A method for compensating optical signals and related apparatus

Abstract

This application provides an optical signal compensation method and related equipment, applied in the field of optical communication. The optical compensation method includes: splitting a received coherent optical signal sequence to obtain K delayed optical signal sequences corresponding to the coherent optical signal sequence. Wherein, the delay of each pair of delayed optical signal sequences in the K delayed optical signal sequences is different, and K is a positive integer greater than or equal to 2. The K delayed optical signal sequences are coupled to obtain a preprocessed optical signal sequence. The preprocessed optical signal sequence is input into the delay reservoir layer of a signal compensation neural network, and a first output signal sequence is obtained through the delay reservoir layer. The first output signal sequence is input into the readout layer of the signal compensation neural network, and the compensated optical signal sequence corresponding to the coherent optical signal sequence is determined based on the output of the readout layer.

Description

Technical Field

[0001] This application relates to the field of optical communication, and more particularly to an optical signal compensation method and related equipment. Background Technology

[0002] With the continuous development of communication technology, people's needs and requirements for information transmission are constantly increasing. Optical communication technology, due to its advantages such as extremely high transmission efficiency, convenient hardware implementation, and resistance to external environmental interference, has developed rapidly in the past 50 years, achieving remarkable success and applications. Compared to earlier direct modulation and detection technologies, coherent optical communication transmission technology can utilize the signal to be transmitted to change the frequency, amplitude, and phase characteristics of the optical carrier, thereby improving the transmission efficiency of the optical signal and the utilization rate of the transmitted signal, making long-distance, high-speed optical communication transmission a reality.

[0003] While coherent optical communication transmission technology addresses the high demands for signal transmission rates, it also introduces new technical challenges. For example, after long-distance, high-speed transmission through optical fibers, optical signals suffer severe distortion. Therefore, how to restore the distorted coherent optical signal at the receiving end has become a crucial factor limiting improvements in the transmission rate and distance of coherent optical signals. Among the technologies developed in recent years, the reservoir neural network algorithm has proven effective for coherent optical communication signal compensation.

[0004] The optical compensation algorithm for the reservoir consists of three modules: optical signal preprocessing, a reservoir neural network, and a readout layer. Optical signal preprocessing typically involves adding a mask to the input signal. This involves multiplying the received optical signal (input signal) at a certain moment by a mask matrix to obtain a high-dimensional preprocessed signal. This preprocessed signal is then input into the reservoir neural network for further processing, and finally, the readout layer retrieves the compensated and restored optical signal. Existing compensation systems usually use digital signal processors to add masks to the input signal. Adding masks to the input signal in the digital domain leads to extremely high signal processing complexity. Therefore, how to achieve input signal preprocessing in the entire optical domain and reduce signal preprocessing complexity has become a pressing problem. Summary of the Invention

[0005] This application provides an optical signal compensation method and related equipment. First, multiple delayed optical signal sequences corresponding to a coherent optical signal sequence are determined. Then, these multiple delayed optical signal sequences are coupled into a preprocessed optical signal sequence to complete the preprocessing of the coherent optical signal sequence. Furthermore, the preprocessed optical signal sequence contains more temporal features, which can improve the accuracy of neural network reconstruction of the coherent optical signal sequence.

[0006] The first aspect of this application provides a method for compensating optical signals, the method comprising: Coherent optical signals suffer loss and distortion during long-distance transmission via optical fiber. Therefore, a compensation neural network is needed to compensate for and restore the transmitted optical signal sequence. First, the distorted coherent optical signal sequence after long-distance transmission is received and preprocessed to obtain a preprocessed optical signal sequence. This preprocessed sequence is then input into the delay reserve layer of the signal compensation neural network. Because the delay reserve layer has a cyclic structure, it can extract the temporal and correlation features between the individual optical signals in the coherent optical signal sequence. The first output signal sequence from the delay reserve layer is then transmitted to the readout layer of the compensation neural network. The readout layer processes and analyzes the temporal and correlation features between the individual optical signals to obtain the compensated optical signal sequence corresponding to the original coherent optical signal sequence.

[0007] In preprocessing coherent optical signal sequences, the received coherent optical signal sequence can be split and transmitted through different paths to obtain multiple delayed optical signal sequences corresponding to the coherent optical signal sequence. Each delayed optical signal sequence has a different time delay. Finally, these multiple delayed optical signal sequences are coupled into a single path to obtain the preprocessed optical signal sequence, completing the preprocessing process of the coherent optical signal sequence.

[0008] The aforementioned compensation method utilizes a multi-path delay optical path architecture to split the same optical signal sequence and delay it at different times. Finally, the multi-path delayed optical signal sequences are combined to complete the preprocessing of the coherent optical signal sequence. This introduces the correlation between the optical signals in the coherent optical signal sequence, allowing the correlation and temporal sequence of the optical signals within the sequence to be discovered. This results in the input of the subsequent signal compensation neural network having more temporal and correlational features, which is more conducive to the compensation and restoration of the coherent optical signal sequence, improving the accuracy and efficiency of compensation. Furthermore, compared to traditional preprocessing using masks, this preprocessing process eliminates the need for a digital signal processor to generate the mask, thus eliminating the need for photoelectric conversion and significantly reducing computational complexity. This greatly increases the processing speed of optical signal compensation and significantly reduces the energy consumption of the entire compensation system.

[0009] In one optional implementation, when splitting the distorted coherent optical signal sequence, the coherent optical signal sequence is first divided into K beams, and then injected into K optical paths respectively. Each optical path has a different length, with the length of the i-th optical path being smaller than the length of the (i+1)-th optical path; i is a positive integer greater than or equal to 1 and less than K-1. The K beams of coherent optical signal sequence are then transmitted using the K optical paths. Due to the different lengths of the optical paths, the end of the first optical path will receive the optical signal from the coherent optical signal sequence first. Once the output of the first optical path receives the first optical signal from the coherent optical signal sequence, recording begins at the output of each optical path until the last optical path receives the last optical signal from the coherent optical signal sequence, at which point recording ends. Based on the recording results for each optical path, the i-th delayed optical signal sequence corresponding to the i-th optical path is then determined. By transmitting the same coherent optical signal sequence through optical paths of different lengths, multiple time-delayed optical signal sequences can be obtained. These time-delayed optical signal sequences can reflect the influence of the transmission line on the coherent optical signal sequence, thus revealing the correlation and temporal sequence of the optical signals in the coherent optical signal sequence with the preceding and following optical signals.

[0010] In one optional implementation, before injecting the coherent optical signal sequence into multiple optical paths, a weight value corresponding to each optical path is first determined. Then, based on the weight value corresponding to each optical path, the initial coherent optical signal sequence is divided into K beams and injected into different optical paths. This allows for a more in-depth exploration of the impact of different optical paths on optical signal transmission.

