A real-signal OFDM underwater acoustic communication method and system

Abstract

This application discloses a real-signal OFDM underwater acoustic communication method and system, relating to the field of underwater acoustic communication technology. The method includes deploying a unitary neural network model at the transmitting end and a deep neural network model at the receiving end; jointly training the unitary neural network model and the deep neural network model to obtain trained unitary neural network and deep neural network models; inputting an information bitstream to the transmitting end and using the trained unitary neural network model to generate pre-coded data based on the information bitstream, which is then transmitted to the receiving end as a real OFDM signal; receiving the real OFDM signal at the receiving end and using the trained deep neural network model to perform signal detection to obtain the information bitstream. This application can reduce computational complexity and improve data detection performance.

Description

Technical Field

[0001] This application relates to the field of underwater acoustic communication technology, and in particular to a real-signal OFDM underwater acoustic communication method and system. Background Technology

[0002] Currently, in OFDM (Orthogonal Frequency Division Multiplexing) underwater acoustic communication (UWAC) systems, due to the complex characteristics of underwater acoustic channels, such as severe multipath effects and Doppler effects, achieving reliable underwater acoustic communication relies on accurate channel estimation and carrier frequency offset (CFO) estimation. However, in order to track these channel effects, a considerable portion of the subcarriers in OFDM underwater acoustic communication systems are used to carry overhead packets. These overhead packets consume limited bandwidth resources, resulting in a significant reduction in bandwidth efficiency.

[0003] To overcome the bandwidth efficiency degradation in OFDM underwater acoustic communication systems caused by additional overhead packets, researchers have proposed various methods. Some studies employ guard intervals to track the underwater acoustic channel; however, these methods assume channel sparsity, which is difficult to guarantee in practical applications, and performance is often affected by inter-block interference. Other studies propose using the Discrete Hartley Transform (DHT) instead of the Discrete Fourier Transform (DFT) to reduce the required overhead packets. However, these methods still require transmitting pilot sequences and may require additional empty subcarriers to track the CFO effect. Furthermore, Superimposed Training (ST) has been introduced as an alternative to pilot symbol-assisted modulation, overcoming the bandwidth reduction problem by superimposing overhead packets onto data symbols; however, this method is limited by interference between pilot and data signals. Therefore, some researchers proposed the data-nulling superimposed pilots method. However, this method only provides acceptable data detection performance when the pilot sequence is short. Practical underwater acoustic communication systems typically use 25% of OFDM symbols to carry the pilot sequence, leading to a sharp decline in the data detection performance of this method. In recent years, deep learning (DL) based methods have been introduced into underwater acoustic communication systems to solve the problems of channel estimation and data detection. However, due to the unpredictability of underwater acoustic channels and the long length of OFDM symbols in underwater acoustic communication, the computational complexity required for training and testing deep learning models is very high, which limits the practical application of this method.

[0004] In summary, current real-signal OFDM underwater acoustic communication methods generally suffer from computational complexity and low data detection performance. Therefore, providing a real-signal OFDM underwater acoustic communication method that is computationally simple and has high data detection performance has become a pressing technical problem to be solved in this field. Summary of the Invention

[0005] The purpose of this application is to provide a real-signal OFDM underwater acoustic communication method and system that can reduce computational complexity and improve data detection performance.

[0006] To achieve the above objectives, this application provides the following solution.

[0007] In a first aspect, this application provides a real-signal OFDM underwater acoustic communication method. The real-signal OFDM underwater acoustic communication method is used for underwater acoustic communication between a transmitter and a receiver. The transmitter is used to input an information bit stream and then send a real OFDM signal to the receiver. The receiver is used to obtain the information bit stream based on the real OFDM signal. The real-signal OFDM underwater acoustic communication method includes the following steps:

[0008] A unitary neural network model is deployed at the transmitting end, and a deep neural network model is deployed at the receiving end. The input of the unitary neural network model is the modulation data corresponding to the information bitstream, and the output is precoded data. The unitary neural network model is used to generate the precoded data based on the input information bitstream. The input of the deep neural network model is the detection data corresponding to the information bitstream, and the output is the detection result corresponding to the information bitstream. The deep neural network model is used to detect the information bitstream.

