Data-driven underwater acoustic communication modulation design method and device

Abstract

The invention relates to the technical field of underwater acoustic communication modulation, in particular to a data-driven underwater acoustic communication modulation design method and device. The method comprises the following steps: generating a training set and a verification set of a channel matrix of underwater acoustic communication; outputting a corresponding modulation matrix based on a neural network structure by using at least two channel matrix samples in the training set, calculating a loss function and a corresponding gradient, and performing gradient accumulation to generate an accumulated gradient; and if the total sampling frequency reaches a certain batch, updating the network by using an accumulated gradient, clearing the gradient, starting a new round until a training batch and a training round are reached, obtaining a to-be-verified underwater acoustic communication modulation model, and verifying by using a verification set to construct a final underwater acoustic communication modulation model. And outputting the data-driven underwater acoustic communication modulation matrix. Therefore, the problem of poor robustness under time delay-Doppler scale domain dual extension caused by inter-subcarrier interference in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of underwater acoustic communication modulation technology, and in particular to a data-driven underwater acoustic communication modulation design method and apparatus. Background Technology

[0002] Integrated air-space-ground-sea networks are considered a key development direction for next-generation wireless communication. As a crucial component, the reliability of underwater communication plays a vital supporting role in the stability of the entire network. In underwater communication, sound waves, due to their low propagation loss, have become the most commonly used underwater communication medium for medium to long distances.

[0003] Among related technologies, OFDM (Orthogonal Frequency Division Multiplexing) modulation technology can divide the channel into multiple orthogonal subcarriers for parallel data transmission, eliminating multipath effects through cyclic prefixes. OTFS (Orthogonal Time Frequency and Space), ODDM (Orthogonal Delay-Doppler Diversity Multiplexing), and AFDM (Affine Frequency Division Multiplexing) modulation techniques are suitable for high-speed mobile scenarios. Meanwhile, deep learning, due to its powerful optimization capabilities, is increasingly valued in communication system design.

[0004] However, among related technologies, modulation techniques such as OFDM generate severe inter-carrier interference, which seriously restricts system reliability and has insufficient anti-interference capability. Modulation techniques such as OTFS, ODDM, and AFDM are generally designed for narrowband dual-dispersion channels and cannot cope with the Doppler scale domain extension characteristics in underwater acoustic communication. They have poor robustness under the dual extension of time delay and Doppler scale domain, which urgently needs to be solved. Summary of the Invention

[0005] This application provides a data-driven underwater acoustic communication modulation design method and apparatus to solve the problems of poor robustness under the dual extension of time delay-Doppler scale domain caused by inter-carrier interference in related technologies.

[0006] The first aspect of this application provides a data-driven underwater acoustic communication modulation design method, comprising the following steps: generating a training set and a validation set for the channel matrix of underwater acoustic communication based on a preset training set size, a validation set size, the maximum delay spread of the channel, and the maximum Doppler scale spread; setting a neural network structure, and setting training rounds and training batches, so as to use at least two channel matrix samples in the training set to output the corresponding modulation matrix based on the neural network structure, calculate the loss function and the corresponding gradient, and perform gradient accumulation to generate a cumulative gradient; if the total number of sampling times has reached the preset batch size, then using the cumulative gradient to update the network and clear the gradient, start a new round, until the training batch and the training round are reached, to obtain the underwater acoustic communication modulation model to be validated, and using the validation set to validate the underwater acoustic communication modulation model to be validated, so as to construct the final underwater acoustic communication modulation model and output the data-driven underwater acoustic communication modulation matrix.

[0007] Through the above technical means, the embodiments of this application can use the channel matrix samples in the training set to output the corresponding modulation matrix in each sampling, calculate the loss function, and after reaching the training end condition, use the validation set to verify, and finally obtain the data-driven underwater acoustic communication modulation matrix, complete the underwater acoustic communication modulation design, obtain the underwater acoustic communication modulation matrix design result, eliminate the need to deploy neural networks in the actual system, reduce deployment complexity, improve the practicality of modulation technology, and improve the anti-interference capability of modulation technology and communication reliability under the dual extension of time delay-Doppler scale domain by calculating the loss function.

[0008] Optionally, in one embodiment of this application, the generation formula for the at least two channel matrix samples may be, but is not limited to, the following:

[0009]

[0010] Where n and n′ are the time-domain sampling indices for reception and transmission, respectively, P is the number of multipath paths, p is the path index, and h p Let a be the gain of the p-th path. p For the time delay, τ p f is the Doppler scaling factor parameter. c Let T be the carrier frequency at which the system operates, T be the Nyquist sampling period of the system, and g(t) be the specific shaping filter used by the system.

[0011] Through the above technical means, the embodiments of this application can generate channel matrix samples based on multi-dimensional physical parameters, improve the physical authenticity and representativeness of the channel matrix samples, broaden the scenario coverage of the channel matrix samples, and then train based on the channel matrix samples to enhance the robustness of the underwater acoustic communication modulation model.

[0012] Optionally, in one embodiment of this application, the loss function may be, but is not limited to:

[0013]

[0014] Where λ is the weighting coefficient, q(F1,F2) represents the consistency between the two, f(H,F) is the index for measuring system performance, H1 and H2 are both channel matrix samples, and F1 and F2 are the modulation matrices corresponding to H1 and H2.

[0015] Through the above technical means, the embodiments of this application can calculate the loss function through the channel matrix sample and the modulation matrix corresponding to the channel matrix sample, which can accurately capture the actual error of the modulation matrix under the channel matrix sample, avoid deviations in underwater acoustic communication modulation design, and guide the direction of underwater acoustic communication modulation design by calculating the loss function, thereby improving the reliability of modulation technology.