[0011] In an optional implementation, if the initially received distorted coherent optical signal sequence includes L optical signals (i.e., the length of the coherent optical signal sequence is L), then the K delayed optical signal sequences need to be truncated according to this length L. Specifically, the first to Lth optical signals in the recording result corresponding to the i-th optical path are determined as the i-th delayed optical signal sequence corresponding to the i-th optical path, where L is a positive integer greater than 1. This maintains the consistency of the delayed optical signal sequence length and is more conducive to the coupling of all delayed optical signal sequences in the end.

[0012] In one alternative implementation, before splitting the received coherent optical signal sequence, the coherent optical signal sequence is first amplified using an optical amplifier, and then the amplified coherent optical signal sequence is split.

[0013] In one optional implementation, the compensation neural network needs to be trained before compensating and restoring the coherent optical signal sequence using the compensation neural network. The training process can be as follows: First, a set of sample sequences is obtained, where each sample in the set includes an initial optical signal sequence and a distorted optical signal sequence within that initial sequence. Then, the distorted optical signal sequence is input into the delay buffer layer of the compensation neural network to obtain a first output result. This first output result is then input into the readout layer of the compensation neural network to obtain a second output result. Finally, the second output result is compared with the initial optical signal sequence, and the parameters of the readout layer are adjusted based on the comparison result. Finally, the training of the compensation neural network ends when certain preset training conditions are met.

[0014] In an optional implementation, before inputting the distorted optical signal sequence into the delay reservoir layer of the compensation neural network during training, the distorted optical signal sequence can be preprocessed. The preprocessing process involves first injecting the distorted sample sequence into K optical paths to obtain K delayed distorted optical signal sequences corresponding to the distorted optical signal sequence; then coupling the K delayed distorted optical signal sequences to obtain the preprocessed distorted optical signal sequence; finally, the preprocessed distorted optical signal sequence is input into the delay reservoir layer of the compensation neural network.

[0015] In one optional implementation, the delay reservoir layer of the compensation neural network is a recurrent network structure, comprising multiple neurons, and the connection between the multiple neurons is optical coupling.

[0016] A second aspect of this application provides an optical signal compensation device, which includes: The processing unit is used to decouple the received coherent optical signal sequence.

[0017] The acquisition unit is used to acquire K delayed optical signal sequences corresponding to the coherent optical signal sequence. The time delay of each pair of delayed optical signal sequences in the K delayed optical signal sequences is different, and K is a positive integer greater than or equal to 2.

[0018] The processing unit is also used to couple K delayed optical signal sequences to obtain a preprocessed optical signal sequence.

[0019] The computing unit is used to input the preprocessed optical signal sequence into the time delay reservoir layer of the signal compensation neural network, and obtain the first output signal sequence through the time delay reservoir layer.

[0020] The computing unit is also used to input the first output signal sequence to the readout layer of the signal compensation neural network, and to determine the compensation optical signal sequence corresponding to the coherent optical signal sequence based on the output result of the readout layer.

[0021] In one optional implementation, the processing unit is specifically used to inject the coherent optical signal sequence into K optical paths. The length of the i-th optical path is less than the length of the (i+1)-th optical path, and i is a positive integer greater than or equal to 1 and less than K-1. The coherent optical signal sequence is transmitted through the K optical paths.

[0022] The acquisition unit is specifically used to record the optical signals received at the output of the i-th optical path when the output of the first optical path receives the first optical signal in the coherent optical signal sequence. Based on the recording results corresponding to the i-th optical path, the i-th delayed optical signal sequence corresponding to the i-th optical path is determined.

[0023] In one optional implementation, the processing unit is specifically used to determine the weight value corresponding to each of the K optical paths, and to inject the coherent optical signal sequence into the K optical paths according to the weight value corresponding to each optical path.

[0024] In one optional implementation, the coherent optical signal sequence includes L optical signals. The acquisition unit is specifically configured to determine the i-th delayed optical signal sequence corresponding to the i-th optical path based on the first to L-th optical signals in the recording result corresponding to the i-th optical path. L is a positive integer greater than 1.

[0025] In one optional implementation, the processing unit is specifically configured to input the received coherent optical signal sequence into an optical amplifier, amplify the coherent optical signal sequence through the optical amplifier, and then split the amplified coherent optical signal sequence.

[0026] In an optional embodiment, the compensation device further includes a training unit. The training unit is used to train the signal compensation neural network. The training process of the signal compensation neural network is as follows: A set of sample sequences is acquired, including an initial sample sequence and a corresponding distorted sample sequence. The distorted sample sequence is input into the delay buffer layer of the compensation neural network to obtain a first output result. The first output result is input into the readout layer of the compensation neural network to obtain a second output result. The second output result is compared with the initial sample sequence, and the parameters of the readout layer are adjusted according to the comparison result. When a preset training condition is met, the training of the compensation neural network ends.

[0027] In an optional implementation, the training unit is further configured to inject the distorted sample sequence into K optical paths to obtain K time-delayed distorted sample sequences corresponding to the distorted sample sequence. The K time-delayed distorted sample sequences are coupled to obtain a preprocessed distorted sample sequence. The preprocessed distorted sample sequence is then input into the time-delay reservoir layer of the compensation neural network.

[0028] In one optional implementation, the delay reserve pool layer is a recurrent network structure, which includes multiple neurons, and the connection between the multiple neurons is optical coupling.

[0029] Thirdly, embodiments of this application provide a training device that may include a processor and a memory coupled together. The memory stores program instructions, and when the program instructions stored in the memory are executed by the processor, the optical signal compensation method described in the first aspect is implemented.

[0030] Fourthly, embodiments of this application provide an execution device that may include a processor and a memory coupled together. The memory stores program instructions, and when the program instructions stored in the memory are executed by the processor, the optical signal compensation method described in the first aspect is implemented.

[0031] Fifthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to perform the optical signal compensation method described in the first aspect.

[0032] In a sixth aspect, embodiments of this application provide a circuit system including a processing circuit configured to perform the optical signal compensation method described in the first aspect above.

[0033] In a seventh aspect, embodiments of this application provide a computer program that, when run on a computer, causes the computer to execute the optical signal compensation method described in the first aspect.