[0009] The unitary neural network model and the deep neural network model are jointly trained to obtain a trained unitary neural network model and a trained deep neural network model.

[0010] The information bitstream is input to the transmitting end, and the pre-coded data is generated based on the information bitstream using the trained unitary neural network model, and then sent to the receiving end in the form of the real OFDM signal.

[0011] The receiver receives the real OFDM signal and uses the trained deep neural network model to perform signal detection on the real OFDM signal to obtain the information bit stream.

[0012] Secondly, this application provides a real-signal OFDM underwater acoustic communication system, which includes a transmitter and a receiver. The transmitter is used to input an information bit stream and then send a real OFDM signal to the receiver. The receiver is used to obtain the information bit stream based on the real OFDM signal. When underwater acoustic communication is performed between the transmitter and the receiver, the real-signal OFDM underwater acoustic communication method described in the first aspect is executed.

[0013] According to the specific embodiments provided in this application, this application has the following technical effects.

[0014] This application provides a real-signal OFDM underwater acoustic communication method and system. By combining a unitary neural network model and a deep neural network model, the two models are deployed and jointly trained at the transmitting and receiving ends, respectively. By using the unitary neural network model to generate pre-coded data at the transmitting end, the method eliminates the need for the underwater acoustic channel during training, thereby reducing computational complexity and solving the problem of excessive computation during training in traditional methods. Simultaneously, this application also deploys a deep neural network model at the receiving end for signal detection, enabling rapid and effective detection of the information bitstream contained in the real OFDM signal, thus improving data detection performance and addressing the performance degradation issue of traditional methods. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a flowchart illustrating a real-signal OFDM underwater acoustic communication method provided in an embodiment of this application.

[0017] Figure 2 This is a schematic diagram of a signal processing flow provided in an embodiment of this application.

[0018] Figure 3 The bit error rate curves of various methods provided in an embodiment of this application under an additive white Gaussian noise channel.

[0019] Figure 4 The bit error rate curves of various methods provided in an embodiment of this application under a perfectly known underwater acoustic channel.

[0020] Figure 5 The bit error rate curves of various methods provided in an embodiment of this application under different CFO values.

[0021] Figure 6 The bit error rate curves of various methods provided in an embodiment of this application with and without the CFO compensation process. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0024] like Figure 1 As shown, this embodiment proposes a real-signal OFDM underwater acoustic communication method. This method is mainly applied in a real-signal OFDM underwater acoustic communication system, which includes a transmitter and a receiver. Underwater acoustic communication occurs between the transmitter and receiver. The transmitter inputs an information bit stream and then sends a real OFDM signal to the receiver. The receiver obtains the information bit stream based on the real OFDM signal. The specific steps of this real-signal OFDM underwater acoustic communication method are as follows:

[0025] Step S1: Deploy a unitary neural network (UNN) model at the transmitting end and a deep neural network (DNN) model at the receiving end.

[0026] This embodiment combines a unitary neural network model and a deep neural network model, deploying them at the transmitter and receiver ends respectively. The unitary neural network model takes modulation data corresponding to the information bitstream as input and outputs precoded data; it generates precoded data based on the input information bitstream. The deep neural network model takes detection data corresponding to the information bitstream as input and outputs detection results; it is used for signal detection to obtain the information bitstream. In this embodiment, the deep neural network model ultimately takes the overhead packet removal signal as input and outputs the model detection signal.

[0027] Step S2: Jointly train the unitary neural network model and the deep neural network model to obtain the trained unitary neural network model and the trained deep neural network model.

[0028] In this embodiment, step S2 involves jointly training the unitary neural network model and the deep neural network model to obtain a trained unitary neural network model and a trained deep neural network model, specifically including the following steps.