[0016] Optionally, in one embodiment of this application, the step of using the validation set to validate the underwater acoustic communication modulation model to construct the final underwater acoustic communication modulation model includes: inputting each channel matrix sample of the validation set into the underwater acoustic communication modulation model to be validated to output the corresponding modulation matrix; calculating the matrix mean based on the corresponding modulation matrix; calculating the modulation matrix design result according to the matrix mean, wherein if the modulation matrix design result satisfies the preset validation conditions, then the final underwater acoustic communication modulation model is determined.

[0017] Through the above technical means, the embodiments of this application can input each channel matrix sample of the validation set into the underwater acoustic communication modulation model to be validated, thereby calculating the modulation matrix design result. It can be verified through the validation set to ensure the scenario generalization of the modulation technology and avoid overfitting during training. It can calculate the mean of the modulation matrix to improve the robustness of the modulation technology. It can calculate the modulation matrix design result to give the modulation matrix orthogonality and improve the anti-interference capability of the modulation technology.

[0018] A second aspect of this application provides a data-driven underwater acoustic communication modulation design apparatus, comprising: a generation module, configured to generate a training set and a validation set for an underwater acoustic communication channel matrix based on a preset training set size, a validation set size, the maximum delay spread of the channel, and the maximum Doppler scale spread; a training module, configured to set a neural network structure, and set training rounds and training batches, so as to use at least two channel matrix samples in the training set to output the corresponding modulation matrix based on the neural network structure, calculate the loss function and the corresponding gradient, and perform gradient accumulation to generate a cumulative gradient; and a design module, configured to update the network and clear the gradient using the cumulative gradient if the total number of samplings has reached the preset batch size, and start a new round until the training batch and the training round are reached, to obtain an underwater acoustic communication modulation model to be validated, and to validate the underwater acoustic communication modulation model to be validated using the validation set, so as to construct the final underwater acoustic communication modulation model and output a data-driven underwater acoustic communication modulation matrix.

[0019] Through the above technical means, the embodiments of this application can use the channel matrix samples in the training set to output the corresponding modulation matrix in each sampling, calculate the loss function, and after reaching the training end condition, use the validation set to verify, and finally obtain the data-driven underwater acoustic communication modulation matrix, complete the underwater acoustic communication modulation design, obtain the underwater acoustic communication modulation matrix design result, eliminate the need to deploy neural networks in the actual system, reduce deployment complexity, improve the practicality of modulation technology, and improve the anti-interference capability of modulation technology and communication reliability under the dual extension of time delay-Doppler scale domain by calculating the loss function.

[0020] Optionally, in one embodiment of this application, the generation formula for the at least two channel matrix samples may be, but is not limited to, the following:

[0021]

[0022] Where n and n′ are the time-domain sampling indices for reception and transmission, respectively, P is the number of multipath paths, p is the path index, and h p Let a be the gain of the p-th path. p For the time delay, τ p f is the Doppler scaling factor parameter. c Let T be the carrier frequency at which the system operates, T be the Nyquist sampling period of the system, and g(t) be the specific shaping filter used by the system.

[0023] Through the above technical means, the embodiments of this application can generate channel matrix samples based on multi-dimensional physical parameters, improve the physical authenticity and representativeness of the channel matrix samples, broaden the scenario coverage of the channel matrix samples, and then train based on the channel matrix samples to enhance the robustness of the underwater acoustic communication modulation model.

[0024] Optionally, in one embodiment of this application, the loss function may be, but is not limited to:

[0025]

[0026] Where λ is the weighting coefficient, q(F1,F2) represents the consistency between the two, f(H,F) is the index for measuring system performance, H1 and H2 are both channel matrix samples, and F1 and F2 are the modulation matrices corresponding to H1 and H2.

[0027] Through the above technical means, the embodiments of this application can calculate the loss function through the channel matrix sample and the modulation matrix corresponding to the channel matrix sample, which can accurately capture the actual error of the modulation matrix under the channel matrix sample, avoid deviations in underwater acoustic communication modulation design, and guide the direction of underwater acoustic communication modulation design by calculating the loss function, thereby improving the reliability of modulation technology.

[0028] Optionally, in one embodiment of this application, the design module includes: an input unit, configured to input each channel matrix sample of the verification set into the underwater acoustic communication modulation model to be verified, so as to output the corresponding modulation matrix; a first calculation unit, configured to calculate the matrix mean based on the corresponding modulation matrix; and a second calculation unit, configured to calculate the modulation matrix design result based on the matrix mean, wherein if the modulation matrix design result satisfies the preset verification conditions, then the final underwater acoustic communication modulation model is determined.

[0029] Through the above technical means, the embodiments of this application can input each channel matrix sample of the validation set into the underwater acoustic communication modulation model to be validated, thereby calculating the modulation matrix design result. It can be verified through the validation set to ensure the scenario generalization of the modulation technology and avoid overfitting during training. It can calculate the mean of the modulation matrix to improve the robustness of the modulation technology. It can calculate the modulation matrix design result to give the modulation matrix orthogonality and improve the anti-interference capability of the modulation technology.

[0030] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the data-driven underwater acoustic communication modulation design method as described in the above embodiments.

[0031] A fourth aspect of this application provides a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the data-driven underwater acoustic communication modulation design method described above.

[0032] A fifth aspect of this application provides a computer program product, including a computer program that, when executed, implements the data-driven underwater acoustic communication modulation design method described above.