[0034] Eighthly, embodiments of this application provide a chip system including a processor for implementing the functions involved in the foregoing aspects, such as transmitting or processing data and / or information involved in the foregoing methods. In one possible design, the chip system further includes a memory for storing program instructions and data necessary for a server or communication device. This chip system may be composed of chips or may include chips and other discrete devices. Attached Figure Description

[0035] Figure 1 A schematic diagram of the algorithm steps of a reserve pool calculation method provided in an embodiment of this application; Figure 2 A system architecture diagram of an optical signal compensation system provided in an embodiment of this application; Figure 3 This application provides a schematic diagram of the structure of an input layer according to an embodiment of the present application. Figure 4 This is a schematic diagram of another input layer structure provided in an embodiment of this application; Figure 5A This is a schematic diagram of the structure of an optical signal compensation system provided in an embodiment of this application; Figure 5B This is a schematic diagram of another optical signal compensation system provided in an embodiment of this application; Figure 6A A schematic flowchart illustrating a training method for a signal compensation neural network provided in an embodiment of this application; Figure 6B A schematic diagram of multiple delayed optical signal sequences provided in the embodiments of this application; Figure 7 A schematic flowchart illustrating an optical signal compensation method provided in an embodiment of this application; Figure 8 A schematic diagram of a structure of an optical signal compensation device provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of an execution device provided in an embodiment of this application; Figure 10 This is a schematic diagram of the structure of a training device provided in an embodiment of this application. Detailed Implementation

[0036] This application provides an optical signal compensation method and related equipment. First, multiple delayed optical signal sequences corresponding to a coherent optical signal sequence are determined. Then, these multiple delayed optical signal sequences are coupled into a preprocessed optical signal sequence to complete the preprocessing of the coherent optical signal sequence. Furthermore, the preprocessed optical signal sequence contains more temporal features, which can improve the accuracy of neural network reconstruction of the coherent optical signal sequence.

[0037] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0038] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0039] First, the overall workflow of the artificial intelligence system is described; please refer to [link / reference]. Figure 1 , Figure 1 The diagram illustrates a structural framework for artificial intelligence (AI). The framework is further elaborated below along two dimensions: the "Intelligent Information Chain" (horizontal axis) and the "IT Value Chain" (vertical axis). The "Intelligent Information Chain" reflects a series of processes from data acquisition to processing. For example, it could be the general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, and intelligent execution and output. In this process, data undergoes a condensation process of "data—information—knowledge—wisdom." The "IT Value Chain" reflects the value that AI brings to the information technology industry, from the underlying infrastructure of human intelligence and information (provided and processed by technology) to the industrial ecosystem of the system.

[0040] (1) Infrastructure

[0041] The infrastructure provides computing power to support artificial intelligence systems, enabling communication with the external world and providing support through a basic platform. Communication with the outside world is achieved through sensors; computing power is provided by intelligent chips, which can specifically employ hardware acceleration chips such as central processing units (CPUs), neural-network processing units (NPUs), graphics processing units (GPUs), application-specific integrated circuits (ASICs), or field-programmable gate arrays (FPGAs). The basic platform includes distributed computing frameworks and related platform guarantees and support, which may include cloud storage and computing, interconnected networks, etc. For example, sensors communicate with the outside world to acquire data, and this data is provided to intelligent chips in the distributed computing system provided by the basic platform for computation.

[0042] (2) Data

[0043] The data at the next layer of infrastructure is used to represent the data sources in the field of artificial intelligence. The data involves graphics, images, voice, text, and IoT data from traditional devices, including business data from existing systems and sensor data such as force, displacement, liquid level, temperature, and humidity.

[0044] (3) Data processing

[0045] Data processing typically includes methods such as data training, machine learning, deep learning, search, reasoning, and decision-making.

[0046] Among them, machine learning and deep learning can perform intelligent information modeling, extraction, preprocessing, and training of data by symbolizing and formalizing it.

[0047] Reasoning refers to the process in which, in a computer or intelligent system, the machine thinks and solves problems by simulating human intelligent reasoning, based on reasoning control strategies and using formalized information. Typical functions include search and matching.

[0048] Decision-making refers to the process of making decisions based on intelligent information after reasoning, and it typically provides functions such as classification, sorting, and prediction.

[0049] (4) General ability

[0050] After the data processing mentioned above, the results of the data processing can be used to form some general capabilities, such as algorithms or a general system, for example, translation, text analysis, computer vision processing, speech recognition, image recognition, etc.

[0051] (5) Smart products and industry applications

[0052] Intelligent products and industry applications refer to products and applications of artificial intelligence systems in various fields. They encapsulate overall artificial intelligence solutions, productize intelligent information decision-making, and realize practical applications. Their application areas mainly include: intelligent terminals, intelligent manufacturing, intelligent transportation, smart homes, intelligent healthcare, intelligent security, autonomous driving, and smart cities.

[0053] Based on the description of the artificial intelligence system above, the application scenario of this application embodiment is in the field of optical communication. With the continuous development of information technology, optical communication has become the main communication method due to its high transmission efficiency, convenient hardware implementation, and resistance to external environmental interference. Early optical communication mainly adopted Intensity Modulation / Direct Detection (IM / DD) technology, which determines different transmission signals by modulating and detecting the intensity of the optical signal. Specifically, at the signal transmitting end, intensity modulation is used to couple the transmission signal onto the optical carrier, and different signal values ​​are represented by different light wave intensity values. The signal receiving end demodulates the optical signal by directly detecting the intensity of the optical signal. Since IM / DD technology cannot measure the phase, polarization, and other information contained in the optical signal, it cannot use phase and polarization information to determine different signal contents, which leads to low utilization of the optical signal and affects the transmission efficiency of the optical signal. Based on this, coherent optical communication technology was developed. In coherent optical communication systems, the signal transmitter employs a coherent modulation format, which modulates the frequency, amplitude, and phase of the optical carrier using the signal to be transmitted (unlike direct modulation and detection systems, which only change the intensity of the optical carrier), thereby fully utilizing the optical carrier's capabilities. This technology requires the optical carrier to have a defined phase and amplitude, necessitating the use of coherent light as the light source. Coherent optical communication systems overcome the limitations of traditional direct modulation and detection systems on information transmission, allowing the bit rate to increase with changes in the modulation format, making long-distance, high-speed optical communication a reality.

[0054] While coherent optical communication transmission technology addresses the high demands of modern society for signal transmission rates, it also introduces new technical challenges. For example, optical signals modulated using the quadrature amplitude modulation (QAM) format experience severe distortion after long-distance, high-speed transmission over optical fibers. Therefore, effectively recovering the distorted coherent optical signal at the receiving end has become a crucial problem to be solved in coherent optical communication transmission systems.

[0055] Generally, distortion in coherent optical signal transmission is mainly classified into linear distortion and nonlinear distortion. Linear distortion primarily includes dispersion, polarization mode dispersion, and fiber loss. Most linear distortion can be effectively compensated at the signal receiver. Nonlinear loss mainly includes scattering effects and the Kerr effect. The self-phase modulation (SPM), cross-phase modulation (XPM), and four-wave mixing (FWM) effects caused by the Kerr effect have a significant impact on optical signal distortion. Therefore, compensating for nonlinear loss caused by the Kerr effect becomes a key issue in optical signal compensation.