[0029] Step S21: Acquire modulation data and modulate the data Divided into Group, modulation data The length is Each group has a length of To obtain the modulated data set .

[0030] Step S22: Based on the modulation data set The first in Group modulation data, generating the first One-positive precoding matrix .

[0031] Step S23, based on the first One-positive precoding matrix Add constraints The weights of the unitary neural network model in the transmitter are initialized as a unitary matrix, where, Indicates transpose. Indicates size is The identity matrix.

[0032] Step S24: Perform singular value decomposition on the weights of the unitary neural network model to force the first... One-positive precoding matrix Satisfying the unitary constraint, we obtain the unitary matrix. unitary matrix .

[0033] In this embodiment, the following formula is used to represent the singular value decomposition of the weights of the unitary neural network model.

[0034] .

[0035] in, Indicates the first A single positive precoding matrix and It is a unitary matrix. This indicates the conjugate transpose.

[0036] Step S25: According to the unitary matrix and the unitary matrix Reconstruct the first One-positive precoding matrix The reconstructed precoding matrix is ​​obtained. .

[0037] In this embodiment, the following formula is used to represent the reconstruction of the first... One-positive precoding matrix .

[0038] .

[0039] in, This represents the reconstructed precoding matrix. and Let ∑ be a unitary matrix, and let ∑ be a rectangular diagonal matrix with non-negative real diagonals. This indicates the conjugate transpose.

[0040] Step S26: Based on the reconstructed precoding matrix To minimize the bit error rate, a deep neural network model at the receiving end is trained to obtain the deep neural network model. .

[0041] In this embodiment, the loss function aims to minimize the bit error rate. The loss function used during the training of the deep neural network model can be expressed as follows.

[0042] .

[0043] in, Represents the loss function. For the first One-positive precoding matrix and deep neural network models The parameters, Represents the reconstructed precoding matrix The output, Representing a deep neural network model The output.

[0044] Step S27: Proceed to step S22 until the modulation data of all groups have been trained to obtain the corresponding first step. One-positive precoding matrix and the deep neural network model The first group of modulation data was trained to obtain the first group of modulation data. One-positive precoding matrix and the neural network model Concatenate sequentially to generate a unitary precoding matrix. and deep neural network models Among them, the unitary precoding matrix The corresponding unitary neural network model is used as the trained unitary neural network model, and the deep neural network model. As a trained deep neural network model.

[0045] Step S3: Input the information bit stream to the transmitting end, and use the trained unitary neural network model to generate the precoded data according to the information bit stream, and send it to the receiving end in the form of the real OFDM signal.

[0046] In this embodiment, step S3 inputs the information bit stream to the transmitting end and uses the trained unitary neural network model to generate the precoded data based on the information bit stream, and sends it to the receiving end in the form of the real OFDM signal. Specifically, it includes the following steps.

[0047] Step S31: Perform source coding and interleaving on the input information bit stream to obtain the source-coded and interleaved information bit stream.

[0048] Step S32: Based on the information bit stream after source coding and interleaving, generate modulation data using a channel encoder and M-ary pulse amplitude modulation.

[0049] Step S33: Input the modulation data into the trained unitary neural network model to obtain precoded data.

[0050] Step S34: Perform equal-interval zeroing on the precoded data and superimpose overhead packets to obtain superimposed overhead packet signals.

[0051] In this embodiment, step S34 performs equal-interval zeroing on the precoded data and superimposes overhead packets to obtain the superimposed overhead packet signal, specifically including the following steps:

[0052] Step S341: Based on the precoded data, determine two diagonal matrices according to the length of the pilot sequence and the number of empty subcarriers; the two diagonal matrices are used to specify the position of the pilot sequence and the position of the empty subcarriers, respectively.

[0053] Step S342: Based on the two diagonal matrices, the positions of specified subcarriers in the precoded data are zeroed to obtain the zeroed precoded data.

[0054] Step S343: Perform overlay overhead packet processing on the precoded data after zeroing to obtain the overlay overhead packet signal.

