An end-to-end channel coding optimization method

Abstract

This invention discloses an end-to-end channel coding optimization method applied to the field of high-speed optical communication technology. It includes the following steps: establishing a direct-modulation / direct-detection (IM / DD) optical fiber channel model to simulate the transmission characteristics of polar-coded signals in single-mode optical fiber; establishing an end-to-end polar-code coding optimization system; integrating the IM / DD optical fiber channel model into the end-to-end polar-code coding optimization system to form a channel-assisted polar-code autoencoder IM / DD optical interconnect end-to-end system; and performing parameter tuning on the system to determine the optimized polar-code coding polarization weight channel index. This invention models the optical fiber channel using an RNN neural network and then utilizes the encoding and decoding structure of an autoencoder to jointly optimize the transmitter, channel, and receiver of the optical fiber system as a whole. Through training, a reliable channel suitable for the optical fiber system is found, thereby obtaining the optimal encoding and decoding system performance.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of high-speed optical communication technology, more particularly to an end-to-end channel coding optimization method. BACKGROUND

[0002] Short-haul optical fiber communication systems play an indispensable role in various practical applications such as local area networks, data centers, and building internal communication networks. The characteristics of short-haul optical fiber communication systems are limited transmission distance and simple modulation scheme. Among the modulation schemes, intensity modulation direct detection (IM / DD) is the most commonly used for short-haul optical fiber communication systems due to its simple structure and high cost performance. However, the inherent chromatic dispersion and nonlinear effects of optical fibers pose challenges to achieving high reliability and high performance transmission. Currently, deep learning technology has been applied to the transmission end, reception end and channel of optical fiber systems to improve system performance by optimizing specific modules. However, the results of this method may be limited by the complexity of the system or device effect, making it difficult to achieve optimality. Therefore, it is necessary to optimize the entire system to achieve optimal end-to-end (E2E) performance.

[0003] Polar codes have received extensive attention in the field of coding theory due to their excellent performance close to the Shannon limit. It has become a promising error correction technology, but its effectiveness depends on the optimization of the coding construction method. By irregular pruning of Polar codes, the decoding complexity and latency are reduced, while structure nesting or cascading improves the decoding performance. In optical fiber communication systems, research on Polar codes has mainly focused on coherent optical communication. This research mainly combines Polar codes with polarization multiplexing or probability shaping to improve the signal-to-noise ratio, or is used for chaos encryption in optical fiber communication systems to improve the security of communication. However, in direct modulation direct detection systems, the improvement of error correction performance is usually achieved by combining selective mapping and flexible Polar codes, and existing research has not included the optimization of coding construction. In the coding process of Polar codes, the input bits of the encoder are divided into two parts, namely non-frozen bits carrying information bits and frozen bits carrying predefined values known to the decoder. Since the input bits are divided into two groups, the problem of how to select the channel index of the non-frozen bits to provide the best error correction performance is referred to as the construction problem of Polar codes. Traditional Polar code construction methods include polarization weight (PW), Bhattacharyya parameter construction, Gaussian approximation (GA) construction, etc. Polarization weight is calculated by calculating the beta expansion of each polarization channel index, and then selecting the K channels with the highest beta expansion value for transmitting information bits. These techniques usually sort the reliability of fixed channels, which may not necessarily have the best error correction performance in the channels of direct modulation direct detection optical fiber communication systems. Therefore, how to provide an end-to-end channel coding optimization method is a problem that those skilled in the art need to solve. SUMMARY

[0004] Therefore, the application provides an end-to-end channel coding optimization method, which models an optical fiber channel by using an RNN neural network, integrates a sending end, a channel and a receiving end of an optical fiber system as a whole by using a coding and decoding structure of an autoencoder, finds a reliable channel index suitable for the optical fiber system by training, and achieves better performance.

[0005] In order to achieve the above object, the application provides the following technical scheme.

[0006] An end-to-end channel coding optimization method comprises the following steps.