[0033] This application embodiment can generate training and validation sets for the channel matrix of underwater acoustic communication. Using at least two channel matrix samples from the training set, a corresponding modulation matrix is ​​output based on a neural network structure. The loss function and corresponding gradient are then calculated, and gradient accumulation is performed to generate a cumulative gradient. If the total number of samplings reaches a certain batch size, the network is updated using the cumulative gradient, and the gradient is cleared to zero, starting a new round. This process continues until the training batch and training rounds are reached, resulting in the underwater acoustic communication modulation model to be validated. The validation set is then used to validate the underwater acoustic communication modulation model, constructing the final underwater acoustic communication modulation model. The final output is a data-driven underwater acoustic communication modulation matrix, which can obtain the design results of the underwater acoustic communication modulation matrix. This eliminates the need to deploy neural networks in the actual system, reducing deployment complexity and improving the practicality of the modulation technology. By calculating the loss function, the anti-interference capability of the modulation technology and the communication reliability under the dual extension of the time-delay-Doppler scale domain are improved. Therefore, this solves the problem of poor robustness under the dual extension of the time-delay-Doppler scale domain caused by inter-carrier interference in related technologies.

[0034] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0035] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0036] Figure 1 This is a flowchart of a data-driven underwater acoustic communication modulation design method according to an embodiment of this application;

[0037] Figure 2 This is a flowchart of a data-driven underwater acoustic communication modulation design method according to an embodiment of this application;

[0038] Figure 3 This is a schematic diagram of a neural network structure according to an embodiment of this application;

[0039] Figure 4 This is a schematic diagram illustrating the calculation of the loss function during the training process according to an embodiment of this application;

[0040] Figure 5 This is a schematic diagram illustrating the bit error rate performance simulation according to an embodiment of this application;

[0041] Figure 6 This is a schematic diagram of a data-driven underwater acoustic communication modulation design apparatus according to an embodiment of this application;

[0042] Figure 7This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0043] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0044] The following description, with reference to the accompanying drawings, illustrates a data-driven underwater acoustic communication modulation design method and apparatus according to embodiments of this application. To address the issue of poor robustness under the dual extension of time delay and Doppler scale domain caused by inter-carrier interference mentioned in the background, this application provides a data-driven underwater acoustic communication modulation design method. In this method, a training set and a validation set of the underwater acoustic communication channel matrix can be generated. At least two channel matrix samples from the training set can be used to output the corresponding modulation matrix based on a neural network structure. Then, the loss function and corresponding gradient are calculated, and gradient accumulation is performed to generate a cumulative gradient. If the total number of sampling times has reached a certain batch size, the network is updated using the cumulative gradient, and the gradient is cleared to zero, starting a new round. This process continues until the training batch and training rounds are reached, resulting in the underwater acoustic communication modulation model to be validated. The validation set is then used to validate the underwater acoustic communication modulation model to construct the final underwater acoustic communication modulation model, ultimately outputting a data-driven underwater acoustic communication modulation matrix. This method can obtain the underwater acoustic communication modulation matrix design results without deploying a neural network in the actual system, reducing deployment complexity and improving the practicality of the modulation technology. By calculating the loss function, the anti-interference capability of the modulation technology and the communication reliability under the dual extension of time delay and Doppler scale domain can be improved. This solves the problem of poor robustness of related technologies under the dual extension of time delay and Doppler scale domain caused by inter-carrier interference.

[0045] Before introducing the data-driven underwater acoustic communication modulation design method provided in the embodiments of this application, the system model of the data-driven underwater acoustic communication modulation design method will be introduced first.

[0046] The underwater acoustic communication channel faced by the system has a maximum delay spread τ max With maximum Doppler scale extension a max Maximum Doppler scale extension a max satisfy Where v max The maximum moving speed is given by r, and the speed of sound is given by c. The signal model of the system is represented as r = HFx + w, where... The sequence is the time-domain sequence obtained by the receiver sampling at the Nyquist sampling frequency, where N is the sequence length at the transmitting end. g To protect the interval length, The channel matrix for underwater acoustic communication. Let F be the modulation matrix, which is a unitary matrix. H F = FF H =I N , where I N It is an N-order identity matrix. Let w be a data symbol vector, and w be a Gaussian white noise vector, where each element follows a mean of 0 and a variance of σ. 2 The cyclic symmetric complex Gaussian distribution.

[0047] Therefore, based on the above system model, this application proposes a data-driven underwater acoustic communication modulation design method and apparatus.

[0048] The following describes a data-driven underwater acoustic communication modulation design method provided by an embodiment of this application.

[0049] Specifically, Figure 1 This is a flowchart of a data-driven underwater acoustic communication modulation design method according to an embodiment of this application.

[0050] like Figure 1 As shown, this data-driven underwater acoustic communication modulation design method includes the following steps:

[0051] In step S101, the training set and validation set of the channel matrix for underwater acoustic communication are generated according to the preset training set size, validation set size, maximum channel delay spread, and maximum Doppler scale spread.

[0052] In the embodiments of this application, the training set can be understood as the core data for learning the neural network model, which contains a large number of known inputs (such as channel matrix samples) and desired outputs (or targets for optimization, such as ideal modulation performance indicators). The size of the training set is related to the complexity and dynamics of the underwater acoustic channel, the diversity and representativeness of the data, etc., and this application does not impose specific limitations.