[0055] Step S35: Perform IDHT (Inverse Discrete Hartley) transform on the superimposed overhead packet signal to obtain the time-domain signal.

[0056] Step S36: Perform cyclic prefix processing and up-conversion processing on the time-domain signal to obtain the real OFDM signal, and send the real OFDM signal to the receiving end.

[0057] Step S4: Receive the real OFDM signal using the receiving end, and perform signal detection on the real OFDM signal using the trained deep neural network model to obtain the information bit stream.

[0058] In this embodiment, step S4 uses the receiving end to receive the real OFDM signal and uses the trained deep neural network model to perform signal detection on the real OFDM signal to obtain the information bit stream, specifically including the following steps.

[0059] Step S41: Determine the actual received signal of the receiving end based on the actual OFDM signal.

[0060] Step S42: Perform synchronization processing, Doppler compensation processing, down-conversion processing, and cyclic prefix removal processing on the actual received signal to obtain the deprecated signal.

[0061] Step S43: Estimate and compensate the phase shift of the deprecated signal to obtain the estimated compensated signal.

[0062] Step S44: Perform DFT transformation on the estimated compensation signal to obtain the frequency domain received signal.

[0063] Step S45: Perform channel estimation on the frequency domain received signal, determine the time domain impulse response of the estimated channel, and further obtain the frequency response of the estimated channel.

[0064] Step S46: Based on the estimated channel frequency response, perform channel equalization on the frequency domain received signal and Transformation, resulting in channel equalization and The received signal is transformed.

[0065] Step S47: After channel equalization and The transformed received signal is processed to remove the overhead packet, resulting in a signal with the overhead packet removed.

[0066] Step S48: Input the removed overhead packet signal into the trained deep neural network model for signal detection to obtain the model detection signal.

[0067] Step S49: Demodulate and decode the model detection signal to obtain the information bit stream.

[0068] To make the technical solution of this embodiment clearer, the specific implementation process of this embodiment will be described in detail below by way of examples.

[0069] like Figure 2 As shown, this embodiment proposes a real-signal OFDM underwater acoustic communication method, which... =2048、 =512 / 256 Taking 64 as an example, the signal processing flow of this method specifically includes the following steps.

[0070] S1: Jointly train the unitary neural network model at the transmitting end and the deep neural network model at the receiving end, including the following steps.

[0071] (1) The length is modulation data Divided into Groups, each group is of length To obtain the modulated data set For the first Groups of modulated data are used for training to generate the first... One-positive precoding matrix and deep neural network models .

[0072] (2) Add constraints The weights of the unitary neural network model are initialized as a unitary matrix, where... Indicates transpose. Indicates size is The identity matrix. Singular value decomposition is performed on the weights of the unitary neural network model to force... Satisfy the unitary constraint, and then use the unitary matrix obtained from the singular value decomposition. unitary matrix To reconstruct the first One-positive precoding matrix The reconstructed precoding matrix is ​​obtained. The process of singular value decomposition is as follows: The reconstruction process is as follows: Here, ∑ is a rectangular diagonal matrix with non-negative real diagonals. This indicates the conjugate transpose.

[0073] (3) Train a deep neural network model to minimize the loss function of the bit error rate. loss function The expression is as follows.

[0074] .

[0075] in, For the first One-positive precoding matrix and deep neural network models The parameters, Represents the reconstructed precoding matrix The output, Representing a deep neural network model The output, and The relationship is as follows.

[0076] .

[0077] in, For DHT matrix, Indicates baseband noise. and for A diagonal matrix whose structure is similar to... and Same. An advanced stochastic gradient descent method with adaptive momentum estimation is used to update... The update process is as follows: ,in, and Let the learning rate and loss function represent the learning rate and loss function, respectively. about gradient, This indicates an update symbol. In this embodiment, Set to 1000, Set to 0.001.

[0078] (4) The unitary precoding matrix obtained from training all groups and deep neural network models Concatenate them sequentially to generate a unitary precoding matrix. and deep neural network models Among them, deep neural network models This refers to a trained deep neural network model.

