Vortex optical communication demodulation method and system based on weighted extreme random forest model

Abstract

The invention relates to the technical field of communication, in particular to a vortex optical communication demodulation method and system based on a weighted extreme random forest model, and the method comprises the following steps: S10, carrying out the demultiplexing of a received vortex optical multiplexing beam to obtain receiving powers of different modes, and converting the receiving powers into digital signals; s20, inputting the digital signals of the receiving power in different modes into a pre-trained weighted extreme random forest model for demodulation to obtain demodulation signals; a splitting mode of the weighted extreme random forest model is that a feature is extracted from a plurality of features included in a sample with different weights, a sample is randomly extracted from a sample set, and splitting is carried out based on the extracted feature and the sample. According to the invention, crosstalk is used as useful information for signal analysis, operation such as equalization is not needed, the received power information can be directly judged, the complexity is reduced, a communication system is greatly simplified, the development period is shortened, the cost is reduced, and the overall performance of vortex optical multiplexing communication is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and particularly to a demodulation method and system for vortex optical communication based on a weighted extreme random forest model. Background Art

[0002] The orbital angular momentum (OAM) carried by a vortex beam provides a new dimensional resource for the spatial domain of light waves, attracting the attention of more and more researchers. During the transmission of vortex optical multiplexing communication, inter-mode distortion is a common problem, which will affect the stability and quality of communication signals. To handle this distortion, traditional demodulation methods such as multi-user detection technology and MIMO detection algorithms are widely used, which can effectively address this challenge to a certain extent. However, its accuracy is low, it cannot handle non-linear interference, and it requires preprocessing of signals such as channel estimation, carrier recovery, and information equalization, and it cannot meet the growing demands of communication for transmission rate and capacity.

[0003] In recent years, the rapid development of machine learning technology has brought new opportunities to the field of OAM optical communication. With its excellent data analysis and information processing capabilities, compared with traditional digital signal processing algorithms, machine learning algorithms have shown significant advantages in analyzing vortex beam distortion and accurately demodulating signals. Such as deep neural networks (DNN), convolutional neural networks (CNN), etc. These algorithms can directly analyze the distortion of light intensity to classify the OAM modes of the received beam, and then demodulate through the method of mapping encoding and decoding. However, these algorithms require a large amount of data for training, and the training time is long, and it is easy to overfit, and it cannot well cope with the new environment. Algorithms such as K-nearest neighbor (KNN), support vector machine (SVM), and random forest (RF) relatively have lower complexity and can better adapt to the new environment with faster changes. However, complexity is still a drawback for its use in demodulation. That is to say, there is a current situation in the current research that both the demodulation accuracy and complexity need to be improved. Summary of the Invention

[0004] The purpose of the present invention is to provide a demodulation method and system for vortex optical communication based on a weighted extreme random forest model to reduce the demodulation complexity while ensuring the demodulation accuracy.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides a demodulation method for vortex optical communication based on a weighted extreme random forest model, including the following steps:

[0007] S10, demultiplex the received vortex optical multiplexed beam to obtain the received powers of different modes, and convert them into digital signals;

[0008] S20. Input the digital signals of received powers in different modes into a pre-trained weighted extreme random forest model for demodulation to obtain demodulated signals. The splitting method of the weighted extreme random forest model is as follows: extract one feature from multiple features included in the samples with different weights, randomly extract one sample from the sample set, and perform splitting based on the extracted feature and the sample.

[0009] The weighted extreme random forest model is obtained through the following training steps:

[0010] S201. Construct a sample data set.

[0011] S202. Randomly extract multiple samples as a sample set from the sample data set in a way of extraction and replacement.

[0012] S203. When splitting is required at each node of the decision tree, extract one feature from multiple features included in the samples with different weights, randomly extract one sample from the sample set, and perform splitting based on the extracted feature and the sample until the splitting termination condition is met.

[0013] S204. Each decision tree takes the class with the largest number of samples in the sample subset of the leaf nodes as the output.

[0014] Execute the operations in steps S202 - S204 for K times to obtain K weighted extreme decision trees, and the K weighted extreme decision trees form the weighted extreme random forest model, where K is an integer greater than 1.

[0015] In the above S203, the process of extracting one feature from multiple features included in the samples with different weights includes:

[0016] Calculate the extraction probability of all features: For the decision tree, when detecting the training for mode l k the extraction probability P kj of feature j is: is the complex conjugate of for the vortex light of mode l k for the beam of mode l m after being distorted by the turbulent transmission distance z of the beam, for the beam of mode l j after being distorted by the turbulent transmission distance z of the beam, and M is the total number of features.

[0017] Assign the extraction probability as the weight to the corresponding feature, and all the features with weights assigned form a feature set.

[0018] ​Randomly extract a feature from the feature set.

[0019] In the above solution, η kj and η km are both crosstalk factors, representing the probability of crosstalk in other modes. The probability of feature extraction is related to crosstalk. Therefore, the weighted extremely randomized forest model can adaptively adjust the splitting weights according to environmental changes, and can dynamically identify and respond to the impact of different features on communication quality, so as to maintain high performance and reliability in a complex and changeable communication environment. Through this adaptive adjustment, problems such as environmental noise and signal attenuation can be processed more effectively, ensuring the stability and efficiency of communication.

[0020] In a second aspect, the present invention provides a vortex optical communication demodulation system based on a weighted extremely randomized forest model, including:

[0021] A data receiving module, configured to receive the received powers of different modes obtained by demultiplexing the vortex optical multiplexed beam and convert them into digital signals;

[0022] An information demodulation module, configured to input the digital signals of the received powers of different modes into a pre-trained weighted extremely randomized forest model for demodulation to obtain a demodulated signal; the splitting method of the weighted extremely randomized forest model is: extract a feature from multiple features included in the sample with different weights, randomly extract a sample from the sample set, and perform splitting based on the extracted feature and sample.