[0007] S1, a direct modulation and direct detection optical fiber channel model is established to simulate transmission characteristics of a polar code coded signal in a single mode optical fiber;

[0008] S2, an end-to-end polar code coding optimization system is established;

[0009] S3, the direct modulation and direct detection optical fiber channel model is integrated into the end-to-end polar code coding optimization system to form a polar code autoencoder direct modulation and direct detection optical interconnection end-to-end system based on channel assistance;

[0010] S4, the polar code autoencoder direct modulation and direct detection optical interconnection system based on channel assistance is parameter tuned to determine an optimized polar code coding polarization weight channel index.

[0011] Optionally, S1 specifically comprises:

[0012] S11, an optical fiber system experiment platform is built, which comprises an arbitrary waveform generator AWG, an electrical amplifier EA, a laser Laser, a modulator MZM, a single mode optical fiber SMF, a variable optical attenuator VOA, a photodetector PD and an oscilloscope OSC;

[0013] S12, a polar code coding modulated by PAM4 is transmitted multiple times, and data sent by the arbitrary waveform generator AWG and data collected by the oscilloscope OSC are recorded;

[0014] S13, the received data is processed by using a sliding window, and the recorded data is proportionally divided into 80% training set and 20% test set;

[0015] S14, a direct modulation and direct detection optical fiber channel model based on RNN is established, data received by the oscilloscope OSC in the training set is taken as input of the RNN model, data sent by the arbitrary waveform generator AWG is taken as a label, parameters are debugged and training is performed, and finally the direct modulation and direct detection optical fiber channel model is obtained;

[0016] S15. Validate the direct-modulation direct-detection fiber optic channel model using a test set. Compare the output data of the direct-modulation direct-detection fiber optic channel model with the labels in the test set to evaluate the accuracy of the direct-modulation direct-detection fiber optic channel model.

[0017] Optionally, S2 is as follows:

[0018] S21. In polar code coding, a three-layer neural network is introduced, with a constant t determined by experimental experience as the input and the channel index in polar code coding as the output.

[0019] S22. Based on the channel index output by the neural network, the information sequence u is polar-coded to obtain the encoded sequence x;

[0020] S23. The encoder and decoder of the autoencoder correspond to the transmitting end and the receiving end in the optical fiber system, respectively. The autoencoder consists of two neural networks with the same structure. The encoder contains an input layer, two hidden layers and an output layer. The input layer contains M neurons, the two hidden layers each contain 2M neurons, and the output layer contains n neurons. The decoder has the same structure as the encoder.

[0021] S24. Perform Polar decoding through multiple iterations of the belief propagation algorithm and calculate the system's bit error rate.

[0022] Optionally, the three-layer neural network in S21 is specifically as follows:

[0023] channel index=Softmax(W2′((W1′×t+b1′))+b2′)

[0024] In the formula, W1′ and W2′ are the weights of the fully connected layer of the three-layer neural network, b1′ and b2′ are the biases of the fully connected layer of the three-layer neural network, and Softmax represents the activation function.

[0025] Optionally, the encoder neural network is specifically:

[0026] z=(3″((2″(L(1″×y+1″))+2″))+3″)

[0027] In the formula, W3″, W2″, and W1″ are the weights of the fully connected layer of the encoder neural network, b1″, b2″, and b3″ are the biases of the fully connected layer of the encoder neural network, Sigmoid represents the activation function, y is the input of the encoder neural network, and z is the output of the encoder neural network.

[0028] Optionally, the decoder neural network is specifically:

[0029] b=(3″′((2″′(U(1″′×a+1″′))+2″′))+3″′)

[0030] In the formula, W3″′, W2″′, and W1″′ are the weights of the fully connected layer of the decoder neural network, b1″′, b2″′, and b3″′ are the biases of the fully connected layer of the decoder neural network, Softmax represents the activation function, a is the input of the decoder neural network, and b is the output of the decoder neural network.

[0031] Optionally, after S4, it also includes:

[0032] S5. Experimentally test the optimized polarization weighted channel index, compare its performance with that of the traditional polar code coding construction method, and calculate the improvement in receiver sensitivity of the optimized method under the same bit error rate.