[0079] S2: The transmitter inputs the information bit stream and sends a real OFDM signal. Specifically, it includes the following steps.

[0080] S201: The input information bitstream is source encoded and interleaved, and then passed through the channel encoder. The output length is The symbol. M-ary pulse amplitude modulation is used to generate modulated data. .

[0081] S202: Modulate data Insert a unitary neural network model to generate precoded data.

[0082] S203: Set the precoded data to zero at equally spaced positions and superimpose the overhead packets to obtain the superimposed overhead packet signal. Specifically, it includes the following steps.

[0083] (1) Determine two respectively diagonal matrix and To specify the positions of the pilot sequence and empty subcarriers, the diagonal elements are represented as follows.

[0084] .

[0085] .

[0086] in, and They are respectively in and Any fixed integer within the range, =4 / 8, =32.

[0087] (2) Set the specified subcarrier position in the precoded data to zero and add the overhead packet to obtain the superimposed overhead packet signal. , expressed as the following formula.

[0088] .

[0089] in, for The identity matrix, for The training pilot vector.

[0090] S204: For superimposed overhead packet signals Perform IDHT transformation to obtain the time-domain signal. .

[0091] S205: For time-domain signals Adding a length of The cyclic prefix is ​​then up-converted to obtain the real OFDM signal. In this embodiment, the roll-off factor is 0.65.

[0092] Although the transmitter sends a real OFDM signal to the receiver However, OFDM signals are usually real. During transmission, various interferences such as channel noise and signal attenuation often exist. Therefore, the signal actually received by the receiver is not the same as the actual OFDM signal transmitted by the transmitter. This also includes channel noise, etc. The signal ultimately received by the receiver in this embodiment is the actual received signal. The actual received signal The expression for is as follows.

[0093] .

[0094] in, Indicates the actual received signal. Representing a path The time-varying decay factor on, Indicates a real OFDM signal. Indicates multipath delay components, Indicates the Doppler scaling factor. Indicates channel noise. Indicates time, L Indicates channel delay truncation. Indicates the carrier frequency. Indicates signal power.

[0095] S3: The receiver receives the actual received signal. and the actual received signal Perform the following signal processing procedure.

[0096] S301: Regarding the actual received signal Synchronization, Doppler compensation, and down-conversion are performed, and the cyclic prefix is ​​removed to obtain the deprecated signal. .

[0097] S302: Deprecated signal Perform phase shift The estimation and compensation are performed to obtain the estimated compensation signal. .

[0098] S303: Estimating the compensation signal Perform DFT transform to obtain the frequency domain received signal and the received signal in the frequency domain Channel estimation is performed to obtain the time-domain impulse response of the estimated channel. The estimation process is expressed as the following formula.

[0099] .

[0100] in, Represents the DFT matrix, For the DFT matrix Submatrix, For frequency domain signal reception, To be assigned to a position The receiving pilot signal at the location.

[0101] S304: Receive frequency domain signal Through frequency domain equalizer and conduct Transformation, resulting in channel equalization and Transformed received signal A minimum mean square error frequency domain equalizer is used, with equalizer coefficients of... ,in Through the conduct N The channel frequency response obtained by point fast Fourier transform. This represents the variance of the channel noise. After channel equalization and... Transformed received signal It is expressed as the following formula.

[0102] .

[0103] in, express The transformation matrix combines the DHT and DFT transformations.

[0104] S305: From after channel equalization and Transformed received signal The overhead packet removal signal is obtained by removing the overhead packet. , represented as Then use a deep neural network model. Detect the removal overhead packet signal The model detection signal is obtained.

[0105] S306: Demodulate and decode the detection signal of this model to obtain the real OFDM signal transmitted by the transmitter. It contains a stream of information bits.