[0023] In a third aspect, the present invention provides a computer program product, including computer-readable instructions, characterized in that the computer-readable instructions, when executed by a processor, implement the steps in the vortex optical communication demodulation method based on the weighted extremely randomized forest model of the present invention.

[0024] In a fourth aspect, the present invention provides a computer-readable storage medium including computer-readable instructions, characterized in that the computer-readable instructions, when executed by a processor, implement the steps in the vortex optical communication demodulation method based on the weighted extremely randomized forest model of the present invention.

[0025] In a fifth aspect, the present invention provides an electronic device, including: a memory for storing program instructions; a processor connected to the memory, executing the program instructions in the memory, and implementing the steps in the vortex optical communication demodulation method based on the weighted extremely randomized forest model of the present invention.

[0026] Compared with the prior art, the present invention has the following technical advantages:

[0027] The present invention is a machine learning demodulation algorithm with low complexity. It treats crosstalk as a kind of useful information for signal analysis, eliminates the need for operations such as equalization, and can directly make a decision after power analog-to-digital conversion of the received signal, greatly simplifying the communication system. Moreover, it can better prevent overfitting, and in a low-noise environment, the present invention has higher accuracy. Through the present invention, the demodulation accuracy and rate can be improved, the communication system can be simplified, the development cycle can be shortened, and the cost can be reduced, thus effectively improving the overall performance of the vortex optical multiplexing communication.

[0028] When the decision tree node splits, different weights are assigned to each feature during feature extraction. The weighting mechanism can highlight important features (generally, the greater the crosstalk factor, the greater the importance) without ignoring other features. Compared with the average extraction of the extremely randomized tree algorithm (each feature has the same probability of being extracted), it can make the key features more likely to be drawn while not ignoring other features, thereby improving the accuracy. It effectively solves the deficiencies in accuracy and limitations in adaptability of traditional digital demodulation methods, and at the same time avoids the high complexity problems commonly existing in machine learning algorithms. It can better cope with the complex environment of vortex optical multiplexing communication.

[0029] It can be seen from the feature extraction probability formula that the extraction probability of a feature is determined by the crosstalk factor, and the crosstalk factor will change with the change of the channel. Therefore, when detecting different modes, the extraction probability of the feature will be different. In practical applications, the WERF model can also adaptively adjust the division weights according to the changes in the environment, and can dynamically identify and respond to the impact of different features on the communication quality, so as to maintain high performance and reliability in a complex and changeable communication environment. Through this adaptive adjustment, problems such as environmental noise and signal attenuation can be more effectively processed, ensuring the stability and efficiency of communication.

[0030] The demodulation method of the present invention is applicable to the marine environment, providing a better demodulation solution for underwater communication, and is also applicable to environments such as the atmosphere and optical fibers.

[0031] For other advantages of the present invention, please refer to the relevant descriptions in the embodiment part. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0033] Figure 1 It is a flowchart of a vortex optical communication demodulation method based on a weighted extremely randomized tree model provided in the embodiment.

[0034] Figure 2 It is the training flow chart of the weighted extreme random forest model provided in the embodiment.

[0035] Figure 3 It is the structural diagram of the communication system.

[0036] Figure 4a 、 Figure 4b They are respectively the structural hierarchy diagrams of the weighted extreme decision tree and the weighted extreme random forest model.

[0037] Figure 5a It is the distribution schematic diagram of the example sample set under different features.

[0038] Figure 5b It is based on Figure 5a The schematic diagram of the splitting process of constructing a decision tree shown in the example.

[0039] Figure 6a 、 Figure 6b 、 Figure 6c They are respectively in the ocean test example SNR 16dB, SNR 6dB, The BER result diagrams under different numbers of training samples in the SNR 12dB environment algorithm.

[0040] Figure 7a 、 Figure 7b 、 Figure 7c They are respectively in the ocean test example 20 meters SNR 15dB, 20 meters SNR 10dB, The BER result diagrams under different numbers of decision trees in the 20 - meter SNR 14dB environment algorithm.

[0041] Figure 8a 、 Figure 8b 、 Figure 8c They are respectively in the ocean test example 20 meters SNR 6dB, 20 meters SNR 10dB, The BER result diagrams under different maximum depths in the 20 - meter SNR 16dB environment algorithm.

[0042] Figure 9a 、 Figure 9b 、 Figure 9c 、 Figure 9d 、 Figure 9e They are respectively in the ocean test example 20 meters, 40 meters, 60 meters, 80 meters, Bit error rate result diagram at different SNRs in a 20-meter environment.

[0043] Figure 10a 、 Figure 10b 、 Figure 10c Are respectively in the atmospheric test cases 1000 meters, 500 meters, Bit error rate result diagrams at different SNRs for 1000 meters.

[0044] Figure 11 Is the block diagram of the electronic device. Detailed implementation manner

[0045] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0046] Please refer to Figure 1 , a vortex optical communication demodulation method based on a weighted extreme random forest model provided in this embodiment includes the following steps:

[0047] S10. Receive the vortex optical multiplexed beam, demultiplex to obtain the received power of different modes, and convert it into a digital signal.

[0048] S20. Input the digital signals of the received powers of different modes into a pre-trained weighted extreme random forest model for demodulation to obtain the demodulated signals, that is, the transmitted information of different modes.