[0056] Currently, the Reservoir Computing Algorithm (RCA) is commonly used to compensate for nonlinear distortion in coherent optical signals. RCA possesses all the advantages of Recurrent Neural Networks (RNNs), significantly reducing computational complexity and improving the speed of distorted signal compensation and restoration. Typically, RCA consists of three layers: an input layer, an intermediate layer, and an output layer. Figure 1 This is a schematic diagram illustrating the algorithm steps of a reserve pool calculation method provided in an embodiment of this application. Figure 1 As shown, the input layer adds a mask to the optical signal. The intermediate layer is the reservoir network, used to extract relevant features of the optical signal. The readout layer analyzes and calculates the relevant features output by the reservoir network to obtain the corresponding compensated and restored signal of the optical signal. The input layer, as the module that directly receives the optical signal to be processed, plays a crucial role in the signal compensation calculation of subsequent modules due to its data processing mode and calculation method. Typically, adding a mask is a preprocessing step for the received optical signal, achieving two objectives. First, it expands the dimension of the input signal; for example, multiplying the input optical signal by a mask weight matrix yields an output sequence with a dimension greater than the input optical signal, thereby enhancing the connectivity between the input optical signal and the various neurons in the reservoir network. Second, different element values ​​in the mask matrix represent the different connection weights of each neuron in the reservoir network to the input optical signal. Since the weight of the same neuron to the input optical signal is fixed, the masking step also introduces the correlation between multiple input optical signals. After the optical signal is preprocessed by adding a mask, more reservoir network states can be activated in the next module reservoir network, thereby improving the reservoir network's compensation performance for signal nonlinear distortion.

[0057] In existing optical signal compensation systems, a digital signal processor (DSP) is typically used to generate a random mask matrix during the preprocessing (masking) of the optical signal to be restored. Specifically, the initial dimensions of the optical signal to be restored are first determined. Then, based on the required weights and the size of the extended dimensions, a weight matrix is ​​randomly generated in the DSP. The optical signal to be restored is then multiplied by the weight matrix to obtain the preprocessed signal with the mask added, completing the preprocessing process. As can be seen from the above, the preprocessing process at the input layer is implemented in the digital domain, requiring the use of a DSP to generate the mask matrix and perform the multiplication of the optical signal to be restored with the mask matrix. Since the subsequent signal restoration and compensation process is performed in the optical domain hardware, this leads to hardware mismatch, necessitating photoelectric conversion. This results in high computational complexity, severely impacts the speed of optical signal compensation and restoration, and causes significant system energy consumption.

[0058] Before providing a detailed description of the optical signal compensation method provided in the embodiments of this application, let's first combine it with... Figure 2 This application describes the optical signal compensation system provided in its embodiments. Please refer to [the relevant documentation / reference]. Figure 2 , Figure 2 This is a system architecture diagram of an optical signal compensation system provided in an embodiment of this application. Figure 2 The optical signal compensation system comprises an input layer 201, a time delay reservoir layer 202, and a readout layer 203. The time delay reservoir layer 202 and the readout layer 203 constitute a signal compensation neural network. The system input signal is the coherent optical signal to be compensated. After preprocessing in the input layer 201, the preprocessing result is input into the time delay reservoir layer 202. After calculation by the time delay reservoir layer 202, the signal enters the output layer. Finally, the readout layer 203 performs decision-making and processing to obtain the final compensated optical signal.

[0059] Figure 3 This is a schematic diagram of the structure of an input layer provided in an embodiment of this application, such as... Figure 3 As shown, the input layer 201 includes an optical amplifier 301, an optical power beam splitter 302, K optical paths (K being a positive integer greater than or equal to 1) 303, and an optical power coupler 304. The input layer 201 is composed of full-domain optical hardware. The input signal to the input layer 201 is the coherent optical signal sequence to be compensated. The length of the sequence is L. After being amplified by optical amplifier 301, the signal is divided into K paths by several optical power beam splitters 302, and injected into K optical paths respectively. The K paths can be denoted as... , , ... For example, the optical path can be an optical fiber path, which is composed of polarization-maintaining fibers of different lengths.

[0060] Each of the K optical paths 303 is composed of polarization-maintaining fibers of different lengths; therefore, the time delay of the coherent optical signal sequence received at the end of each optical path is also different. At the end of each optical path, there is an optical power coupler 304, whose function is to couple the K delayed optical signal sequences together. The coupled optical signal sequence is then a preprocessed optical signal sequence, which will be input into subsequent network layers for further processing and calculation. Figure 4 This is a schematic diagram of a specific input layer hardware architecture provided in an embodiment of this application. It is implemented by a silicon photonics chip, and its operation is the same as described above, so it will not be repeated here. Among them, Power Demuxer is optical power beam splitter 302, and PowerMuxer is optical power coupler 304.

[0061] Based on the above description of input layer 201, Figure 5A This is a schematic diagram of an optical signal compensation system provided in an embodiment of this application. Figure 5A As shown, the input layer 201 is followed by the delay reservoir layer 202, and the delay reservoir layer 202 is followed by the readout layer 203. The delay reservoir layer 202 is a recurrent network structure comprising multiple nonlinear nodes (neurons) 2021. The connections between neurons 2021 can be optical or electrical coupling. The nonlinearity of each neuron 2021 can be implemented using different optical or electrical devices, which is not limited here. Specifically, the coherent optical signal sequence A and coherent optical signal sequence B are first preprocessed by the input layer 201. The preprocessed optical signal sequence obtained after preprocessing... The input will be fed into the corresponding delay reserve pool layer 202, which performs nonlinear processing on it to discover the delay and correlation features of the optical signals in the relevant optical signal sequence, obtaining the first output result. Then, the first input result is fed into the readout layer 203, which calculates the compensated optical signal sequences for the coherent optical signal sequences A and B, thus achieving the compensation and restoration of the distorted optical signal. This is understandable. Figure 5B This is a schematic diagram of another optical signal compensation system provided in an embodiment of this application. Figure 5B The working principle of the optical signal compensation system shown is the same as Figure 5A The optical signal compensation systems shown operate on the same principle. The only difference is the structure of the delay storage pool layer 202. Figure 5A The delay reserve pool layer 202 shown includes one nonlinear node and several virtual nonlinear nodes. Figure 5B The delay reserve pool layer 202 shown includes multiple nonlinear nodes, but no virtual nonlinear nodes. This application does not impose specific limitations on the structure of the delay reserve pool layer 202.

[0062] Based on the above description, the specific implementation process of the training and inference stages of the optical signal compensation method provided in the embodiments of this application will now be described.

[0063] I. Training Phase of Signal Compensation Neural Network

[0064] In this embodiment, the training phase of the signal compensation neural network describes how the training device uses a set of sample sequences for training. For details, please refer to [link to relevant documentation]. Figure 6A , Figure 6A This is a flowchart illustrating a training method for a signal compensation neural network provided in an embodiment of this application. The training method may include: 601. The training equipment acquires a set of sample sequences.