[0106] The real-signal OFDM underwater acoustic communication method proposed in this embodiment is specifically applied in an underwater acoustic communication scenario. In this scenario, the real-signal OFDM underwater acoustic communication system is a single-input single-output system. In this system, the transmitter is the transmitting end, and the receiver is the receiving end. The transmitter and receiver are placed at a water depth of 4m, and the distance between them is 1km. The channel delay is truncated as follows. =128, channel coding rate is 0.5.

[0107] In this embodiment, the number of transmitted symbols is 104, and the number of subcarriers is [missing information]. =2048, where the number of pilot subcarriers is set as follows: =512 and =256, the number of empty subcarriers is =64, the number of guard interval subcarriers is =512, using M-PAM (M-element pulse amplitude modulation). The method proposed in this embodiment is compared with traditional UWAC-OFDM (Orthogonal Frequency Division Multiplexing Water Acoustic Communication), real-signal UWAC-OFDM, and traditional ST-based OFDM to verify the effectiveness of the proposed method. Under ideal CFO conditions, the bit error rates of the above methods in an additive white Gaussian noise channel are as follows: Figure 3 As shown, the bit error rate under a perfectly known underwater acoustic channel is as follows: Figure 4 As shown. Overall, compared with the traditional ST-based method, the method proposed in this embodiment overcomes the problem of data detection performance degradation caused by precoding data loss in the traditional ST-based method, and therefore has better bit error rate performance than the traditional ST-based method. In the case of a perfectly known UWAC channel, when When CFO = 8, the method proposed in this embodiment achieves diversity by using a unitary neural network model at the transmitter, resulting in better bit error rate performance than traditional OFDM. Under UWAC channel estimation conditions, the bit error rates of each method at different CFO values ​​are as follows: Figure 5 As shown. When the CFO value is 0.05, traditional DFT-OFDM is quite sensitive to the CFO value, while the method proposed in this embodiment still provides acceptable performance. When the CFO value is 0.15, the bit error rate performance of all methods deteriorates due to the need for CFO compensation. The bit error rates of each method in the cases with and without CFO compensation are shown below. Figure 6 As shown in the figure, the method proposed in this embodiment not only avoids the spectral efficiency loss caused by the additional transmission of empty subcarriers in traditional real-signal OFDM underwater acoustic communication systems, but also exhibits stronger bit error rate performance.

[0108] This embodiment uses a unitary neural network model to generate precoded data at the transmitter, eliminating the need for an underwater acoustic channel during training. This solves the problem of high computational complexity in previous methods, and the trained model can be used in any simulation and experimental environment, demonstrating universal applicability and a wider range of applications. At the transmitter, after eliminating 25% of the symbol subcarriers, pilot sequences are used to replace them, overcoming interference caused by superimposed pilot sequences. Carrier frequency offset estimation and compensation are performed using empty subcarriers superimposed on the transmitted data to track the CFO effect without sacrificing additional subcarriers, thus improving the transmission rate of the real-signal OFDM underwater acoustic communication system. By deploying a deep neural network model at the receiver for data detection, performance degradation in data detection is avoided.

[0109] This embodiment proposes a real-signal OFDM underwater acoustic communication method based on deep learning and superposition training. This method replaces the discrete Fourier transform with a discrete Hartley transform. At the transmitter, a unitary neural network model is used to generate precoded data, and after eliminating 25% of the symbol subcarriers, these subcarriers are replaced with pilot sequences, thus completely overcoming the interference caused by the additional pilot sequences. At the receiver, another deep neural network model is deployed to detect the transmitted information bitstream. The two models are jointly trained to avoid data detection performance degradation, thereby enhancing data detection performance. Furthermore, since the precoded data is generated using a unitary neural network model, maximum diversity gain can be obtained in multipath channels. This eliminates the need to include the underwater acoustic channel during training, allowing the trained model to be used in any simulation and experimental environment, meeting the application requirements of various simulation scenarios and experimental environments.