[0053] In the embodiments of this application, the validation set can be understood as a separate set of data independent of the training set, used to evaluate the generalization ability of the neural network model (i.e., its adaptability to unseen data) during training and to adjust the hyperparameters of the neural network model (such as the dimension of the modulation matrix, the step size of the optimization algorithm, etc.). The size of the validation set is related to the complexity and dynamics of the underwater acoustic channel, as well as evaluation metrics (such as mean square error, bit error rate), etc., and this application does not impose specific limitations on them.

[0054] In the embodiments of this application, the maximum delay spread can reflect the maximum difference in arrival time of multipath signals in the channel, and is related to the number and distribution of paths, path length, environmental parameters, communication distance, etc., and is not specifically limited in this application. The maximum Doppler scale spread can reflect the maximum range of Doppler frequency shift in the channel, and is related to relative motion speed, path incident angle, environmental parameters, etc., and is not specifically limited in this application.

[0055] In actual implementation, the embodiments of this application can generate a training set and a validation set for the channel matrix of underwater acoustic communication based on a certain training set size, validation set size, maximum channel delay spread, and maximum Doppler scale spread.

[0056] For example, such as Figure 2 As shown, the main steps for generating the training and validation sets of the channel matrix for underwater acoustic communication can be as follows:

[0057] Step S201: Generate training set and validation set.

[0058] In this embodiment of the application, the training set size N can be determined according to a certain training set size. train With the size of the validation set N val and the maximum delay spread τ of the channel max Maximum Doppler scale extension a max Training and validation sets for generating the channel matrix of underwater acoustic communication.

[0059] The embodiments of this application can generate training and validation sets for the channel matrix of underwater acoustic communication based on various parameters, which can improve the scene coverage of the training and validation sets and enhance the adaptability and robustness of modulation design.

[0060] Optionally, in one embodiment of this application, the generation formula for at least two channel matrix samples may be, but is not limited to, the following:

[0061]

[0062] Where n and n′ are the time-domain sampling indices for reception and transmission, respectively, P is the number of multipath paths, p is the path index, and h p Let a be the gain of the p-th path. p For the time delay, τ p f is the Doppler scaling factor parameter. c Let T be the carrier frequency at which the system operates, T be the Nyquist sampling period of the system, and g(t) be the specific shaping filter used by the system.

[0063] In the embodiments of this application, channel matrix samples can be understood as data used to quantify the influence of the channel on signal transmission at a specific time or location, and are discretized data reflecting the random characteristics of the channel. Because the channel is random and time-varying, the attenuation, multipath distribution, Doppler effect, etc., of the channel may be completely different at the same location at different times, or at different locations at the same time. Therefore, a single channel matrix cannot fully describe the statistical characteristics of the channel, and channel matrix samples are needed to capture this diversity.

[0064] In some embodiments, the generation of channel matrix samples is based on the physical characteristics of multipath channels. The transmission process from transmitted to received signals (which may include, but is not limited to, the effects of attenuation, delay, Doppler scale spread, filtering, etc.) can be quantized into a matrix form. Through discretization, the transmission coefficients from the time-domain sampling points at the transmitting end to the time-domain sampling points at the receiving end are calculated, and finally a complete channel matrix sample is formed.

[0065] The following examples illustrate the process of generating the channel matrix. For each channel matrix in the training set, denoted as... Where 1≤i≤N train Let its index be in the training set, and its generating formula can be, but is not limited to, the following:

[0066]

[0067] Where n and n′ are the time-domain sampling indices for reception and transmission, respectively; P is the number of multipath paths, which can change or remain constant during the training set generation process; p is the path index; and h... p Let a be the gain of the p-th path. p For delay, -a max ≤a p ≤a max , τ p Let τ be the Doppler scaling factor parameter, 0 ≤ τ p ≤τ max h p a p and τ p The channel matrix is ​​generated randomly each time it is generated. The specific random generation scheme can be set by those skilled in the art according to the actual situation, and this application does not impose specific restrictions. c Let T be the carrier frequency at which the system operates, T be the Nyquist sampling period of the system, and g(t) be the specific shaping filter used by the system, such as a raised cosine filter. Finally, the channel matrix samples in the validation set are generated using the same scheme.

[0068] The embodiments of this application can generate channel matrix samples based on multi-dimensional physical parameters, improve the physical authenticity and representativeness of the channel matrix samples, broaden the scenario coverage of the channel matrix samples, and then train based on the channel matrix samples to enhance the robustness of the underwater acoustic communication modulation model.

[0069] In step S102, the neural network structure is set, and the training rounds and training batches are set so that the corresponding modulation matrix is ​​output based on the neural network structure using at least two channel matrix samples in the training set. The loss function and the corresponding gradient are calculated, and gradient accumulation is performed to generate the cumulative gradient.

[0070] In the embodiments of this application, the neural network structure can be understood as the overall framework of the basic building blocks, hierarchical divisions, connection rules, and core components that constitute an artificial neural network. Training epochs can be understood as the number of times the entire training set is fully learned, and training batches can be understood as dividing the training set into small data blocks, with the neural network model calculating the loss function using only one batch of data at a time, rather than the entire training set.

[0071] In the embodiments of this application, the gradient can be understood as the partial derivative of the loss function with respect to network parameters (such as weighting coefficients), which reflects the degree of influence of small changes in network parameters on the loss function. The direction of the gradient can reflect whether the network parameters should be increased or decreased to reduce the loss; the absolute value of the gradient can reflect the degree of influence of the network parameters on the loss.

[0072] In actual implementation, the embodiments of this application can define the calculation rules for the mapping between channel matrix samples and corresponding modulation matrices by setting the neural network structure, control the learning pace by setting training rounds and training batches, evaluate the error degree of the modulation matrix using the loss function, and adjust network parameters to reduce errors by calculating and accumulating gradients.