[0049] The weighted extreme random forest model is an innovation based on the extreme random forest algorithm. The training complexity of the random forest algorithm is O(n*log(n)), and the testing complexity is O(n). The training complexity of the extreme random forest algorithm is O(n), and the testing complexity is O(n). However, if the extreme random forest algorithm is directly used for demodulation, the bit error rate will be relatively high, especially in the case of medium-intensity crosstalk, the bit error rate will be even higher. The weighted extreme random forest algorithm reduces the bit error rate while having the advantage of low complexity of the extreme random forest algorithm by changing the splitting method of the random forest algorithm. The weighted extreme random forest is to extract a feature from the subset of samples (M features) with different weights when each node of the decision tree needs to be split, and then randomly extract a sample from this feature for splitting. Since the features are extracted from the subset of samples based on different weights, it has a certain pertinence, and thus the bit error rate can be reduced.

[0050] Please refer to Figure 2 , the above-mentioned weighted extreme random forest model is trained through the following steps:

[0051] S201. Construct a sample data set.

[0052] Refer to Figure 3 , at the transmitting end, the information to be sent is loaded onto the laser transmitter through modulation (such as using on-off keying OOK modulation); after the light beam generates different-mode vortex beams through different-mode vortex beam converters, the multiplexed beam is obtained through a vortex light multiplexer, and the multiplexed beam is sent through a transmitting antenna and reaches the receiving end through the channel; at the receiving end, according to the orthogonality of OAM, the vortex light multiplexed beam is demultiplexed to obtain the received power of different modes and converted into digital signals.

[0053] The transmitting end generates a string of signal sequences, and the digital signals of the received power of different modes obtained by demultiplexing form a sample. If the transmitting end generates several strings of signal sequences, several samples can be obtained, and all the samples constitute a sample data set.

[0054] The actual application scenario can also refer to Figure 3 , in actual applications, after the receiving end receives the power, the demodulation method of the present invention (WERF demodulation) can directly make a decision on the digital signal converted from analog to digital, saving a lot of preprocessing operations on the signal, so the structure of the communication system can be greatly simplified.

[0055] S202. Randomly select n samples as a sample set from the sample data set in a way of sampling with replacement (Bootstrap sampling). n is an integer greater than 1.

[0056] One sample set is used to construct (or train) a weighted extreme decision tree (WEDT). Repeating multiple times can construct multiple different weighted extreme decision trees.

[0057] It should be noted that each decision tree is independent of each other. The process of constructing the decision tree can be executed in parallel or serially. If it is executed serially, steps S202 - S204 are looped until a set number of decision trees are generated. If it is executed in parallel, the same operations are independently executed in parallel according to steps S202 - S204.

[0058] S203. When each node of the weighted extreme decision tree needs to be split, a feature is randomly selected from the M features included in the sample with different weights, and a sample is randomly selected from the sample set. The split is performed based on the selected sample and feature until the split termination condition is met.

[0059] Refer to Figure 4aTaking a simple structure as an example, the WEDT structure consists of a root node, internal nodes (i.e., decision nodes), and leaf nodes. Starting from the root node, adopting a "divide and conquer" strategy, the sample set is partitioned according to the splitting point selection mechanism. First, the splitting point is determined, that is, a specific value selected on a certain feature in the samples is determined. Then, according to this splitting point, the samples in the sample set are divided into two sample subsets according to the size of the feature values, and the two newly generated sample subsets are recursively iterated by the above method until the splitting termination condition is met. The position of the sample subset that no longer splits finally is the leaf node.

[0060] In this embodiment, the splitting termination conditions include: 1. Whether all samples belong to one type (each sample has a corresponding result, which represents high level or low level in communication. If all samples belong to one type, it means that in this node, if all are high / low levels, then there is no need to split anymore, and it can be directly determined that the samples falling in this node are high / low levels. For example Figure 5b the first leaf node shown); 2. Whether the maximum depth is reached; 3. Whether the minimum number of samples is reached. The traditional random forest algorithm also has a splitting termination condition, that is, whether the variance is less than eps. However, this splitting termination judgment condition is deleted in the solution of this embodiment because through experiments, the role of this parameter is too low. Setting it cannot significantly improve the accuracy, but it requires computational effort to calculate, and the cost performance is low. Therefore, this parameter is chosen to be removed.

[0061] The traditional random forest algorithm randomly selects m features (m < M) from the samples (M features), and then selects 1 attribute as the splitting attribute of this node from these m attributes by adopting a certain strategy (such as information gain). The extremely randomized random forest algorithm randomly selects one feature from M features (each feature has the same probability of being selected), and then randomly selects a sample from the sample set for splitting. Although the complexity of the extremely randomized random forest algorithm is lower, its randomness is strong, so the error rate is high.

[0062] In the solution of the present invention, one feature is extracted according to weights from M features, and a sample is randomly selected from the sample set for splitting. Therefore, the complexity is the same as that of the extremely randomized random forest. However, when extracting features, they are extracted with different weights instead of randomly, so it has a certain pertinence, and then the error rate can be greatly reduced.

[0063] It is easy to understand that extracting features and extracting samples are independent, and there is no order of execution. That is, one feature can be extracted from the M features included in a sample first, and then a sample is extracted from the sample set; or a sample can be extracted from the sample set first, and then one feature is extracted from the M features included in a sample.

[0064] S204. Each weighted extreme decision tree outputs the class with the largest number of samples in the sample subset of the leaf node. That is, in the leaf node, the class that occupies the majority of the sample quantity is determined as the output class.

[0065] In the testing stage, for each weighted extreme decision tree, the unknown sample starts from the root node and, according to the division of the decision node, finally reaches a certain leaf node and outputs the classification result of that leaf node.

[0066] Refer to Figure 4b , the weighted extreme random forest (WERF) model consists of multiple WEDTs. Each WEDT is independently trained and can process information in parallel. When performing prediction, each WEDT will output a prediction result for the same data respectively, and then summarize the prediction results of each WEDT, and use the voting mechanism to determine the class with the most votes as the final output.