[0033] As can be seen from the above technical solution, compared with the prior art, the present invention provides an end-to-end channel coding optimization method, which has the following beneficial effects:

[0034] 1. The present invention discloses an overall optimization learning strategy for direct-modulation and direct-detection optical fiber systems. By modeling the transmitter, channel, and receiver as an end-to-end system in a unified manner and combining it with an autoencoder algorithm for global training, optimal system performance is achieved.

[0035] 2. This invention considers the effects of dispersion and nonlinearity in the modeling of direct-modulation direct-detection fiber optic systems, optimizes the polarization weight channel index in polarization code encoding, and achieves performance improvement for direct-modulation direct-detection fiber optic systems, significantly improving receiver sensitivity.

[0036] 3. By adjusting different fiber lengths, this invention can obtain polarization weighted channel indexes for different channel conditions, which has higher adaptability and versatility, and effectively reduces costs. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0038] Figure 1 This is a flowchart of the end-to-end channel coding optimization method of the present invention;

[0039] Figure 2 This is a schematic diagram of the end-to-end channel coding optimization method of the present invention;

[0040] Figure 3 This is a schematic diagram of the measured code distance of the experimental results in the embodiments of the present invention;

[0041] Figure 4 This is a schematic diagram illustrating the bit error rate performance of the experimental results in this embodiment of the invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] This invention discloses an end-to-end channel coding optimization method, such as... Figure 1 and Figure 2 As shown, it includes the following steps:

[0044] S1. Establish a direct-modulation and direct-detection optical fiber channel model to simulate the transmission characteristics of polar code encoded signals in single-mode optical fiber;

[0045] S2. Establish an end-to-end polar code encoding optimization system;

[0046] S3. Integrate the direct-modulation and direct-detection optical fiber channel model into the end-to-end polar code coding optimization system to form a channel-assisted polar code autoencoder direct-modulation and direct-detection optical interconnect end-to-end system.

[0047] S4. Parameter tuning is performed on the channel-assisted polar code autoencoder direct-modulation and direct-detection optical interconnect system to determine the optimized polar code coding polarization weight channel index.

[0048] Furthermore, S1 specifically refers to:

[0049] S11. Build an experimental platform for the fiber optic system, including an arbitrary waveform generator (AWG), an electrical amplifier (EA), a laser (Laser), a modulator (MZM), a single-mode fiber (SMF), a variable optical attenuator (VOA), a photodetector (PD), and an oscilloscope (OSC).

[0050] S12. Transmit polar code encoded by PAM4 modulation multiple times, and record the data sent by the arbitrary waveform generator AWG and the data collected by the oscilloscope OSC.

[0051] S13. Process the received data using a sliding window, and divide the recorded data into 80% training set and 20% test set according to the proportion.

[0052] S14. Establish a direct-modulation and direct-detection fiber optic channel model based on RNN. Use the data received by the OSC oscilloscope in the training set as the input of the RNN model, and the data sent by the AWG arbitrary waveform generator as the tag. Debug the parameters and train to finally obtain the direct-modulation and direct-detection fiber optic channel model.

[0053] S15. Validate the direct-modulation direct-detection fiber optic channel model using a test set. Compare the output data of the direct-modulation direct-detection fiber optic channel model with the labels in the test set to evaluate the accuracy of the direct-modulation direct-detection fiber optic channel model.

[0054] Furthermore, S2 specifically refers to:

[0055] S21. In polar code coding, a three-layer neural network is introduced, with a constant t determined by experimental experience as the input and the channel index in polar code coding as the output.

[0056] In this embodiment of the invention, the value of t is selected based on the specific requirements of the system and the nature of the data being processed, in order to effectively optimize the performance of the neural network. To achieve the goal of outputting channel indices, the outputs of the three-layer neural network need to be constrained to ensure that their number is N, and that their size is an integer between 1 and N. Furthermore, these values ​​used as channel indices must not be repeated. Simultaneously, the propagation of gradients within the three-layer neural network must be maintained to ensure that the parameters of the encoder and decoder neural networks can be effectively updated during training.