[0073] For example, such as Figure 2 As shown, in this embodiment of the application, the modulation matrix corresponding to the channel matrix sample is output, and then the loss function and the corresponding gradient are calculated. Gradient accumulation is then performed. The main steps for generating the accumulated gradient are as follows:

[0074] Step S202: Initialization.

[0075] Among them, the neural network structures that can be used in the embodiments of this application are as follows: Figure 3 As shown, M = N + N gTo determine the total number of samples at the receiver, the neural network first treats the real and imaginary parts of the complex channel matrix as two channels and performs a one-dimensional expansion of the total floating-point number. Then, it continuously extracts features through fully connected layers and activation function layers. Finally, it shapes the matrix and uses QR decomposition to obtain a unitary modulation matrix F matching the input H. The specific neural network structure can be configured by those skilled in the art according to actual conditions, and this application does not impose specific limitations.

[0076] Step S203: Start the training process.

[0077] In this embodiment, the training round e = 1.

[0078] Step S204: Training batch b = 1.

[0079] Step S205: Number of samplings t = 1.

[0080] Step S206: Randomly sample two channel matrix samples from the training set.

[0081] In this embodiment of the application, two channel matrix samples are randomly selected from the training set, denoted as H1 and H2.

[0082] Step S207: Calculate the loss function and corresponding gradient based on the same neural network structure.

[0083] Step S208: Accumulate gradients.

[0084] The embodiments of this application can calculate the loss function and gradient of the channel matrix and the corresponding modulation matrix, and then accumulate the gradient. This allows for the quantification of the compatibility between the modulation technique and the channel, clarification of the optimization direction of the modulation technique, and the use of gradients to capture the coupling degree between the channel matrix and the modulation matrix, thereby achieving precise optimization of the modulation technique. Accumulated gradients can also reduce noise interference in the underwater acoustic channel and improve the robustness of the modulation technique.

[0085] Optionally, in one embodiment of this application, the loss function may be, but is not limited to:

[0086]

[0087] Where λ is the weighting coefficient, q(F1,F2) represents the consistency between the two, f(H,F) is the index for measuring system performance, H1 and H2 are both channel matrix samples, and F1 and F2 are the modulation matrices corresponding to H1 and H2.

[0088] In some embodiments, the loss function can measure the consistency between two modulation matrices, measure the performance of the system to reflect the reliability of the modulation technique, and assign priority to different differences through weighting coefficients.

[0089] The following examples illustrate the process of calculating the loss function. Figure 4 As shown, H1 and H2 are passed through a pre-designed network, with the network weights remaining unchanged during the process, resulting in corresponding modulation matrix outputs F1 and F2. The loss function and corresponding gradients are then calculated and accumulated. The formula for calculating the loss function can be, but is not limited to, the following:

[0090]

[0091] Where λ is the weighting coefficient of the two terms, and q(F1,F2) represents the consistency between the two terms, designed as an NMSE (Normalized Mean Squared Error) function, i.e.

[0092]

[0093] in, Let f(H,F) represent the F norm of matrix A. F(H,F) is a metric for system performance, specifically representing the fairness of the MSE (Mean Square Error) after LMMSE (Linear Minimum Mean Square Error) equilibrium. Its calculation method is as follows:

[0094] Calculate the covariance matrix C after LMMSE equilibrium. H =σ 2 F H (H H H+σ 2 I N ) -1 F;

[0095] The MSE vector e of the calculated data is C H A vector consisting of the diagonal elements;

[0096] The optimal MSE vector is calculated as follows: Among them, trace(C) H ) is the covariance matrix C H traces, 1 N It is an N-dimensional vector in which all elements are 1;

[0097] calculate

[0098] in, This represents the 2-norm of vector a.

[0099] The embodiments of this application can calculate the loss function through the channel matrix sample and the modulation matrix corresponding to the channel matrix sample. This can accurately capture the actual error of the modulation matrix under the channel matrix sample, avoid deviations in the underwater acoustic communication modulation design, and improve the anti-interference capability of the modulation technology and the communication reliability under the dual extension of the time delay-Doppler scale domain by calculating the loss function.

[0100] In step S103, if the total number of samplings has reached the preset batch size, the network is updated and the gradient is cleared using the cumulative gradient, and a new round is started until the training batch and training round are reached, so as to obtain the underwater acoustic communication modulation model to be verified. The underwater acoustic communication modulation model to be verified is then verified using the validation set to construct the final underwater acoustic communication modulation model and output the data-driven underwater acoustic communication modulation matrix.

[0101] In some embodiments, the cumulative gradient reflects the sum of the direction and magnitude of the cumulative adjustments to network parameters within a certain batch size. Resetting the gradient to zero avoids confusion with the gradient of the next batch. After each batch of updates is completed, the neural network model continues to process the next batch of samples until all samples in the training set have been traversed, at which point one round is completed, and a new round begins. In a single round, the neural network model may only learn local features of the training set (such as the multipath distribution of some channel samples); multiple rounds of iteration allow the neural network model to repeatedly access all samples, gradually correcting network parameter deviations, and ultimately capturing the global patterns of the training set (such as the common feature of the time delay-Doppler scale domain dual extension of the underwater acoustic channel). When the number of training rounds reaches a certain value, the network parameters stop updating, and the embodiments of this application can obtain the underwater acoustic communication modulation model to be verified. The verification set screening eliminates overfitted underwater acoustic communication modulation models to be verified, ensuring that the final underwater acoustic communication modulation model can work reliably in complex and variable underwater acoustic environments to output a data-driven underwater acoustic communication modulation matrix.