[0067] As a typical representative of vortex beams, Laguerre-Gaussian beams are favored for the simplicity of their optical field expressions and ease of experimental generation. The mathematical expression of its optical field is as follows:

[0068]

[0069] In the formula: l is the mode number, r is the radius in the cylindrical coordinate; θ represents its azimuth angle; z is the propagation distance; p is the radial exponent; Rayleigh length λ is the wavelength of the beam; wave number ω0 is the waist radius of the Gaussian beam; is the generalized Laguerre polynomial.

[0070] The basic principle of vortex optical multiplexing communication is to use vortex beams with different OAM modes as the carriers of modulation signals and achieve the transmission of multiple signals through superposition. s m (t) is the modulation signal, and multiple multiplexed vortex beams U MUX (r, θ, t) can be expressed as:

[0071] When simulating the influence of the ocean or atmospheric environment on the beam propagation, the power spectrum inversion method is used to generate a series of random phase screens. This process can be roughly divided into the following three steps: First, construct a complex Gaussian matrix, where the elements are complex values randomly generated according to the Gaussian distribution, which represents the initial spatial distribution of the refractive index in the turbulent medium; then, use the channel power spectrum function to filter the complex Gaussian matrix, which ensures that the generated phase screen can accurately reflect the statistical characteristics of the channel; finally, by performing the inverse Fourier transform, the filtered complex Gaussian matrix is converted into a random phase screen. This phase screen will then be used to simulate the propagation process of the beam in the channel.

[0072] During the transmission of the vortex beam, mode crosstalk occurs regardless of whether it is in a fiber, atmospheric, or ocean environment, affecting communication. The demodulation algorithm of the present invention is applicable to the demodulation of all orbital angular momentum multiplexing communications. Hereinafter, the ocean environment will be taken as an example for introduction.

[0073] Here, the spatial power spectrum model of ocean turbulence refractive index fluctuation proposed by researchers such as Nikishov is adopted. This model comprehensively considers various factors such as temperature, salinity, and refractive index fluctuation, and is widely used to generate random phase screens of ocean turbulence. In a homogeneous and isotropic seawater environment, the expression of this model is as follows:

[0074]

[0075] In the formula: k x k y are the frequency components in the x-axis and y-axis directions in the spatial frequency domain respectively. ε is the turbulent kinetic energy dissipation rate per unit fluid, reflecting the intensity of ocean turbulence fluctuations; χ T is the mean square temperature dissipation rate, reflecting the influence of temperature fluctuations on ocean turbulence; ψ is the temperature-salinity gradient rate; ζ is the inner scale of ocean turbulence; other parameters are A T = 1.863×10 -2 , A S = 1.9×10 -4 , A TS = 9.341×10 -3 , δ = 8.284(κη) 4 / 3 + 12.978(κη) 2 .

[0076] The principle of vortex light multiplexing communication is to regard the vortex beam as the carrier of the modulation signal. s m (t) is the modulation signal corresponding to mode l m , is the Laguerre-Gaussian beam (vortex light) of mode l m . The multiplexed vortex beams U MUX (r, θ, t) carrying information can be expressed as

[0077] The multiplexed vortex beam after turbulence transmission is expressed as where is the beam distorted after the beam of mode l m has passed through the turbulence transmission distance z, and n(t) is the additive Gaussian white noise of the channel.

[0078] After the multiplexed beam after turbulence reaches the receiving end, demultiplexing can be performed according to the orthogonality between different modes of the vortex light. Therefore, mode lk Received signal y k can be expressed as:

[0079]

[0080] where is the complex conjugate of, and η km is the crosstalk factor, representing the probability that the feature m under the mode l k is crosstalked by other modes, and s k (t) is the modulation signal corresponding to the mode l k , and η kk represents the detection probability, and n k (t) is the output of the additive Gaussian white noise passing through the k-th distorted OAM state matching filter, and η km and n k (t) are expressed mathematically as follows:

[0081]

[0082] In matrix form, the signal received at the receiving end can be expressed as:

[0083]

[0084] which can be simplified to: Y = HS + N.

[0085] where Y is the received signal, S is the transmitted signal, H is the channel matrix, and N is the additive Gaussian white noise.

[0086] In the above step S203, the process of extracting a feature with different weights includes:

[0087] (1) Calculate the extraction probability of all features (the extraction probability is the probability that the feature is extracted): For the decision tree, when training to detect the mode l k , the extraction probability P kj of the feature j is: η kj is the crosstalk factor, representing the probability that the feature j under the mode l k is crosstalked by other modes, and can reflect the distortion of the vortex beam of different modes (including the mode l k ) on the vortex beam of the mode l k . is the distorted beam of the beam of the mode l j after passing through the turbulence transmission distance z.

[0088] (2) Assign the extraction probability as the weight to the corresponding feature, and all the features with weights form a feature set;

[0089] (3) Randomly extract a feature from the feature set.

[0090] Since different weights are assigned to each feature when extracting features, the weighting mechanism can highlight the key points without neglecting other features. Compared with the average extraction of the extreme random forest algorithm (the extraction probability of each feature is the same), it can make the key features more likely to be drawn, thus improving the accuracy.

[0091] In practical applications, when obtaining the crosstalk factor, a certain mode of vortex light can be transmitted only, and then the information received by all receivers is normalized to obtain the crosstalk factor corresponding to the mode. By repeatedly transmitting the vortex light of all modes, the crosstalk factors of all modes can be obtained. This weight probability is not absolute, and slight fluctuations will not cause an impact, so it is not afraid of noise interference. Depending on the situation, it can also be repeated several times and averaged. For example, as shown in the above matrix expression, when only transmitting modes s1, s2 - s n not transmitting, the received y1 - y n is η 11 * s1 + n ~ η 1n * s1 + n. By normalizing, the crosstalk factor can also be obtained to adjust the weight. Next, transmit mode s2 until s n to obtain all the crosstalk factors in the matrix.