[0057] S22. Based on the channel index output by the neural network, the information sequence u is polar-coded to obtain the encoded sequence x;

[0058] In this embodiment of the invention, since the transmitted sequence x consists of only 0 and 1 bits, the one-hot encoding method is used to encode the 2 bits of x into 4 bits and send them to the input of the autoencoder.

[0059] S23. The encoder and decoder of the autoencoder correspond to the transmitting end and the receiving end in the optical fiber system, respectively. The autoencoder consists of two neural networks with the same structure. The encoder contains an input layer, two hidden layers and an output layer. The input layer contains M neurons, the two hidden layers each contain 2M neurons, and the output layer contains n neurons. The decoder has the same structure as the encoder.

[0060] S24. Perform Polar decoding through multiple iterations of the belief propagation algorithm and calculate the system's bit error rate.

[0061] Furthermore, the three-layer neural network in S21 is specifically as follows:

[0062] channel index=Softmax(W2′((W1′×t+b1′))+b2′)

[0063] In the formula, W1′ and W2′ are the weights of the fully connected layer of the three-layer neural network, b1′ and b2′ are the biases of the fully connected layer of the three-layer neural network, and Softmax represents the activation function.

[0064] Furthermore, the encoder neural network is specifically as follows:

[0065] z=(3″((2″(L(1″×y+1″))+2″))+3″)

[0066] In the formula, W3″, W2″, and W1″ are the weights of the fully connected layer of the encoder neural network, b1″, b2″, and b3″ are the biases of the fully connected layer of the encoder neural network, Sigmoid represents the activation function, y is the input of the encoder neural network, and z is the output of the encoder neural network.

[0067] Furthermore, the decoder neural network is specifically as follows:

[0068] b=(3″′((2″′(U(1″′×a+1″′))+2″′))+3″′)

[0069] In the formula, W3″′, W2″′, and W1″′ are the weights of the fully connected layer of the decoder neural network, b1″′, b2″′, and b3″′ are the biases of the fully connected layer of the decoder neural network, Softmax represents the activation function, a is the input of the decoder neural network, and b is the output of the decoder neural network.

[0070] In this embodiment of the invention, the loss function is calculated by the input y of the encoder neural network and the output b of the decoder neural network. During training, the cross-entropy loss function is used as the standard for training the neural network. The optimizer uses the Adam algorithm, sets the learning rate to 0.001, and uses stochastic gradient descent to optimize the parameters of the encoder neural network and the decoder neural network.

[0071] Since the input data of the direct-modulation direct-detection fiber optic channel model is determined by a sliding window, the output data of the corresponding encoder needs to be processed through the sliding window before it can be sent into the direct-modulation direct-detection fiber optic channel model.

[0072] Furthermore, after S4, it also includes:

[0073] S5. Experimentally test the optimized polarization weighted channel index, compare its performance with that of the traditional polar code coding construction method, and calculate the improvement in receiver sensitivity of the optimized method under the same bit error rate.

[0074] In this embodiment of the invention, an experimental platform for an optical fiber system was built. Polar-coded signals modulated by PAM4 were transmitted multiple times, collecting 2.4 million data sets, of which 1.92 million were used for training and 480,000 for testing. The data was input into the established direct-modulation direct-detection optical fiber channel model, and after training, the direct-modulation direct-detection optical fiber channel model was obtained. The experiment transmitted a polarization-weighted channel index optimized through end-to-end training, a signal encoded with polar codes and modulated by PAM4, at a transmission rate of 50 GBaud over a distance of 5 kilometers. By calculating the bit error rate performance of this signal and comparing it with the traditional Polar construction method, experimental results were plotted. It can be seen that the proposed end-to-end polar code encoding optimization method has a certain improvement in receiver sensitivity performance.