[0110] Based on the same inventive concept, this application also provides a real-signal OFDM underwater acoustic communication system for implementing the real-signal OFDM underwater acoustic communication method described above. The solution provided by this system is similar to the implementation described in the above method; therefore, the specific limitations in the embodiments of the real-signal OFDM underwater acoustic communication system provided below can be found in the limitations of the real-signal OFDM underwater acoustic communication method described above, and will not be repeated here.

[0111] In one feasible embodiment, a real-signal OFDM underwater acoustic communication system is provided. This system includes a transmitter and a receiver. The transmitter inputs an information bitstream and then sends a real OFDM signal to the receiver. The receiver obtains the information bitstream based on the real OFDM signal. When underwater acoustic communication occurs between the transmitter and receiver, the described real-signal OFDM underwater acoustic communication method is executed.

[0112] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0113] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A real signal OFDM underwater acoustic communication method, characterized in that, The real-signal OFDM underwater acoustic communication method is used for underwater acoustic communication between a transmitting end and a receiving end, the transmitting end is used for transmitting a real OFDM signal to the receiving end after inputting an information bit stream, and the receiving end is used for obtaining the information bit stream according to the real OFDM signal; The real-signal OFDM underwater acoustic communication method comprises: The unitary neural network model is deployed at the transmitting end, and the deep neural network model is deployed at the receiving end; the input of the unitary neural network model is modulation data corresponding to the information bit stream, and the output is precoding data, the unitary neural network model is used for generating the precoding data according to the input information bit stream; the input of the deep neural network model is detection data corresponding to the information bit stream, and the output is a detection result corresponding to the information bit stream, the deep neural network model is used for detecting the information bit stream; The unitary neural network model and the deep neural network model are jointly trained to obtain a trained unitary neural network model and a trained deep neural network model; The information bit stream is input to the transmitting end, and the trained unitary neural network model is used to generate the precoding data according to the information bit stream, and the real OFDM signal is transmitted to the receiving end; The real OFDM signal is received by the receiving end, and the trained deep neural network model is used for signal detection on the real OFDM signal to obtain the information bit stream; The unitary neural network model and the deep neural network model are jointly trained to obtain a trained unitary neural network model and a trained deep neural network model, specifically comprising: Acquiring modulation data , and dividing the modulation data into groups, the length of the modulation data in each group being , the length of each group being , to obtain a modulation data set ; generate a first unitary precoding matrix based on a first group of modulation data in the set of modulation data ​​​​ based on the first unitary precoding matrix , a constraint is added to initialize the weights of the unitary neural network model in the transmitting end to a unitary matrix, wherein denotes transposition, denotes a unit matrix of size . performing singular value decomposition on weights of the unitary neural network model to enforce the first unitary precoding matrix satisfying the unitary constraint to obtain a unitary matrix and the unitary matrix ​ According to the unitary matrix and the unitary matrix reconstruct the first unitary precoding matrix , to obtain a reconstructed precoding matrix ; based on the reconstructed precoding matrix The deep neural network model of the receiving end is trained to minimize the bit error rate, and a deep neural network model is obtained ; Jump to step "Based on the modulation data set" The first in Group modulation data, generating the first One-positive precoding matrix "until the modulation data of all groups have been trained to obtain the corresponding first..." One-positive precoding matrix and the deep neural network model The first group of modulation data was trained to obtain the first group of modulation data. One-positive precoding matrix and the neural network model Concatenate sequentially to generate a unitary precoding matrix. and deep neural network models The unitary precoding matrix The corresponding unitary neural network model serves as the trained unitary neural network model, and the deep neural network model This serves as the trained deep neural network model.

2. The real signal OFDM underwater acoustic communication method of claim 1, wherein, The singular value decomposition of the weight of the unitary neural network model is represented by the following formula: ； wherein denotes the th unitary precoding matrix, and is a unitary matrix, denotes the conjugate transpose.

3. The real signal OFDM underwater acoustic communication method of claim 1, wherein, The reconstructed first unitary precoding matrix is represented by the following equation: :​ ； wherein denotes the reconstructed precoding matrix, and is an orthogonal matrix, and ∑ is a rectangular diagonal matrix with non-negative real diagonal entries, denotes the conjugate transpose.