[0092] During the training process, in step S203, detections will be performed for all modes. From the formula of the extraction probability, it can be seen that the extraction probability of a feature is determined by the crosstalk factor. Therefore, when detecting different modes, the extraction probability of the feature will be different. In practical applications, the WERF model can also adaptively adjust the division weight according to the changes in the environment, and can dynamically identify and respond to the impact of different features on the communication quality, so as to maintain high performance and reliability in a complex and changeable communication environment. Through this adaptive adjustment, problems such as environmental noise and signal attenuation can be processed more effectively, ensuring the stability and efficiency of communication. Therefore, the weighted extreme random forest model will adaptively adjust following the changes in the environment during use, rather than remaining fixed after one training.

[0093] In terms of complexity comparison, Table 1 shows the training and test complexities of various commonly used and low-complexity machine learning demodulation methods. n is the number of training samples, is the number of test samples. It can be seen that the Support Vector Machine (SVM) algorithm has a high complexity in the training phase, while the K-Nearest Neighbor (KNN) algorithm, although it does not require a training process, has a high complexity in the testing phase. In contrast, the RF algorithm has a lower maximum complexity than the former two. As for the more simplified ERF and WERF algorithms, they obviously have a lower complexity than RF. The WERF algorithm makes a change in the random extraction probability of the splitting attributes in the ERF algorithm, and the complexities of the two are the same.

[0094] Table 1: Comparison Table of Complexities of Several Low-Complexity Machine Learning Algorithms

[0095]

[0096] To facilitate a better understanding of the splitting method of the random forest model, a simple example is given here for illustration.

[0097] Please refer to Figure 5a and Figure 5b , Figure 5a where X1 represents the light intensity information received by Feature 1, X2 represents the light intensity information received by Feature 2, and the higher up, the greater the received power.

[0098] The circular symbol represents the high-level state of Mode 1, and the square symbol represents the low-level state of Mode 1. During the training process, the circle and the square are known states. The number in the middle of the figure represents the serial number of the sample.

[0099] During the training process, first start to judge from Feature 1 (X1). Samples with a Y X1 value (the value after analog-to-digital conversion of the light intensity corresponding to Feature X1 received at the receiving end) greater than or equal to 6 can be classified as high level (because the data in Figure 5a shows that samples with a Y X1 value greater than 6 are all circles, that is, in the high-level state. Therefore, if a newly emerging unknown sample falls within this range, it is very likely to be high level). Similarly, samples with a Y X1 value less than or equal to 3 are classified as low level (because samples within this range are all squares, that is, in the low-level state). However, for samples with serial numbers 6, 7, and 8, since there are both circles and squares among them, they cannot be directly judged. At this time, Feature 2 (X2) can be analyzed. At this time, only samples with serial numbers 6, 7, and 8 need to be considered because the level states of other samples have been determined. Observing Figure 5a it can be found that the Y X2 values of samples with serial numbers 7 and 8 are both greater than 5, while the Y X2 value (the value after analog-to-digital conversion of the light intensity corresponding to Feature X2 received at the receiving end) of the sample with serial number 6 is less than 5. Based on this, Y X2Samples with values greater than 5 are divided into squares, and those less than 5 are divided into circles. The specific splitting process can be referred to Figure 5b .

[0100] Figure 5a The triangles in

[0100] represent the unknown level state. After the model is trained, the unknown level state can be classified. For example, according to the above training results, it can be judged that the sample with serial number 13 corresponds to a circular symbol, representing the high level state of mode 1; while serial numbers 14 and 15 correspond to square symbols, representing the low level state of mode 1. This process shows the key importance of split feature selection. If features with weak correlation are initially selected for splitting, it will not only increase the computational burden in vain, but also may lead to misjudgment. The weighted feature selection method proposed in this embodiment effectively solves this problem, improving the efficiency and accuracy of the algorithm. Although more complex situations may be encountered in practical applications, by constructing numerous independent WEDTs, a high degree of accuracy and robustness can still be maintained.

[0101] From the above splitting process, it can be seen that in the demodulation method of the present invention, crosstalk is regarded as a useful information for signal analysis. Without operations such as equalization, it can directly make a decision after power modulus conversion of the received signal, reducing the complexity, greatly simplifying the communication system, shortening the development cycle, and reducing costs, thereby effectively improving the overall performance of vortex optical communication.

[0102] To test the application effect of the algorithm proposed in the present invention in the field of underwater vortex optical communication, detailed numerical simulation experiments were carried out with the aid of the MATLAB platform in the research. A multiplexing communication system of four-way underwater vortex optical was built, and the OOK modulation technology was adopted to directly load the information to be sent onto the laser generator. In this experiment, the focus was on studying the influence of communication distance and turbulence intensity on communication performance. By adjusting the parameter Z to simulate different communication distances, and changing the parameter to control the intensity of turbulence, while keeping other system parameters unchanged (refer to Table 2) to ensure the accuracy and consistency of the experiment. For the ocean environment, was set as strong turbulence, was set as medium turbulence. In the experiment, a data stream of 200,000 bits was transmitted each time, and 1000 transmission experiments were repeated under each set condition to calculate the average performance index, so as to obtain more representative experimental data.

