless, and laser underwater transmission. The constructed basic equalization model can capture common distortion patterns such as optoelectronic device nonlinearity and fiber dispersion, avoiding repeated learning of the same features. The amplitude distribution characteristics of other target shaping rate signals and the unique distortion characteristics of the adapted links are extracted, and the parameters of the basic model are fine-tuned to obtain the corresponding target equalization model, which can specifically adapt to the unique transmission characteristics of different rate signals in the hybrid system. The receiver calls the corresponding model for equalization according to the signal shaping rate, which can efficiently correct multi-link composite distortion and restore the original bit sequence, thus achieving compatible equalization of multi-rate PS-OFDM signals in the hybrid RoF-UOWC system without requiring full model retraining, significantly improving equalization efficiency and system adaptability.

[0026] In one embodiment, the transmitter generates a PS-OFDM signal with a first shaping rate by: performing distribution matching processing on the original bit sequence to generate a non-uniform amplitude signal, constructing a PS symbol based on the non-uniform amplitude signal and the check bits generated by forward error correction coding, and performing Hermitian symmetric and inverse Fourier transform on the PS symbol to obtain the PS-OFDM signal.

[0027] In this embodiment, a non-uniform amplitude signal is generated by distribution matching to take advantage of the capacity of probability shaping. Combined with forward error correction coding check bits to construct PS symbols, anti-distortion redundancy can be introduced in advance. Then, signal modulation is completed by Hermite symmetry and inverse Fourier transform, which allows the generated signal to be naturally adapted to the transmission characteristics of the multi-link hybrid RoF-UOWC system. This lays the foundation for resisting distortion and ensuring signal integrity in complex links, while ensuring that the signal format meets the processing requirements of the system's optical-electrical-optical conversion.

[0028] In one embodiment, the neural network model includes an input layer, at least two hidden layers, and an output layer. The hidden layers employ nonlinear activation functions to learn the common distortion characteristics introduced by optoelectronic nonlinearity, fiber dispersion, and underwater scattering in the hybrid RoF-UOWC system.

[0029] In this embodiment, the architecture of the input layer, at least two hidden layers, and the output layer, combined with a nonlinear activation function, can accurately capture the complex characteristics of various common distortions in the hybrid RoF-UOWC system, such as nonlinearity of optoelectronic devices, fiber dispersion, and underwater scattering. This avoids the problem of insufficient learning of composite distortions in hybrid links by a single structure, enabling the trained basic equalization model to have a more comprehensive ability to correct common distortions and reducing the difficulty of fine-tuning when adapting to signals of different rates.

[0030] In one embodiment, preprocessing includes sequentially performing photoelectric conversion, analog-to-digital conversion, symbol synchronization, cyclic prefix removal, and Fourier transform on the received PS-OFDM signal to convert the time-domain signal into a frequency-domain signal.

[0031] In this embodiment, the received signal is sequentially subjected to photoelectric conversion for data acquisition, analog-to-digital conversion for digital signal processing, symbol synchronization to cancel link delay interference, cyclic prefix removal to eliminate OFDM inter-symbol interference, and Fourier transform to convert the time-domain signal to the frequency domain for easy equalization processing. This can effectively eliminate format interference and timing deviations introduced by multi-link transmission in the hybrid RoF-UOWC system, process the signal into a standard form that adapts to the equalization model, and ensure that subsequent equalization operations can accurately apply to the link distortion part.

[0032] In one embodiment, the links of the hybrid RoF-UOWC system include a fiber optic link, a wireless link, and an underwater optical wireless link connected in sequence; the fiber optic link is a single-mode fiber optic link, the wireless link is a millimeter-wave transmission link, and the underwater optical wireless link is a blue-green light-based optical transmission link.

[0033] In this embodiment, single-mode fiber ensures long-distance, low-attenuation transmission, millimeter waves meet the high bandwidth requirements of the wireless band, and blue-green light is adapted to the low-attenuation characteristics underwater.

[0034] In one embodiment, parameter fine-tuning includes fixing some network layer parameters of the base equalization model and adjusting only the network layer weights related to the amplitude distribution characteristics and link distortion characteristics of the target forming rate signal.

[0035] In one embodiment, the PS inverse transform includes inverse distribution matching and low-density parity-check decoding to restore the equalized signal to the original bit sequence.

[0036] In one embodiment, the method further includes: comparing the recovered original bit sequence with the original bit sequence sent by the transmitter to calculate the bit error rate; if the bit error rate exceeds a preset threshold, generating a rate adjustment command and feeding it back to the transmitter to trigger the transmitter to adjust the current shaping rate to a lower shaping rate.