[0102] For example, such as Figure 2 As shown in the embodiment of this application, when the total number of samplings has reached a certain batch size, the network can be updated and the gradient cleared using the cumulative gradient to start a new round until the underwater acoustic communication modulation model to be verified is obtained. The main steps for finally outputting the data-driven underwater acoustic communication modulation matrix can be as follows:

[0103] Step S209: Number of samplings t = t + 1.

[0104] Step S210: t < T.

[0105] In this embodiment, it is determined whether t is less than a certain batch size T. If t < T, the process returns to step S206; if t ≥ T, the process executes step S211.

[0106] Step S211: Update network parameters.

[0107] In this embodiment, the cumulative gradient can be used to update the network parameters and clear the gradient to zero.

[0108] Step S212: Training batch b = b + 1.

[0109] Step S213: b < B.

[0110] In this embodiment, it is determined whether b is less than the maximum number of training batches B. If b < B, return to step S205; if b ≥ B, set e = e + 1 and execute step S214.

[0111] Step S214: Training termination condition.

[0112] In this embodiment, it is determined whether the training termination condition has been met. If the training termination condition has not been met, the process returns to step S204; if the training termination condition has been met, step S215 is executed. The training termination condition may include, but is not limited to, the maximum number of training epochs, or the loss function on the validation set no longer decreasing after multiple epochs. The specific condition can be set by those skilled in the art according to the actual situation, and this application does not impose any specific limitations.

[0113] Step S215: Validate the underwater acoustic communication modulation model to be validated using the validation set.

[0114] After the training termination condition is met, the embodiment of this application can be verified using a validation set to finally obtain the data-driven underwater acoustic communication modulation matrix, complete the underwater acoustic communication modulation design, and obtain the underwater acoustic communication modulation matrix design result. There is no need to deploy a neural network in the actual system, which improves the practicality of the modulation technology.

[0115] Furthermore, in one embodiment of this application, the underwater acoustic communication modulation model to be verified is verified using a verification set to construct the final underwater acoustic communication modulation model. This includes: inputting each channel matrix sample of the verification set into the underwater acoustic communication modulation model to be verified to output the corresponding modulation matrix; calculating the matrix mean based on the corresponding modulation matrix; and calculating the modulation matrix design result based on the matrix mean. If the modulation matrix design result meets certain verification conditions, then the final underwater acoustic communication modulation model is determined.

[0116] In actual implementation, the embodiments of this application can verify the mapping between the channel matrix samples and the corresponding modulation matrix through the validation set, so as to check whether the neural network model can generate a reasonable modulation matrix in the untrained channel scenario, avoid the neural network model losing its generalization ability, and then calculate the matrix mean and the modulation matrix design result, which can smooth the interference in extreme scenarios, retain the common requirements of most scenarios, and thus determine the final underwater acoustic communication modulation model.

[0117] For example, embodiments of this application can perform verification on each channel matrix sample H in the verification set. i The modulation matrix F is obtained through the trained neural network model. i Then calculate the mean. Then calculate The QR decomposition yields the final modulation matrix design result F.

[0118] Specifically, such as Figure 5 As shown, the simulation verification results presented are for a channel environment with 8 paths, an average delay spread of 8ms, a carrier frequency of 12.5kHz, a system bandwidth of 5kHz, 256 transmitter sequence samples, 64 guard interval points, LMMSE equalization at the receiver, a maximum motion speed of 20kn, and training and validation set sizes of 20000 and 1000 respectively. The signal-to-noise ratio (SNR) used during training is 20dB. It should be noted that in this embodiment, after design completion, a uniform F is used in all SNR and channel tests without adjustment. It can be seen that the modulation in this embodiment achieves a bit error rate below 10... -2 It exhibited optimal performance across the entire range. When the signal-to-noise ratio reached 30dB, the bit error rate of the embodiments of this application was less than 10. -5 This result is also lower than that of ODDM modulation designed for narrowband dual-dispersion channels. Furthermore, considering that the number of channels at each signal-to-noise ratio point is 100,000, which far exceeds the size of the training and validation sets, this means that the embodiments of this application can not only handle underwater acoustic channels with delay-Doppler scale dual extension well, but also have good generalization in different channel environments.

[0119] In this embodiment, each channel matrix sample of the validation set can be input into the underwater acoustic communication modulation model to be validated, thereby calculating the modulation matrix design result. The modulation technology can be validated through the validation set, ensuring the scenario generalization of the modulation technology and avoiding overfitting during training. The mean of the modulation matrix can be calculated to improve the robustness of the modulation technology. The modulation matrix design result can be calculated to give the modulation matrix orthogonality and improve the anti-interference capability of the modulation technology.

[0120] The following section describes the data-driven underwater acoustic communication modulation design method proposed in this application, using a specific embodiment as an example.

[0121] Example 1:

[0122] The embodiments of this application can be based on a certain training set size N. train With the size of the validation set N val and the maximum delay spread τ of the channel max Maximum Doppler scale extension a max Generate the channel matrix training and validation sets.

[0123] Furthermore, in this embodiment, a neural network structure can be designed and initialized; the training process can be started and the training round e=1 and the training batch b=1 can be set. Two channel matrix samples can be randomly selected from the training set, denoted as H1 and H2; then H1 and H2 can be passed through the designed network respectively, with the network weights remaining unchanged during this period, to obtain the corresponding modulation matrix outputs F1 and F2 respectively; then the loss function and the corresponding gradient are calculated, and the gradients are accumulated.