[0075] Figure 3 The code distance distribution is shown with and without end-to-end optimization. It can be seen that when the code distance is 4 and 6, the code distance with end-to-end optimization exceeds that without. The total code distances with and without end-to-end optimization are calculated to be 1294 and 1259.5, respectively. Therefore, the higher total code distance with end-to-end optimization improves the fault tolerance of the optical fiber communication system and achieves better BER performance. Figure 4 The superior bit error rate performance of optimized polarization weights in Polar code transmission is demonstrated. Experimental results show that, at 1×10⁻⁶, the bit error rate is significantly lower. -3 At the specified bit error rate, the receiver power sensitivity of the back-to-back transmission (BTB) link is improved by 0.9 dB. In a 5 km single-mode fiber transmission link, considering the impact of frequency attenuation caused by dispersion on the signal, the proposed end-to-end coding scheme can additionally improve optical power loss by 0.4 dB compared with the traditional Polar code encoder.

[0076] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0077] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An end-to-end channel coding optimization method, characterized by, Comprise the following steps: S1, a direct modulation direct detection optical channel model is established to simulate the transmission characteristics of polar code coded signals in a single mode optical fiber; S2, an end-to-end polar code coding optimization system is established; S3, the direct modulation direct detection optical channel model is integrated into the end-to-end polar code coding optimization system to form a channel-assisted polar code autoencoder direct modulation direct detection optical interconnection end-to-end system; S4, the parameters of the channel-assisted polar code autoencoder direct modulation direct detection optical interconnection system are optimized to determine the optimized polar code coding polar weight channel index; S1 is specifically: S11, an optical fiber system experimental platform is built, including an arbitrary waveform generator AWG, an electrical amplifier EA, a laser Laser, a modulator MZM, a single mode optical fiber SMF, a variable optical attenuator VOA, a photodetector PD and an oscilloscope OSC; S12, the PAM4 modulated polar code is transmitted multiple times, and the data sent by the arbitrary waveform generator AWG and the data collected by the oscilloscope OSC are recorded; S13, the received data is processed by using a sliding window, and the recorded data is divided into 80% training set and 20% test set according to the proportion; S14, a direct modulation direct detection optical channel model based on RNN is established, the data received by the oscilloscope OSC in the training set is taken as the input of the RNN model, the data sent by the arbitrary waveform generator AWG is taken as the label, the parameters are debugged and trained, and finally the direct modulation direct detection optical channel model is obtained; S15, the test set is used to verify the direct modulation direct detection optical channel model, and the output data of the direct modulation direct detection optical channel model is compared with the label in the test set to evaluate the accuracy of the direct modulation direct detection optical channel model; S2 is specifically: S21, a three-layer neural network is introduced in the polar code coding, a constant t determined by experimental experience is taken as the input, and the channel index in the polar code coding is taken as the output; S22, the information sequence u is polar code coded according to the channel index output by the neural network to obtain the coded sequence x; S23, the encoder and the decoder of the autoencoder correspond to the sending end and the receiving end in the optical fiber system respectively, the autoencoder is composed of two neural networks with the same structure, the encoder includes an input layer, two hidden layers and an output layer, the input layer includes M neurons, the two hidden layers each include 2M neurons, and the output layer includes n neurons, and the decoder has the same structure as the encoder; S24, the Polar decoding is carried out by the belief propagation algorithm multiple times, and the bit error rate of the system is calculated.

2. The method of claim 1, wherein, The three-layer neural network in S21 is specifically: wherein , is a weight of a fully connected layer of a three-layer neural network, , Softmax is a bias of a fully connected layer of a three-layer neural network, denotes an activation function.

3. The method of claim 1, wherein, The encoder neural network is specifically: wherein , , are weights of the fully connected layer of the encoder neural network, , , are biases of the fully connected layer of the encoder neural network, Sigmoid and ReLU denotes an activation function, y is an input of the encoder neural network, z is an output of the encoder neural network.

4. The method of claim 3, wherein, The decoder neural network is specifically: wherein , , are weights of the fully connected layer of the decoder neural network, , , are biases of the fully connected layer of the decoder neural network, Softmax denotes an activation function, a is an input of the decoder neural network, b is an output of the decoder neural network.

5. The method of claim 1, wherein, After S4, it further comprises: S5, the optimized polar weight channel index is tested, the performance of the traditional polar code coding construction method is compared, and the improvement of the optimization method in the receiving sensitivity under the same bit error rate condition is calculated.

Images