[0037] In this embodiment, the recovered original bit sequence is compared with the original sequence at the transmitter to calculate the bit error rate. When the bit error rate exceeds a preset threshold, the transmitter is triggered to reduce the shaping rate. This can dynamically adapt to the real-time state of the hybrid RoF-UOWC system link, such as enhanced underwater link scattering or sudden increase in fiber link loss. By adjusting the rate, the system avoids continuous high bit error rates and ensures stable transmission even when link conditions fluctuate, thereby improving the system's adaptability to complex link environments.

[0038] In one embodiment, the method further includes: calculating generalized mutual information based on the recovered original bit sequence; comparing the generalized mutual information with a performance threshold; if it is lower than the performance threshold, extracting the residual features of the ideal symbol mapped from the original bit sequence and the symbol after equalization; and incrementally fine-tuning the equalization model corresponding to the current signal forming rate based on the ideal symbol and the residual features to obtain the updated equalization model.

[0039] In this embodiment, the generalized mutual information is calculated based on the recovered bit sequence and compared with the performance threshold. When the value is below the threshold, the residual features of the ideal symbol and the equalized symbol are extracted to fine-tune the model. This can specifically optimize the equalization model's ability to handle residual distortions such as fiber dispersion and underwater scattering in the hybrid RoF-UOWC system, and realize dynamic iterative optimization of the model. This allows the equalization effect to continuously improve as the system operates, further ensuring the transmission reliability of multi-rate signals in the hybrid link.

[0040] In one specific embodiment, such as Figure 2 As shown, a schematic diagram of adaptive equalization and verification in a hybrid RoF-UOWC system is provided, wherein, Figure 2 (a) is a flowchart illustrating the transfer learning process. Figure 2 (b) is a schematic diagram of the signal processing flow of the hybrid RoF-UOWC system. Figure 2 In (a), Task 1: Using the PS-OFDM signal (signal 1) at the first forming rate, extract its general distortion features in the optical and electrical domains to train and obtain Model 1, learn the general distortion features of the hybrid system, and obtain the initial output Model 1; Task 2: Based on Model 1 (transferring learned knowledge), extract specific features using PS-OFDM signals (signal 2) at other target forming rates, and fine-tune to obtain output 2 (Model 1). Figure 2In (b), the signal processing flow at the transmitting end, multi-link transmission and receiving end is as follows: At the transmitting end: the transmitting end digital signal processing module (PS-OFDMTXDSP) generates a signal, which is output by an arbitrary waveform generator (AWG) and combined with DC bias (DC-Bias) to drive the Mach-Zehnder modulator (MZM); the external cavity laser (ECL1) provides an optical carrier, which is modulated and split into two paths by a polarization controller (PC) and an optical coupler (OC), one of which enters a standard single-mode fiber (SSMF) analog fiber link. In the transmission link: In the fiber optic link, after transmission via SSMF, the signal is amplified by an erbium-doped fiber amplifier (EDFA), its power is adjusted by a variable optical attenuator (VOA), and then converted into an electrical signal by a photodetector (PD). This signal then enters the wireless free-space link, passing through a transmit antenna (TA) and a receive antenna (RA). Subsequently, the signal undergoes modulation and electro-optical conversion via an envelope detector (ED), an electrical amplifier (EA), and a laser diode (LD) before entering the underwater optical wireless link for transmission. Further processing is performed by an avalanche photodiode (APD) and a neutral density filter (NDF). At the receiving end: Digital signal processing (PS-OFDM / RXDSP) combined with a digital storage oscilloscope (DSO) completes signal acquisition and equalization recovery.

[0041] The technical effects of this invention were verified through experiments. Specifically, after the signal was transmitted via standard single-mode fiber (SSMF), a wireless link, and underwater optical wireless communication (UOWC), the generalized mutual information (GMI) performance and constellation diagram of different equalizers were verified for PS-OFDM signals with shaping rates of PS(R1) and PS(R2) within a received optical power (ROP) range of 0dBm to 4dBm. The schematic diagram of the GMI performance of different equalizers under PS(R1) and PS(R2) signals as a function of received optical power (ROP, ranging from 0dBm to 4dBm) is shown below. Figure 3As shown, the performance of three schemes—traditional equalizer (Con.), artificial neural network equalizer (ANN), and adaptive transfer learning equalizer (ATL)—is compared under PS(R1) and PS(R2) signals. For PS(R1), the GMI of the artificial neural network equalizer is significantly better than that of the traditional equalizer; when the ROP is 2dBm, the GMI is improved from 3.03 bits / symbol to 3.23 bits / symbol; and when using the ATL equalizer proposed in this invention, the performance is even better, with a further improvement of 0.12 bits / symbol in GMI compared to the ANN equalizer when the ROP is 2dBm. For PS(R2) (which represents a lower degree of shaping, and achieving probabilistic shaping requires a higher signal-to-noise ratio (SNR), but in the hybrid RoF-UOWC system, the signal attenuates and is severely distorted after transmission, resulting in an extremely low overall SNR, making it difficult to guarantee the GMI of PS(R2): In the lower ROP range of 0dBm to 2dBm, the ANN equalizer still has a significant gain compared to the traditional equalizer; however, at higher ROPs, due to the difference in distribution matching rate between PS(R2) and PS(R1), the GMI of the ANN equalizer approaches that of the traditional equalizer (after the channel condition is improved, the nonlinear gain brought by ANN weakens); while the ATL equalizer, due to the introduction of the new characteristics of PS(R2), still shows superiority, with a GMI improvement of 0.2 bits / symbol compared to the ANN equalizer at a ROP of 2dBm. Furthermore, the constellation diagrams of PS(R1) and PS(R2) after ANN equalization and the ATL equalization of this invention at a ROP of 2dBm are as follows: Figure 4 As shown in (ad), the results show that the ATL equalizer of the present invention can significantly suppress nonlinear distortion. Compared with the ANN equalizer, the noise between the constellation points of PS(R1) and PS(R2) as well as in the center and edge regions are significantly reduced.