[0065] The training equipment needs to acquire multiple samples to train the compensation neural network. First, the training equipment acquires a set of sample sequences, including the initial sample sequence and the corresponding distorted sample sequences. The initial sample sequence is the initial coherent optical signal sequence transmitted by the optical signal transmitter in the optical communication system, while the distorted sample sequence refers to the distorted coherent optical signal sequence received by the optical signal receiver in the optical communication system after long-distance fiber optic transmission. Our goal is to restore and compensate the distorted coherent optical signal sequence back to the initial coherent optical signal sequence, enabling the optical signal receiver to receive accurate information.

[0066] 602. The training device divides the distorted sample sequence input into K paths and injects them into the K optical paths respectively.

[0067] Assume the distorted sample sequence is a coherent optical signal sequence of length L. Then we need to first... After being amplified by an optical amplifier, it is then divided according to a certain weight value. The light is divided into K beams, each injected into one of K optical paths. Since the lengths of the K optical paths are all different, the light received at the end of each path is... The required time also varies. After transmission through K optical paths, it can achieve [the desired effect]. Different delay effects, for example, the delayed optical signal sequences can be denoted as follows: , , ... .

[0068] Understandable. The optical path is divided into paths based on the weight value corresponding to each path. In practical applications, the weight value corresponding to each optical path can be adjusted according to requirements, and no limitation is made here. The original... Due to the optical amplifier and the weighted splitting, each optical signal sequence is essentially multiplied by a corresponding weight value. Therefore, the optical signal sequence injected into the K optical paths is actually a weighted coherent optical signal sequence.

[0069] 603. The training device acquires K delay distortion sample sequences corresponding to K optical paths.

[0070] K optical paths transmit the weighted K coherent optical signal sequences to obtain a delay distortion sample sequence for each optical path. It's understandable that since the optical paths are all of different lengths, the shortest optical path receives the signal first. For example, in the K optical paths, the length of the i-th optical path is less than the length of the (i+1)-th optical path, where i is a positive integer greater than or equal to 1 and less than K-1. That is, the K optical paths are arranged in order of length. The first optical path is the shortest and receives the first optical signal in the coherent optical signal sequence first. When the first optical path receives the first optical signal, the training device can record it. Assuming t is the delay unit length, it records the sequence every t. For example, when the first optical path receives the first optical signal, after time t, the first optical path will receive the second optical signal. At this time, the second optical path may receive the first optical signal, so the reception status of the second optical path also needs to be recorded.

[0071] To ensure that the length of each delayed optical signal sequence remains L, it is only necessary to record the sequence up to the Lth signal. For example... Figure 6B As shown, A schematic diagram of the corresponding multiple delayed optical signal sequences. , , ... There is a time delay between them.

[0072] 604. The training device couples K delayed distortion sample sequences to obtain preprocessed distortion samples.

[0073] After obtaining K time-delay distortion sample sequences, the training device uses an optical power coupler to couple the K time-delay distortion sample sequences into one path, thus obtaining preprocessed distortion samples. That's understandable. Each signal is no longer its original value, but a weighted sum of the previous and next signals. This makes... The signal sequence introduces continuity between adjacent signals, increases the time delay and correlation characteristics of the optical signal sequence, and is more conducive to the subsequent network to restore the distorted sample sequence.

[0074] 605. The training device inputs the preprocessed distorted samples into the delay reservoir layer of the compensation neural network and obtains the first output result of the delay reservoir layer.

[0075] When the training device inputs preprocessed distorted samples into the time-delay reservoir layer of the compensation neural network, the time-delay reservoir layer can discover the correlation and temporal sequence of optical signals in the coherent optical signal sequence with the preceding and following optical signals. The structure and parameters of the time-delay reservoir layer are fixed; by calculating the preprocessed distorted samples through the time-delay reservoir layer, a first output result containing more correlated features can be obtained.

[0076] 606. The training device inputs the first output result into the readout layer of the compensation neural network and obtains the second output result of the delay reserve pool layer.

[0077] The training device then inputs the first output result into the readout layer of the compensation neural network, where it performs final analysis and data processing, resulting in a final second output result. This second output result is the compensated sample sequence of the distorted sample sequence. Understandably, the closer the second output result is to the initial sample sequence, the stronger the optical signal compensation capability of the compensation neural network and the higher its compensation accuracy.

[0078] 607. The training device compares the second output result with the initial sample sequence and adjusts the parameters of the readout layer based on the comparison result.

[0079] Therefore, the training device needs to compare the final output of the compensation neural network with the initial sample sequence to obtain the loss value, and then pass the loss value in reverse to adjust the parameters of the readout layer according to the loss value in order to improve the performance of the compensation neural network.

[0080] 608. When the preset training conditions are met, the training device ends the training of the compensation neural network.

[0081] Understandably, the training equipment needs to repeatedly execute the above training process until the performance of the compensation neural network reaches the required level and the accuracy of the output results reaches the preset conditions. For example, reaching the required number of training iterations, or the accuracy of the second output result reaching the preset standard. At this point, the training of the compensation neural network can be stopped, and the trained compensation neural network can be obtained.

[0082] II. Reasoning Stage

[0083] In this embodiment, the inference phase describes how the execution device uses the optical signal compensation system to compensate the coherent optical signal to be compensated. For details, please refer to [link to relevant documentation]. Figure 7 , Figure 7This is a schematic flowchart of an optical signal compensation method provided in an embodiment of this application. The optical signal compensation method provided in an embodiment of this application may include: 701. The device receives a sequence of coherent optical signals.

[0084] After the coherent optical signal sequence is transmitted over a long optical fiber, the execution device receives the coherent optical signal sequence to be compensated. The length of the coherent optical signal sequence is L, meaning it includes L optical signals. The execution device needs to reconstruct the coherent optical signal sequence to obtain its corresponding compensated optical signal sequence. For example, after receiving the coherent optical signal sequence, the execution device can amplify it using an optical amplifier, and then perform subsequent operations on the amplified coherent optical signal sequence.

[0085] 702. The execution device divides the coherent optical signal sequence into K beams and injects them into K optical paths respectively.

[0086] If the coherent optical signal sequence is a coherent optical signal sequence of length L Then we need to first... After being amplified by an optical amplifier, it is then divided according to a certain weight value. The light is divided into K beams, each injected into one of K optical paths. Since the lengths of the K optical paths are all different, the light received at the end of each path is... The required time also varies. After transmission through K optical paths, it can achieve [the desired effect]. Different delay effects, for example, the delayed optical signal sequences can be denoted as follows: , , ... .

[0087] Understandable. The optical path is divided into paths based on the weight value corresponding to each path. In practical applications, the weight value corresponding to each optical path can be adjusted according to requirements, and no limitation is made here. The original... Due to the optical amplifier and the weighted splitting, each optical signal sequence is essentially multiplied by a corresponding weight value. Therefore, the optical signal sequence injected into the K optical paths is actually a weighted coherent optical signal sequence.

[0088] 703. The execution device acquires the K delayed optical signal sequences corresponding to the K optical paths.