[0103] Table 2: Parameter configuration table

[0104] Parameter Meaning Value l Vortex beam mode [1,2,3,4] p Vortex beam radial index 0 λ Gaussian beam wavelength 532nm <![CDATA[ω0]]> Beam waist radius of Gaussian beam <![CDATA[3×10 -3 > ψ Temperature - salinity gradient rate -4 ζ Inner scale of ocean turbulence <![CDATA[1×10 -3 > - Matrix size of phase screen and laser 512×512 Phase screen interval 10m

[0105] As Figure 6a - Figure 6cAs shown, it presents the comparison of the demodulation bit error rates of the RF (Random Forest) and WERF (Weighted Extreme Random Forest) algorithms when transmitting 20 meters under different numbers of training samples, with the number of decision trees being 25 and the maximum depth being 10, at various levels of turbulence intensity and signal-to-noise ratio. It is used to compare the dependence of the two algorithms on the number of training samples (Number of training samples). Figure 6a It is the case of a signal-to-noise ratio of 16 dB in a strong turbulence environment, Figure 6b 、 Figure 6c which are respectively the cases of signal-to-noise ratios of 6 dB and 12 dB in a medium turbulence environment. Through the comparative analysis of these three figures, it can be observed that under different transmission environments, the increase in the number of training samples has basically the same impact on the bit error rate convergence of the two algorithms. The number of training samples is a key factor affecting the algorithm complexity. When the number of training samples exceeds 1800, the bit error rate basically tends to converge. In addition, these figures also show that the two algorithms require relatively few training samples, which means that in practical applications, good demodulation effects can be achieved even with fewer training samples. Compared with other machine learning algorithms that require a large amount of training data, this indirectly reflects the low-complexity characteristic of the WERF algorithm.

[0106] As Figure 7a - Figure 7c shown, it presents the impact of different numbers of decision trees on the demodulation bit error rates of the RF and WERF algorithms when transmitting 20 meters under various levels of turbulence intensity and noise levels, with the number of training samples being 1500 and the maximum depth being 10. These figures aim to analyze the dependence of the two algorithms on the number of decision trees. Figure 7a 、 Figure 7b respectively present the cases of signal-to-noise ratios of 15 dB and 10 dB in a medium turbulence environment, while Figure 7c shows the case of a signal-to-noise ratio of 14 dB in a strong turbulence environment. From Figure 7a 、 Figure 7b it can be observed that when the noise level is low, in order to achieve the optimal demodulation effect, the WERF algorithm needs to configure more decision tree numbers, and its final performance may even exceed that of the RF algorithm. This finding indicates that under specific conditions, the WERF algorithm can effectively alleviate the problem of premature convergence of the RF algorithm in a low-noise environment. Generally speaking, the greater the turbulence intensity, the more decision tree numbers the algorithm requires. In most cases, the number of decision trees required for the two algorithms to achieve good effects is approximately around 30. However, decision trees can process information in parallel, so increasing the number of decision trees does not bring additional complexity.

[0107] As Figure 8a - Figure 8c ​​As shown, it presents the comparison of the demodulation error rates of the RF and WERF algorithms for different maximum depth settings under various turbulence intensities and noise levels when transmitting 20 meters with 1500 training samples and 25 decision trees. These charts aim to explore the dependence of the two algorithms on the maximum depth of the decision tree. Figure 8a , Figure 8b respectively show the cases of signal-to-noise ratios of 6 dB and 10 dB in a medium-turbulence environment, while Figure 8c shows the case of a noise of 16 dB in a strong-turbulence environment. From these three figures, it can be seen that in order to achieve better demodulation performance, the WERF algorithm requires a larger maximum depth than the RF algorithm, and this gap is approximately 3. To improve the fault tolerance of the algorithm, the maximum depths of both algorithms can be set to about 13 to achieve the convergence of the error rate. Especially from Figure 8b it can be observed that when the maximum depth exceeds 10, the demodulation performance of the WERF algorithm is better than that of the RF algorithm, which indicates that in some cases, the demodulation performance of the WERF algorithm is indeed better than that of the RF algorithm.

[0108] By deeply studying the influence of the three key parameters of the number of training samples, the number of decision trees, and the maximum depth on the convergence of the demodulation error rate, the results show that the WERF algorithm is more sensitive to the two parameters of the number of decision trees and the maximum depth than the RF algorithm, but the dependence on the key parameter of the number of training samples is basically the same. Therefore, the overall complexity of the WERF algorithm is still significantly lower than that of the RF algorithm. Since there are many parameters involved in the algorithm, the experimental results in this part can only reflect the influence trend of the parameters on the algorithm performance and cannot be used for the comparison of the optimal performance of the algorithms. The specific performance comparison will be based on Figure 9a - Figure 9e introduced.

[0109] In addition, in the research, the error rate performances of four algorithms, namely RF, ERF (Extremely Random Forest), WERF, and MMSE (Minimum mean square error), under different ocean turbulence intensities, transmission distances, and signal-to-noise ratios were deeply compared. In terms of the algorithm parameter settings, according to the results in the previous algorithm parameter selection experiment, the parameters that can achieve the best demodulation performance were selected for different transmission environments. Figure 9a - Figure 9d They are respectively the signal-to-noise ratio SNR result charts for transmitting 20, 40, 60, and 80 meters under medium turbulence, Figure 9e and the signal-to-noise ratio SNR result chart for transmitting 20 meters under strong turbulence.