4. The real signal OFDM underwater acoustic communication method of claim 1, wherein, The loss function used in the training of the deep neural network model is represented by: ； wherein, represents a loss function, is the th unitary precoding matrix and the parameters of the deep neural network model , represents the output of the reconstructed precoding matrix , represents the output of the deep neural network model .

5. The real signal OFDM underwater acoustic communication method of claim 1, wherein, The information bit stream is input to the transmitting end, and the trained unitary neural network model is used to generate the precoding data according to the information bit stream, and the real OFDM signal is transmitted to the receiving end, specifically comprising: The input information bit stream is source encoded and interleaved to obtain a source encoded and interleaved information bit stream; The modulation data is generated by using a channel encoder and M-pulse amplitude modulation according to the source encoded and interleaved information bit stream; The modulation data is input into the trained unitary neural network model to obtain precoding data; The precoding data is processed by equal-interval position zeroing and overhead packet superposition to obtain an overhead packet superposition signal; The overhead packet superposition signal is subjected to IDHT transformation to obtain a time domain signal; The time domain signal is subjected to cyclic prefix processing and up-conversion processing to obtain the real OFDM signal, and the real OFDM signal is transmitted to the receiving end.

6. The real signal OFDM underwater acoustic communication method of claim 5, wherein, The precoding data is processed by equal-interval position zeroing and overhead packet superposition to obtain an overhead packet superposition signal, specifically comprising: Two diagonal matrices are determined according to the length of the pilot sequence and the number of null subcarriers based on the precoding data; the two diagonal matrices are respectively used to specify the positions of the pilot sequence and the positions of the null subcarriers; The positions of the subcarriers specified in the precoding data are zeroed according to the two diagonal matrices, to obtain zeroed precoding data; The zeroed precoding data is subjected to overhead packet superposition processing to obtain the overhead packet superimposed signal.

7. The real signal OFDM underwater acoustic communication method of claim 1, wherein, The real OFDM signal is received by the receiving end, and the real OFDM signal is subjected to signal detection by using the trained deep neural network model to obtain the information bit stream, specifically including: The actual receiving signal of the receiving end is determined according to the real OFDM signal; The actual receiving signal is subjected to synchronization processing, Doppler compensation processing, down-conversion processing and cyclic prefix removal processing to obtain a de-prefix signal; The de-prefix signal is subjected to phase offset estimation and compensation processing to obtain an estimated and compensated signal; The estimated and compensated signal is subjected to DFT transformation to obtain a frequency domain receiving signal; The frequency domain receiving signal is subjected to channel estimation to determine the time domain impulse response of the estimated channel, to obtain the frequency response of the estimated channel; channel equalizing the frequency domain received signal according to the estimated channel frequency response, and transforming, to obtain a channel equalized and transformed received signal; performing overhead removal on the channel-equalized and transformed received signal to obtain an overhead-removed signal; The overhead packet removed signal is input into the trained deep neural network model for signal detection to obtain a model detection signal; The model detection signal is subjected to demodulation and decoding to obtain the information bit stream.

8. The real signal OFDM underwater acoustic communication method of claim 7, wherein, The actual receiving signal is represented by the following formula: ； in, Indicates the actual received signal. Representing a path The time-varying decay factor on, Indicates a real OFDM signal. Indicates multipath delay components, Indicates the Doppler scaling factor. Indicates channel noise. Indicates time, L Indicates channel delay truncation. Indicates the carrier frequency. Indicates signal power.

9. A real signal OFDM underwater acoustic communication system characterized by, The real signal OFDM underwater acoustic communication system includes a transmitting end and a receiving end, the transmitting end is used to input an information bit stream and then send a real OFDM signal to the receiving end, and the receiving end is used to obtain the information bit stream according to the real OFDM signal; when the transmitting end and the receiving end perform underwater acoustic communication, the real signal OFDM underwater acoustic communication method in any one of claims 1-8 is executed.

Images