[0089] K optical paths transmit the weighted K coherent optical signal sequences, thereby obtaining the delayed optical signal sequence corresponding to each optical path. It's understandable that the optical paths are all of different lengths, so the shortest optical path receives the optical signal first. For example, in the K optical paths, the length of the i-th optical path is less than the length of the (i+1)-th optical path, where i is a positive integer greater than or equal to 1 and less than K-1. That is, the K optical paths are arranged in order of length. Therefore, the first optical path is the shortest and receives the first optical signal in the coherent optical signal sequence first. When the first optical path receives the first optical signal, the training device can record it. Assuming t is the delay unit length, it records once every t. For example, when the first optical path receives the first optical signal, after time t, the first optical path will receive the second optical signal. At this time, the second optical path may receive the first optical signal, so the reception status of the second optical path also needs to be recorded. For example, in order to ensure that the length of each delayed optical signal sequence is still L, it is only necessary to record from the beginning to the Lth signal.

[0090] 704. The execution device couples K delayed optical signal sequences to obtain a preprocessed optical signal sequence.

[0091] After obtaining K delayed optical signal sequences, the execution device uses an optical power coupler to couple the K delayed optical signal sequences into one channel to obtain a preprocessed optical signal sequence. That's understandable. Each signal is no longer its original value, but a weighted sum of the previous and next signals. This makes... The introduction of continuity between adjacent signals in the signal sequence increases the time delay and correlation characteristics of the optical signal sequence, which is more conducive to the subsequent reconstruction of the coherent optical signal sequence by the network.

[0092] 705. The execution device inputs the preprocessed optical signal sequence into the time delay storage pool layer of the signal compensation neural network, and obtains the first output signal sequence through the time delay storage pool layer.

[0093] When the execution device inputs the preprocessed optical signal sequence into the delay reservoir layer of the compensation neural network, the delay reservoir layer can discover the correlation and temporal sequence between the optical signals in the coherent optical signal sequence and the preceding and following optical signals. The structure and parameters of the delay reservoir layer are fixed; by calculating the preprocessed optical signal sequence through the delay reservoir layer, a first output result containing more correlated features can be obtained.

[0094] 706. The execution device inputs the first output signal sequence into the readout layer of the signal compensation neural network, and determines the compensation optical signal sequence corresponding to the coherent optical signal sequence based on the output result of the readout layer.

[0095] The execution device then inputs the first output result into the readout layer of the compensation neural network, where it performs final analysis and data processing to obtain a final second output result. This second output result is the compensated optical signal sequence of the correlated optical signal sequence.

[0096] The aforementioned compensation method utilizes a multi-path delay optical path architecture to split the same optical signal sequence and delay it at different times. Finally, the multi-path delayed optical signal sequences are combined to complete the preprocessing of the coherent optical signal sequence. This introduces the correlation between the optical signals in the coherent optical signal sequence, allowing the correlation and temporal sequence of the optical signals within the sequence to be discovered. This results in the input of the subsequent signal compensation neural network having more temporal and correlational features, which is more conducive to the compensation and restoration of the coherent optical signal sequence, improving the accuracy and efficiency of compensation. Furthermore, compared to traditional preprocessing using masks, this preprocessing process eliminates the need for a digital signal processor to generate the mask, thus eliminating the need for photoelectric conversion and significantly reducing computational complexity. This greatly increases the processing speed of optical signal compensation and significantly reduces the energy consumption of the entire compensation system.

[0097] exist Figures 1 to 7 Based on the corresponding embodiments, in order to better implement the above-described solutions of this application, related equipment for implementing the above solutions is also provided below. See details. Figure 8 , Figure 8 This is a schematic diagram of the structure of an optical signal compensation device provided in an embodiment of this application. Figure 8 As shown, the compensation device includes: The processing unit 801 is used to decouple the received coherent optical signal sequence.

[0098] The acquisition unit 802 is used to acquire K delayed optical signal sequences corresponding to the coherent optical signal sequence. The time delay of each pair of delayed optical signal sequences in the K delayed optical signal sequences is different, and K is a positive integer greater than or equal to 2.

[0099] The processing unit 801 is also used to couple K delayed optical signal sequences to obtain a preprocessed optical signal sequence.

[0100] The computing unit 803 is used to input the preprocessed optical signal sequence into the time delay reservoir layer of the signal compensation neural network, and obtain the first output signal sequence through the time delay reservoir layer.

[0101] The computing unit 803 is also used to input the first output signal sequence to the readout layer of the signal compensation neural network, and to determine the compensation optical signal sequence corresponding to the coherent optical signal sequence based on the output result of the readout layer.

[0102] In an optional implementation, the processing unit 801 is specifically used to inject a coherent optical signal sequence into K optical paths. The length of the i-th optical path is less than the length of the (i+1)-th optical path, and i is a positive integer greater than or equal to 1 and less than K-1. The coherent optical signal sequence is transmitted through the K optical paths.

[0103] The acquisition unit 802 is specifically used to record the optical signals received at the output of the i-th optical path when the output of the first optical path receives the first optical signal in the coherent optical signal sequence. Based on the recording results corresponding to the i-th optical path, the i-th delayed optical signal sequence corresponding to the i-th optical path is determined.

[0104] In an optional implementation, the processing unit 801 is specifically used to determine the weight value corresponding to each of the K optical paths, and to inject the coherent optical signal sequence into the K optical paths according to the weight value corresponding to each optical path.

[0105] In one optional implementation, the coherent optical signal sequence includes L optical signals. The acquisition unit 802 is specifically used to determine the i-th delayed optical signal sequence corresponding to the i-th optical path based on the first to L-th optical signals in the recording result corresponding to the i-th optical path. L is a positive integer greater than 1.

[0106] In one optional implementation, the processing unit 801 is specifically configured to input the received coherent optical signal sequence into an optical amplifier, amplify the coherent optical signal sequence through the optical amplifier, and then split the amplified coherent optical signal sequence.

[0107] In an optional embodiment, the compensation device further includes a training unit 804. The training unit 804 is used to train the signal compensation neural network. The training process of the signal compensation neural network is as follows: A sample sequence set is acquired, including an initial sample sequence and a corresponding distorted sample sequence. The distorted sample sequence is input into the delay reservoir layer of the compensation neural network to obtain a first output result. The first output result is input into the readout layer of the compensation neural network to obtain a second output result. The second output result is compared with the initial sample sequence, and the parameters of the readout layer are adjusted according to the comparison result. When a preset training condition is met, the training of the compensation neural network ends.

[0108] In an optional implementation, the training unit 804 is further configured to inject the distorted sample sequence into K optical paths to obtain K time-delayed distorted sample sequences corresponding to the distorted sample sequence. The K time-delayed distorted sample sequences are coupled to obtain a preprocessed distorted sample sequence. The preprocessed distorted sample sequence is then input into the time-delay reservoir layer of the compensation neural network.