[0110] It can be seen from Figure 9a - Figure 9d that the RF algorithm and the WERF algorithm basically reach the forward error correction coding (FEC) of 3.8×10 -3Tolerance. In this regard, it can be said that within a transmission distance of 80 meters under medium-intensity turbulence, the demodulation effects of the two algorithms are similar, and the difference in the required signal-to-noise ratio will not exceed 0.5 dB. Overall, the signal-to-noise ratio required for the bit error rate to reach the FEC tolerance will increase with the increase of turbulence intensity and transmission distance. In these five environments shown in the figure, the bit error rate can reach the FEC tolerance at a signal-to-noise ratio of approximately 9.5 dB to 15 dB. In most cases, the RF algorithm has a better demodulation effect in a strong-noise environment, but the difference is not significant. In a low-noise environment, the WERF algorithm has a better demodulation effect, which is particularly obvious in an environment with a longer transmission distance or stronger turbulence intensity. The demodulation effect of the ERF algorithm will improve when the transmission distance is longer and under strong turbulence. Especially under strong turbulence intensity, it can reach the demodulation effect of the RF algorithm. However, overall, the applicable conditions of this algorithm are limited, and in most cases, the demodulation effect is much lower than that of the WERF algorithm. The MMSE algorithm has a better demodulation effect than RF and WERF in a low-noise environment within a transmission distance of 40 meters under medium-intensity turbulence. However, for the FEC tolerance, its effect is much lower than that of RF and WERF, and this situation becomes more obvious as the transmission distance increases, with the signal-to-noise ratio requirement gap between 2 dB and 10 dB. Considering the overall situation, the applicable environment of the MMSE algorithm is relatively limited, far less applicable than the RF and WERF algorithms. Figure 9e It can be seen that in a strong-turbulence environment with a transmission distance of 20 meters, the demodulation effects of the RF and WERF algorithms are similar to those of transmitting 80 meters under medium turbulence, and the demodulation effect of the MMSE algorithm is significantly lower than that of the other three algorithms. The signal-to-noise ratios of the three algorithms, RF, ERF, and WERF, are approximately 15 dB when reaching the FEC threshold. In a low-noise environment, the WERF algorithm has the best demodulation effect. Through comprehensive comparison, the WERF algorithm can not only meet the accuracy of the RF algorithm but also has the low complexity of the ERF algorithm, and can adapt to various transmission environments. It is an algorithm that can better meet the demodulation of underwater vortex optical multiplexing communication.

[0111] To verify the application effect of the algorithm proposed in the present invention in the field of atmospheric vortex optical communication, detailed numerical simulation experiments were carried out with the help of the MATLAB platform in the research. A four-channel multiplexing communication system for atmospheric vortex light was built, and the OOK modulation technology was adopted to directly load the information to be sent onto the laser generator. In this experiment, the focus was on studying the influence of communication distance and turbulence intensity on communication performance. By adjusting the parameter Z to simulate different communication distances and changing the parameter to control the turbulence intensity while keeping other system parameters unchanged to ensure the accuracy and consistency of the experiment. According to the above modeling of the ocean communication system, by replacing the ocean power spectrum with the atmospheric turbulence power spectrum, the conversion of the channel can be realized through the same steps. The atmospheric turbulence power spectrum used is as follows:

[0112]

[0113] In the formula, k x k y are the frequency components in the x-axis and y-axis directions in the spatial frequency domain, κ0 = 2π / L0, κ l = 3.3 / l0, κ L = 1 / L0, where L0 and l0 are the outer scale and inner scale of the turbulence respectively.

[0114] Under the atmospheric turbulence experiment, Figure 10a - Figure 10c are respectively for transmission of 1000 m, for transmission of 500 m, for transmission of 1000 m, the BER result diagrams for different algorithms and different SNRs. The experimental parameter settings are L0 = 50 m, l0 = 1 mm, the beam wavelength is set to 1550 nm, and the phase screen interval is 100 m. Other parameters involved are the same as those in Table 2. The parameters of the RF, ERF, and WERF algorithms are all selected, with the number of training samples being 1800, the decision tree being 30, and the maximum depth being 10. The comparison of the demodulation BER is as shown in the figure. According to the FEC threshold, when transmitting 1000 m, the WERF algorithm can reach the FEC threshold at about SNR = 9 dB, which is basically the same as the RF algorithm, about 3 dB less than the MMSE algorithm, and much lower than the ERF algorithm. When transmitting 500 m, the WERF can reach the FEC threshold at about SNR 14 dB, which is basically the same as the RF algorithm, about 2 dB less than the ERF, and about 7 dB less than the MMSE algorithm. When transmitting 1000 m, the WERF can reach the FEC threshold at about SNR 17 dB, which is about 1 dB less than the RF, about 5 dB less than the ERF, and about 7 dB less than the MMSE algorithm. It can be seen that the demodulation method and the WERF proposed in this embodiment are also applicable to the demodulation of vortex optical multiplexing communication in atmospheric turbulence and have good demodulation effects.

[0115] Based on the same inventive concept, in this embodiment, a vortex optical communication demodulation system based on a weighted extreme random forest model is also provided, including a data receiving module and an information demodulation module. Among them, the data receiving module is used to demultiplex the received vortex optical multiplexed beam to obtain the received power of different modes and convert it into a digital signal; the information demodulation module is used to input the digital signals of the received power of different modes into a pre-trained weighted extreme random forest model for demodulation to obtain a demodulated signal; the splitting method of the weighted extreme random forest model is: extracting a feature from multiple features included in the sample with different weights, randomly extracting a sample from the sample set, and performing splitting based on the extracted feature and sample.

[0116] The specific processing procedure of the information demodulation module may refer to the relevant descriptions in the foregoing method, which will not be elaborated here.

[0117] As Figure 11 shown, this embodiment also provides an electronic device, which may include a processor 41 and a memory 42, where the memory 42 is coupled to the processor 41. It should be noted that this figure is exemplary, and other types of structures can also be used to supplement or replace this structure to implement data extraction, report generation, communication, or other functions.

[0118] As Figure 11 shown, the electronic device may further include: an input unit 43, a display unit 44, and a power supply 45. It should be noted that the electronic device does not necessarily have to include Figure 11 all the components shown in Figure 11 In addition, the electronic device may further include components not shown in

[0119] The processor 41 is sometimes also referred to as a controller or an operation control unit, and may include a microprocessor or other processor devices and / or logic devices. The processor 41 receives inputs and controls the operations of the various components of the electronic device.