[0124] Furthermore, when the total number of samplings t has reached a certain batch size T, this embodiment of the application can update the network using the accumulated gradients and clear the gradients to zero; otherwise, sampling is repeated. When the total number of training batches b has reached the maximum training batch B in each round, a new round begins; otherwise, training is repeated. When the termination training condition is met, this embodiment of the application can obtain a validated underwater acoustic communication modulation model, which is then validated using a validation set to construct the final underwater acoustic communication modulation model and output a data-driven underwater acoustic communication modulation matrix.

[0125] According to the data-driven underwater acoustic communication modulation design method proposed in this application, a training set and a validation set for the channel matrix of underwater acoustic communication can be generated. At least two channel matrix samples from the training set can be used to output the corresponding modulation matrix based on a neural network structure. Then, the loss function and corresponding gradient are calculated, and gradient accumulation is performed to generate a cumulative gradient. If the total number of sampling times has reached a certain batch size, the network is updated using the cumulative gradient, and the gradient is cleared to zero, starting a new round until the training batch and training rounds are reached, thus obtaining the underwater acoustic communication modulation model to be validated. The validation set is used to validate the underwater acoustic communication modulation model to construct the final underwater acoustic communication modulation model, ultimately outputting a data-driven underwater acoustic communication modulation matrix. This method can obtain the design results of the underwater acoustic communication modulation matrix without deploying a neural network in the actual system, reducing deployment complexity and improving the practicality of the modulation technology. By calculating the loss function, the anti-interference capability of the modulation technology and the communication reliability under the dual extension of the time delay-Doppler scale domain can be improved. Therefore, it solves the problem of poor robustness under the dual extension of the time delay-Doppler scale domain caused by inter-carrier interference in related technologies.

[0126] Next, with reference to the accompanying drawings, a data-driven underwater acoustic communication modulation design apparatus according to an embodiment of this application is described.

[0127] Figure 6 This is a schematic diagram of a data-driven underwater acoustic communication modulation design device according to an embodiment of this application.

[0128] like Figure 6 As shown, the data-driven underwater acoustic communication modulation design device 60 includes: a generation module 100, a training module 200, and a design module 300.

[0129] The generation module 100 is used to generate the training set and validation set of the channel matrix for underwater acoustic communication according to the preset training set size, validation set size, maximum channel delay spread, and maximum Doppler scale spread.

[0130] The training module 200 is used to set the neural network structure, training rounds and training batches, so as to use at least two channel matrix samples in the training set to output the corresponding modulation matrix based on the neural network structure, calculate the loss function and the corresponding gradient, and perform gradient accumulation to generate the cumulative gradient.

[0131] The design module 300 is used to update the network and clear the gradient by using the cumulative gradient if the total number of sampling times has reached the preset batch size, and start a new round until the training batch and training round are reached, so as to obtain the underwater acoustic communication modulation model to be verified, and to verify the underwater acoustic communication modulation model to be verified using the validation set, so as to construct the final underwater acoustic communication modulation model and output the data-driven underwater acoustic communication modulation matrix.

[0132] Optionally, in one embodiment of this application, the generation formula for at least two channel matrix samples may be, but is not limited to, the following:

[0133]

[0134] Where n and n′ are the time-domain sampling indices for reception and transmission, respectively, P is the number of multipath paths, p is the path index, and h p Let a be the gain of the p-th path. p For the time delay, τ p f is the Doppler scaling factor parameter. c Let T be the carrier frequency at which the system operates, T be the Nyquist sampling period of the system, and g(t) be the specific shaping filter used by the system.

[0135] Optionally, in one embodiment of this application, the loss function may be, but is not limited to:

[0136]

[0137] Where λ is the weighting coefficient, q(F1,F2) represents the consistency between the two, f(H,F) is the index for measuring system performance, H1 and H2 are both channel matrix samples, and F1 and F2 are the modulation matrices corresponding to H1 and H2.

[0138] Optionally, in one embodiment of this application, the design module 300 includes: an input unit, a first calculation unit, and a second calculation unit.

[0139] The input unit is used to input each channel matrix sample of the validation set into the underwater acoustic communication modulation model to be validated, so as to output the corresponding modulation matrix.

[0140] The first calculation unit is used to calculate the matrix mean based on the corresponding modulation matrix.

[0141] The second calculation unit is used to calculate the modulation matrix design result based on the matrix mean. If the modulation matrix design result meets the preset verification conditions, then please determine the final underwater acoustic communication modulation model.

[0142] It should be noted that the foregoing explanation of the data-driven underwater acoustic communication modulation design method embodiment also applies to the data-driven underwater acoustic communication modulation design device of this embodiment, and will not be repeated here.

[0143] The data-driven underwater acoustic communication modulation design device proposed in this application can generate training and validation sets for the channel matrix of underwater acoustic communication. It can use at least two channel matrix samples from the training set to output the corresponding modulation matrix based on a neural network structure, then calculate the loss function and corresponding gradient, and accumulate the gradient to generate a cumulative gradient. If the total number of samplings has reached a certain batch size, the network is updated using the cumulative gradient and the gradient is cleared, starting a new round until the training batch and training rounds are reached, thus obtaining the underwater acoustic communication modulation model to be validated. The validation set is then used to validate the underwater acoustic communication modulation model to construct the final underwater acoustic communication modulation model, ultimately outputting a data-driven underwater acoustic communication modulation matrix. This allows for the acquisition of underwater acoustic communication modulation matrix design results without the need to deploy neural networks in the actual system, reducing deployment complexity and improving the practicality of the modulation technology. By calculating the loss function, the anti-interference capability of the modulation technology and the communication reliability under the dual extension of the time delay-Doppler scale domain can be improved. Therefore, it solves the problem of poor robustness under the dual extension of the time delay-Doppler scale domain caused by inter-carrier interference in related technologies.