[0109] In one optional implementation, the delay reserve pool layer is a recurrent network structure, which includes multiple neurons, and the connection between the multiple neurons is optical coupling or electrical coupling.

[0110] It should be noted that the information interaction and execution process between the modules / units in the optical signal compensation device are different from those in this application. Figures 1 to 7 The various method embodiments are based on the same concept, and the details can be found in the descriptions of the method embodiments shown above in this application, which will not be repeated here.

[0111] The following describes an execution device provided in an embodiment of this application. Please refer to [link / reference]. Figure 9 , Figure 9 This is a schematic diagram of an execution device provided in an embodiment of this application. The execution device 900 can specifically be a virtual reality (VR) device, a mobile phone, a tablet, a laptop computer, a smart wearable device, a monitoring data processing device, or a radar data processing device, etc., and is not limited thereto. The execution device 900 may be equipped with... Figure 8 The optical signal compensation device described in the corresponding embodiment is used to implement Figure 7 The corresponding embodiment executes the function of the device. Specifically, the execution device 900 includes: a receiver 901, a transmitter 902, a processor 903, and a memory 904 (wherein the execution device 900 may have one or more processors 903). Figure 9 (Taking a processor as an example), processor 903 may include application processor 9031 and communication processor 9032. In some embodiments of this application, receiver 901, transmitter 902, processor 903 and memory 904 may be connected via a bus or other means.

[0112] Memory 904 may include read-only memory and random access memory, and provides instructions and data to processor 903. A portion of memory 904 may also include non-volatile random access memory (NVRAM). Memory 904 stores processor and operation instructions, executable modules, or data structures, or subsets thereof, or extended sets thereof, wherein the operation instructions may include various operation instructions for implementing various operations.

[0113] Processor 903 controls the operation of the execution device. In specific applications, the various components of the execution device are coupled together through a bus system, which may include not only the data bus, but also power buses, control buses, and status signal buses. However, for clarity, all buses are referred to as the bus system in the diagram.

[0114] The methods disclosed in the embodiments of this application can be applied to or implemented by processor 903. Processor 903 can be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above methods can be completed by integrated logic circuits in the hardware or by instructions in software form within processor 903. Processor 903 can be a general-purpose processor, a digital signal processor (DSP), a microprocessor, or a microcontroller, and may further include application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. Processor 903 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 904, and processor 903 reads the information from memory 904 and, in conjunction with its hardware, completes the steps of the above method.

[0115] Receiver 901 can be used to receive input digital or character information, and to generate signal inputs related to the settings and function control of the execution device. Transmitter 902 can be used to output digital or character information through the first interface; transmitter 902 can also be used to send instructions to the disk group through the first interface to modify the data in the disk group; transmitter 902 may also include a display device such as a display screen.

[0116] In this embodiment of the application application, the application processor 9031 in the processor 903 is used to execute... Figure 7 The optical signal compensation method executed by the execution device in the corresponding embodiment. It should be noted that the specific manner in which the application processor 9031 executes each step is different from that in this application. Figure 7 The various method embodiments are based on the same concept, and the technical effects they bring are the same as those in this application. Figure 7 The corresponding method embodiments are the same, and for details, please refer to the description in the method embodiments shown above in this application, which will not be repeated here.

[0117] This application also provides a training device; please refer to [link / reference]. Figure 10 , Figure 10 This is a schematic diagram of a training device provided in an embodiment of this application. The training device 1000 can be equipped with... Figure 8 The optical signal compensation device described in the corresponding embodiment is used to implement the function of the training device in the corresponding embodiment of FIG6. Specifically, the training device 1000 is implemented by one or more servers. The training device 1000 can vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 1022 (e.g., one or more processors) and memory 1032, and one or more storage media 1030 (e.g., one or more mass storage devices) for storing application programs 1042 or data 1044. The memory 1032 and storage media 1030 can be temporary or persistent storage. The program stored in the storage media 1030 may include one or more modules (not shown in the figure), each module may include a series of instruction operations on the training device. Furthermore, the CPU 1022 may be configured to communicate with the storage media 1030 and execute the series of instruction operations in the storage media 1030 on the training device 1000.

[0118] The training device 1000 may also include one or more power supplies 1026, one or more wired or wireless network interfaces 1050, one or more input / output interfaces 1058, and / or one or more operating systems 1041, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, etc.

[0119] In one embodiment of this application, the central processing unit 1022 is used to execute... Figure 8 The steps performed by the optical signal compensation device described in the corresponding embodiment. The specific manner in which the central processing unit 1022 performs the aforementioned steps is the same as described in this application. Figure 6A The various method embodiments are based on the same concept, and the technical effects they bring are the same as those in this application. Figure 6A The corresponding method embodiments are the same, and for details, please refer to the description in the method embodiments shown above in this application, which will not be repeated here.

[0120] This application also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned actions. Figure 7 The steps performed by the execution device in the method described in the illustrated embodiment.

[0121] This application embodiment also provides a computer-readable storage medium storing a program for performing signal processing, which, when run on a computer, causes the computer to perform the aforementioned actions. Figure 7 The steps performed by the execution device in the method described in the illustrated embodiment.

[0122] The optical signal compensation device, execution device, and training device provided in this application embodiment can specifically be a chip. The chip includes a processing unit and a communication unit. The processing unit can be, for example, a processor, and the communication unit can be, for example, an input / output interface, pins, or circuits. The processing unit can execute computer execution instructions stored in the storage unit to cause the chip to perform the aforementioned operations. Figure 7 The illustrated embodiment describes a method for compensating optical signals. Optionally, the storage unit is an in-chip storage unit, such as a register or cache. Alternatively, the storage unit can be an external storage unit located within the wireless access device, such as a read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, such as random access memory (RAM).

[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0124] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0125] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for compensating optical signals, characterized in that, The compensation method includes: The received coherent optical signal sequence is split to obtain K delayed optical signal sequences corresponding to the coherent optical signal sequence; the time delay of each pair of delayed optical signal sequences in the K delayed optical signal sequences is different; K is a positive integer greater than or equal to 2; The K delayed optical signal sequences are coupled together to obtain a preprocessed optical signal sequence; The preprocessed optical signal sequence is input into the time delay reservoir layer of the signal compensation neural network, and the first output signal sequence is obtained through the time delay reservoir layer. The first output signal sequence is input to the readout layer of the signal compensation neural network, and the compensation optical signal sequence corresponding to the coherent optical signal sequence is determined based on the output result of the readout layer.