[0120] Among them, the memory 42 may be, for example, one or more of a buffer, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory, or other suitable devices, and can store information such as the configuration information of the foregoing processor 41 and the instructions executed by the processor 41. The processor 41 may execute the programs stored in the memory 42 to implement information storage or processing, etc. In one embodiment, the memory 42 further includes a buffer memory, that is, a buffer, to store intermediate information.

[0121] This embodiment of the present invention also provides a computer program product, including computer-readable instructions. When the computer-readable instructions are executed in an electronic device, the program product causes the electronic device to execute the operation steps included in the method of the present invention.

[0122] This embodiment of the present invention also provides a storage medium storing computer-readable instructions, and the computer-readable instructions cause the electronic device to execute the operation steps included in the method of the present invention.

[0123] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0124] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0125] The above-described embodiments are only specific implementation manners of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications, substitutions, and improvements, etc. These modifications, substitutions, and improvements should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A demodulation method for vortex optical communication based on a weighted extreme random forest model, characterized in that, It includes the following steps: S10, Demultiplex the received vortex optical multiplexed beam to obtain the received power of different modes, and convert it into a digital signal; S20, Input the digital signals of the received power of different modes into a pre-trained weighted extreme random forest model for demodulation to obtain a demodulated signal; The splitting method of the weighted extreme random forest model is: Extract a feature from multiple features included in the sample with different weights, randomly extract a sample from the sample set, and perform splitting based on the extracted feature and the sample.

2. The demodulation method of vortex optical communication based on the weighted extreme random forest model according to claim 1, wherein The weighted extreme random forest model is obtained through the following steps of training: S201, Construct a sample data set; S202, Randomly extract multiple samples as a sample set from the sample data set in a way of extraction and replacement; S203, When each node of the decision tree needs to be split, extract a feature from multiple features included in the sample with different weights, randomly extract a sample from the sample set, and perform splitting based on the extracted feature and the sample until the splitting termination condition is met; S204, Each decision tree takes the category with the largest number of samples in the sample subset of the leaf node as the output; Execute the operations in steps S202 - S204 K times to obtain K weighted extreme decision trees, and the K weighted extreme decision trees form the weighted extreme random forest model, where K is an integer greater than 1.

3. A demodulation method for vortex optical communication based on a weighted extreme random forest model according to claim 2, characterized in that In the S201, the processing of constructing the sample data set includes: At the transmitting end, load the information to be sent onto the laser transmitter through modulation; After the light beam generates different mode vortex light beams through different mode vortex beam converters, it passes through a vortex optical multiplexer to obtain a multiplexed beam, and the multiplexed beam is sent through a transmitting antenna and reaches the receiving end through a channel; The receiving end demultiplexes the vortex optical multiplexed beam according to the orthogonality of OAM to obtain the received power of different modes, and converts it into a digital signal to form a sample; Loop and execute the above operations until a set number of samples are obtained, and all the samples form a sample data set.

4. A demodulation method for vortex optical communication based on a weighted extreme random forest model according to claim 2, characterized in that, In the S203, the processing of extracting a feature from multiple features included in the sample with different weights includes: Calculate the extraction probability of all features: For the decision tree, when training the detection of pattern l k the extraction probability P of feature j kj is: is the complex conjugate of, is the vortex light of pattern l k , is the beam distorted after the beam of pattern l m propagates through the turbulence transmission distance z, is the beam distorted after the beam of pattern l j propagates through the turbulence transmission distance z, and M is the total number of features; Assign the extraction probability as the weight to the corresponding feature, and all the features assigned with weights form a feature set; Randomly extract a feature from the feature set.

5. A demodulation method for vortex optical communication based on a weighted extreme random forest model according to claim 2, characterized in that In the S203, the splitting termination condition is one of the following three:

1. Whether all samples belong to the same category; 2. Whether the maximum depth is reached; 3. Whether the minimum number of samples is reached.

6. A vortex optical communication demodulation system based on a weighted extreme random forest model, characterized in that, It includes: A data receiving module, which is used to demultiplex the received vortex optical multiplexed beam to obtain the received power of different modes, and convert it into a digital signal; An information demodulation module, which is used to input the digital signals of the received power of different modes into a pre-trained weighted extreme random forest model for demodulation to obtain a demodulated signal; The splitting method of the weighted extreme random forest model is: Extract a feature from multiple features included in the sample with different weights, randomly extract a sample from the sample set, and perform splitting based on the extracted feature and the sample.

7. A vortex optical communication demodulation system based on a weighted extreme random forest model according to claim 6, characterized in that When extracting features, the information demodulation module performs the following operations: Calculate the extraction probability of all features. For the decision tree, when training to detect pattern l k the extraction probability P of feature j kj is: is the complex conjugate of, is the vortex light of pattern l k and, is the beam distorted after the beam of pattern l m propagates through the turbulent transmission distance z, is the beam distorted after the beam of pattern l j propagates through the turbulent transmission distance z, and M is the total number of features; Assign the extraction probability as the weight to the corresponding feature, and all the features assigned with weights form a feature set; Randomly extract a feature from the feature set.

8. A computer program product comprising computer-readable instructions, characterized in that, The computer-readable instructions, when executed by a processor, implement the steps in the vortex optical communication demodulation method based on a weighted extremely randomized trees model according to any one of claims 1-5.

9. A computer-readable storage medium including computer-readable instructions, characterized in that, The computer-readable instructions, when executed by a processor, implement the steps in the vortex optical communication demodulation method based on a weighted extremely randomized trees model according to any one of claims 1-5.

10. An electronic device, characterized in that, Comprising: a memory for storing program instructions; a processor connected to the memory, executing the program instructions in the memory, and implementing the steps in the vortex optical communication demodulation method based on a weighted extremely randomized trees model according to any one of claims 1-5.

Images