[0144] Figure 7 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. The electronic device may include:

[0145] The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0146] When the processor 702 executes the program, it implements the data-driven underwater acoustic communication modulation design method provided in the above embodiments.

[0147] Furthermore, electronic devices also include:

[0148] Communication interface 703 is used for communication between memory 701 and processor 702.

[0149] The memory 701 is used to store computer programs that can run on the processor 702.

[0150] The memory 701 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0151] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0152] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0153] The processor 702 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0154] This application also provides a non-volatile computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the data-driven underwater acoustic communication modulation design method described above.

[0155] This application also provides a computer program product, including a computer program that, when executed, implements the data-driven underwater acoustic communication modulation design method described above.

[0156] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0157] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0158] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0159] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0160] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or more of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0161] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0162] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0163] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A data-driven underwater acoustic communication modulation design method, characterized in that, Includes the following steps: The training set and validation set of the channel matrix for underwater acoustic communication are generated based on the preset training set size, validation set size, maximum channel delay spread, and maximum Doppler scale spread. Set up a neural network structure, and set up training rounds and training batches to use at least two channel matrix samples in the training set to output the corresponding modulation matrix based on the neural network structure, calculate the loss function and corresponding gradient, and perform gradient accumulation to generate a cumulative gradient. If the total number of samplings has reached the preset batch size, the network is updated and the gradient is cleared using the accumulated gradient, and a new round is started until the training batch and the training round are reached, so as to obtain the underwater acoustic communication modulation model to be verified. The underwater acoustic communication modulation model to be verified is then verified using the validation set to construct the final underwater acoustic communication modulation model and output the data-driven underwater acoustic communication modulation matrix.

2. The method according to claim 1, characterized in that, The formula for generating the at least two channel matrix samples is: Where n and n′ are the time-domain sampling indices for reception and transmission, respectively, P is the number of multipath paths, p is the path index, and h p Let a be the gain of the p-th path. p For the time delay, τ p f is the Doppler scaling factor parameter. c Let T be the carrier frequency at which the system operates, T be the Nyquist sampling period of the system, and g(t) be the specific shaping filter used by the system.

3. The method according to claim 1, characterized in that, The loss function is: Where λ is the weighting coefficient, q(F1,F2) represents the consistency between the two, f(H,F) is the index for measuring system performance, H1 and H2 are both channel matrix samples, and F1 and F2 are the modulation matrices corresponding to H1 and H2.

4. The method according to claim 1, characterized in that, The step of using the validation set to validate the underwater acoustic communication modulation model to construct the final underwater acoustic communication modulation model includes: Each channel matrix sample of the validation set is input into the underwater acoustic communication modulation model to be validated, so as to output the corresponding modulation matrix; Calculate the matrix mean based on the corresponding modulation matrix; Calculate the modulation matrix design result based on the mean of the matrix. If the modulation matrix design result meets the preset verification conditions, then determine the final underwater acoustic communication modulation model.

5. A data-driven underwater acoustic communication modulation design device, characterized in that, include: The generation module is used to generate the training set and validation set of the channel matrix for underwater acoustic communication based on the preset training set size, validation set size, maximum channel delay spread, and maximum Doppler scale spread. The training module is used to set the neural network structure, and set the training rounds and training batches, so as to use at least two channel matrix samples in the training set to output the corresponding modulation matrix based on the neural network structure, calculate the loss function and the corresponding gradient, and perform gradient accumulation to generate a cumulative gradient. The design module is used to update the network and clear the gradient to zero if the total number of samplings has reached the preset batch size, and start a new round until the training batch and the training round are reached, so as to obtain the underwater acoustic communication modulation model to be verified, and to verify the underwater acoustic communication modulation model to be verified using the verification set, so as to construct the final underwater acoustic communication modulation model and output the data-driven underwater acoustic communication modulation matrix.

6. The apparatus according to claim 5, characterized in that, The design module includes: The input unit is used to input each channel matrix sample of the verification set into the underwater acoustic communication modulation model to be verified, so as to output the corresponding modulation matrix; The first calculation unit is used to calculate the matrix mean based on the corresponding modulation matrix; The second calculation unit is used to calculate the modulation matrix design result based on the mean of the matrix. If the modulation matrix design result meets the preset verification conditions, then the final underwater acoustic communication modulation model should be determined.

7. The apparatus according to claim 5, characterized in that, The formula for generating the at least two channel matrix samples is: Where n and n′ are the time-domain sampling indices for reception and transmission, respectively, P is the number of multipath paths, p is the path index, and h p Let a be the gain of the p-th path. p For the time delay, τ p f is the Doppler scaling factor parameter. c Let T be the carrier frequency at which the system operates, T be the Nyquist sampling period of the system, and g(t) be the specific shaping filter used by the system.

8. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the data-driven underwater acoustic communication modulation design method as described in any one of claims 1-4.

9. A non-volatile computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the data-driven underwater acoustic communication modulation design method as described in any one of claims 1-4.

10. A computer program product, comprising a computer program, characterized in that, The computer program is executed to implement the data-driven underwater acoustic communication modulation design method as described in any one of claims 1-4.

Images