2. The compensation method according to claim 1, characterized in that, The process of splitting the received coherent optical signal sequence includes: The coherent optical signal sequence is injected into K optical paths; wherein the length of the i-th optical path is less than the length of the (i+1)-th optical path; and i is a positive integer greater than or equal to 1 and less than K-1. The coherent optical signal sequence is transmitted through the K optical paths; The step of obtaining the K delayed optical signal sequences corresponding to the coherent optical signal sequence includes: When the output of the first optical path receives the first optical signal in the coherent optical signal sequence, the optical signals received by the output of the i-th optical path are recorded sequentially. Based on the recording results corresponding to the i-th optical path, the i-th delayed optical signal sequence corresponding to the i-th optical path is determined.

3. The compensation method according to claim 2, characterized in that, The step of injecting the coherent optical signal sequence into K optical paths includes: Determine the weight value corresponding to each of the K optical paths; Based on the weight value corresponding to each optical path, the coherent optical signal sequence is injected into K optical paths.

4. The compensation method according to any one of claims 2 to 3, characterized in that, The coherent optical signal sequence includes L optical signals; determining the i-th delayed optical signal sequence corresponding to the i-th optical path includes: Based on the first to Lth optical signals in the recording results corresponding to the i-th optical path, determine the i-th delayed optical signal sequence corresponding to the i-th optical path; where L is a positive integer greater than 1.

5. The compensation method according to any one of claims 1 to 3, characterized in that, The process of splitting the received coherent optical signal sequence includes: The received coherent optical signal sequence is input into an optical amplifier, and the coherent optical signal sequence is amplified by the optical amplifier; The amplified coherent optical signal sequence is split.

6. The compensation method according to any one of claims 1 to 3, characterized in that, Before inputting the processed optical signal sequence into the delay reservoir layer of the signal compensation neural network, the method further includes: The signal compensation neural network is trained; The training process of the signal compensation neural network is as follows: Obtain a set of sample sequences, the set of sample sequences including an initial sample sequence and a distorted sample sequence corresponding to the initial sample sequence; The distorted sample sequence is input into the time delay reservoir layer of the compensation neural network to obtain the first output result; The first output result is input into the readout layer of the compensation neural network to obtain the second output result; Compare the second output result with the initial sample sequence, and adjust the parameters of the readout layer according to the comparison result; The training of the compensation neural network ends when the preset training conditions are met.

7. The compensation method according to claim 6, characterized in that, The step of inputting the distorted sample sequence into the time delay reservoir layer of the compensation neural network includes: The distorted sample sequence is injected into K optical paths to obtain K time-delayed distorted sample sequences corresponding to the distorted sample sequence; The K delayed distortion sample sequences are coupled to obtain a preprocessed distortion sample sequence; The preprocessed distorted sample sequence is input into the time delay reservoir layer of the compensation neural network.

8. The compensation method according to claim 6, characterized in that, The time delay reservoir layer is a recurrent network structure, and the time delay reservoir layer includes multiple neurons, which are connected by optical coupling.

9. A compensation device for optical signals, characterized in that, The compensation device includes: The processing unit is used to decompose the received coherent optical signal sequence; The acquisition unit is used to acquire K delayed optical signal sequences corresponding to the coherent optical signal sequence; the time delay of each pair of delayed optical signal sequences in the K delayed optical signal sequences is different; and K is a positive integer greater than or equal to 2. The processing unit is also used to couple the K delayed optical signal sequences to obtain a preprocessed optical signal sequence; The computing unit is used to input the preprocessed optical signal sequence into the time delay reservoir layer of the signal compensation neural network, and obtain the first output signal sequence through the time delay reservoir layer; The computing unit is further configured to input the first output signal sequence into the readout layer of the signal compensation neural network, and determine the compensation optical signal sequence corresponding to the coherent optical signal sequence based on the output result of the readout layer.

10. The compensation device according to claim 9, characterized in that, The processing unit is specifically used to inject the coherent optical signal sequence into K optical paths; wherein the length of the i-th optical path is less than the length of the (i+1)-th optical path; i is a positive integer greater than or equal to 1 and less than K-1; and to transmit the coherent optical signal sequence through the K optical paths. The acquisition unit is specifically used to record the optical signals received by the output end of the i-th optical path when the output end of the first optical path receives the first optical signal in the coherent optical signal sequence; and to determine the i-th delayed optical signal sequence corresponding to the i-th optical path based on the recording results corresponding to the i-th optical path.

11. The compensation device according to claim 10, characterized in that, The processing unit is specifically used to determine the weight value corresponding to each of the K optical paths; and to inject the coherent optical signal sequence into the K optical paths according to the weight value corresponding to each optical path.

12. The compensation device according to any one of claims 10 to 11, characterized in that, The coherent optical signal sequence includes L optical signals; the acquisition unit is specifically used to determine the i-th delayed optical signal sequence corresponding to the i-th optical path based on the first to L-th optical signals in the recording result corresponding to the i-th optical path; where L is a positive integer greater than 1.

13. The compensation device according to any one of claims 9 to 11, characterized in that, The processing unit is specifically used to input the received coherent optical signal sequence into an optical amplifier, amplify the coherent optical signal sequence through the optical amplifier, and split the amplified coherent optical signal sequence.

14. The compensation device according to any one of claims 9 to 11, characterized in that, The compensation device also includes a training unit: The training unit is used to train the signal compensation neural network; The training process of the signal compensation neural network is as follows: Obtain a set of sample sequences, the set of sample sequences including an initial sample sequence and a distorted sample sequence corresponding to the initial sample sequence; The distorted sample sequence is input into the time delay reservoir layer of the compensation neural network to obtain the first output result; The first output result is input into the readout layer of the compensation neural network to obtain the second output result; Compare the second output result with the initial sample sequence, and adjust the parameters of the readout layer according to the comparison result; The training of the compensation neural network ends when the preset training conditions are met.

15. The compensation device according to claim 14, characterized in that, The training unit is further configured to inject the distorted sample sequence into K optical paths to obtain K time-delayed distorted sample sequences corresponding to the distorted sample sequence; couple the K time-delayed distorted sample sequences to obtain a preprocessed distorted sample sequence; and input the preprocessed distorted sample sequence into the time-delay reservoir layer of the compensation neural network.

16. The compensation device according to any one of claims 9 to 11, characterized in that, The time delay reservoir layer is a recurrent network structure, and the time delay reservoir layer includes multiple neurons, which are connected by optical coupling.

17. A computer program product, characterized in that, When the computer program product is run on a computer, it causes the computer to perform the compensation method as described in any one of claims 1 to 8.

18. A computer-readable storage medium, characterized in that, Includes a program that, when run on a computer, causes the computer to perform the compensation method as described in any one of claims 1 to 8.

19. An execution device, characterized in that, It includes a processor and a memory, wherein the processor is coupled to the memory. The memory is used to store programs; The processor is configured to execute a program in the memory, causing the execution device to perform the method as described in any one of claims 1 to 5.

20. A training device, characterized in that, It includes a processor and a memory, wherein the processor is coupled to the memory. The memory is used to store programs; The processor is configured to execute a program in the memory, causing the training device to perform the method as described in any one of claims 6 to 8.

Images