[0042] It should be understood that, although Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed 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 some of the sub-steps or stages of other steps.

[0043] Specific limitations regarding the transfer learning-based PS-OFDM adaptive equalization device can be found in the limitations of the transfer learning-based PS-OFDM adaptive equalization method described above, and will not be repeated here. Each module in the aforementioned transfer learning-based PS-OFDM adaptive equalization device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in hardware or independent of the processor in a computer device, or stored in software in the memory of a computer device, so that the processor can call and execute the corresponding operations of each module.

[0044] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in 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.

[0045] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A PS-OFDM signal adaptive equalization method based on transfer learning, characterized in that, The method includes: The PS-OFDM signal with a first shaping rate generated at the transmitter is used as a training sample. The pre-built neural network model is trained using the training sample to learn the general distortion characteristics of the training sample when it is transmitted in the hybrid RoF-UOWC system, and a basic equalization model is obtained. The unique features of the PS-OFDM signals of other target forming rates are extracted, and the parameters of the basic equalization model are fine-tuned using the unique features to obtain the target equalization model corresponding to each target forming rate; the unique features include the amplitude distribution features corresponding to the target forming rate and the link distortion features in the hybrid RoF-UOWC system that are adapted to the target forming rate. The receiving end receives the PS-OFDM signal transmitted through the hybrid RoF-UOWC system link, performs preprocessing, calls the corresponding equalization model according to the signal shaping rate to perform nonlinear distortion correction on the preprocessed PS-OFDM signal, and then recovers the original bit sequence through PS inverse transform. The neural network model includes an input layer, at least two hidden layers, and an output layer. The hidden layers employ nonlinear activation functions to learn the common distortion characteristics introduced by optoelectronic device nonlinearity, fiber dispersion, and underwater scattering in the hybrid RoF-UOWC system. The links of the hybrid RoF-UOWC system include an optical fiber link, a wireless link, and an underwater optical wireless link connected in sequence; the optical fiber link is a single-mode optical fiber link, the wireless link is a millimeter-wave transmission link, and the underwater optical wireless link is a blue-green light-based optical transmission link. The parameter fine-tuning includes: The network layer parameters of the fixed basic equalization model are adjusted only, and the network layer weights related to the amplitude distribution characteristics and link distortion characteristics of the target forming rate signal are adjusted. The method further includes: The generalized mutual information is calculated based on the recovered original bit sequence. The generalized mutual information is compared with a performance threshold. If it is lower than the performance threshold, the residual features of the ideal symbol mapped from the original bit sequence and the symbol after equalization are extracted. The equalization model corresponding to the current signal forming rate is incrementally fine-tuned based on the ideal symbol and the residual features to obtain the updated equalization model.

2. The method according to claim 1, characterized in that, The transmitter generates a PS-OFDM signal with a first shaping rate, including: The original bit sequence is subjected to distribution matching processing to generate a non-uniform amplitude signal. A PS symbol is constructed based on the non-uniform amplitude signal and the check bits generated by forward error correction coding. The PS symbol is subjected to Hermitian symmetric and inverse Fourier transform to obtain the PS-OFDM signal.

3. The method according to claim 1, characterized in that, The preprocessing includes: The received PS-OFDM signal is sequentially subjected to photoelectric conversion, analog-to-digital conversion, symbol synchronization, cyclic prefix removal, and Fourier transform to convert the time-domain signal into a frequency-domain signal.

4. The method according to claim 1, characterized in that, The PS inverse transform includes inverse distribution matching processing and low-density parity-check decoding, which are used to restore the equalized signal to the original bit sequence.

5. The method according to claim 1, characterized in that, The method further includes: The recovered original bit sequence is compared with the original bit sequence sent by the transmitter, and the bit error rate is calculated. If the bit error rate exceeds a preset threshold, a rate adjustment command is generated and fed back to the transmitter, triggering the transmitter to adjust the current forming rate to a lower forming rate.

Images