Intelligent Adaptive Underwater Acoustic Communication System

Abstract

This invention relates to the field of self-learning underwater acoustic communication technology, and provides an intelligent adaptive underwater acoustic communication system. The multi-standard transmitter includes a data transmission unit for transmitting communication data, a standard switching unit for switching the modulation mode of the transmitted signal, and a transmitted signal modulation unit for modulating the signal. The intelligent multi-standard receiver includes a synchronization unit, a Doppler estimation and compensation unit, an intelligent modulation mode identification unit, a data processing unit, an intelligent signal decoding unit, and a digital demodulation unit. The channel information identification unit identifies the signal-to-noise ratio (SNR) and maximum channel delay of the underwater acoustic channel and transmits them to the standard switching unit. The standard switching unit adjusts and switches the transmitted signal modulation mode according to the SNR and maximum channel delay. The intelligent modulation mode identification unit uses a lightweight neural network to intelligently identify the signal modulation mode of the underwater acoustic channel signal. The intelligent signal decoding unit uses a deep learning neural network to intelligently decode the underwater acoustic communication signal to recover the communication data.

Description

Technical Field

[0001] This invention relates to the field of self-learning underwater acoustic communication technology, and in particular to an intelligent adaptive multi-mode underwater acoustic communication system. Background Technology

[0002] Given the limitations of optical and radio frequency (RF) communications in underwater applications, underwater acoustic communication is currently the only solution for long-distance underwater wireless communication and a crucial component in marine information acquisition, transmission, and processing. With the increasing demand for long-term real-time and large-scale dynamic marine observation, the need for high-quality underwater acoustic communication technology is becoming increasingly urgent. However, the underwater acoustic channel presents far more challenging transmission conditions than terrestrial RF wireless channels, exhibiting large and time-varying transmission delays (underwater acoustic propagation speed is five orders of magnitude slower than RF propagation speed), complex environmental noise (additive and multiplicative noise), energy fading related to both distance and frequency, strong multipath effects, and strong Doppler frequency spread. Especially in complex application scenarios such as long-term real-time marine information observation (e.g., real-time observation of engineering geological parameters during natural gas hydrate extraction), complex underwater network observation (e.g., coral reef ecosystem monitoring and protection), and large-scale dynamic underwater network observation (e.g., underwater target detection and seabed resource exploration based on dynamic networks), the time-varying characteristics and their impact on the underwater acoustic channel become even more pronounced, posing a greater challenge to the realization of high-quality underwater acoustic communication technology.

[0003] In existing technologies, underwater acoustic communication systems use a single carrier mode for underwater acoustic signals, such as FSK, PSK, and OFDM. The system performance is limited by the worst underwater acoustic channels in complex application scenarios, which restricts the application of underwater acoustic communication systems in complex application scenarios such as long-term real-time underwater observation, underwater network observation in complex environments, and large-scale dynamic networking. It is difficult to meet the requirements of high-quality underwater acoustic communication that makes full use of channel bandwidth, so it is necessary to develop channel-adaptive underwater acoustic communication technology.

[0004] Channel-adaptive underwater acoustic communication systems can adjust their parameters according to the underwater acoustic channel in complex application scenarios. Based on the signal modulation method, channel-adaptive underwater acoustic communication systems are mainly divided into two categories: single-mode multi-parameter adaptive underwater acoustic communication systems using a single signal modulation method, and multi-mode adaptive underwater acoustic communication systems using multiple signal modulation methods. Single-mode multi-parameter adaptive underwater acoustic communication systems mainly adapt to changes in the underwater acoustic channel through baseband adjustment coding methods and digital modulation methods. However, limited by the single signal modulation method, the channel adaptability range of single-mode multi-parameter adaptive underwater acoustic communication systems is limited. Multi-mode adaptive underwater acoustic communication systems can use different signal modulation methods to adapt to changes in the underwater acoustic channel, such as single-carrier, multi-carrier, spread spectrum, and combinations thereof. This allows for better adaptation to changes in the underwater acoustic channel and full utilization of the underwater acoustic channel bandwidth. However, existing multi-mode adaptive underwater acoustic communication systems use receivers based on traditional signal processing techniques for channel estimation and equalization. Limited by the performance of channel estimation and equalization, the performance of multi-mode adaptive underwater acoustic communication systems based on traditional signal processing techniques is constrained.

[0005] To address the need for improved performance in underwater acoustic communication systems with complex time-varying underwater acoustic channels, intelligent underwater acoustic communication technologies based on machine learning and deep learning networks have emerged. However, existing intelligent underwater acoustic communication technologies are based on a single signal modulation method, which still suffers from the problem of limited channel adaptability due to the single signal modulation method. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent adaptive underwater acoustic communication system. In response to the high-quality underwater acoustic communication application requirements of complex time-varying underwater acoustic channels in complex underwater application scenarios, and taking into account the limitations of existing underwater acoustic communication technologies, this invention proposes an intelligent adaptive multi-mode underwater acoustic communication system based on deep learning.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0008] An intelligent adaptive underwater acoustic communication system, comprising:

[0009] Multi-standard transmitter: includes a data transmission unit for transmitting communication data, a standard switching unit for switching the modulation mode of the transmitted signal, and a transmitted signal modulation unit for modulating the signal; the data transmission unit, the standard switching unit, and the transmitted signal modulation unit are connected in sequence; the output of the multi-standard transmitter is connected to an underwater acoustic channel;

[0010] The intelligent multi-standard receiver includes a synchronization unit, a Doppler estimation and compensation unit, an intelligent modulation scheme identification unit, a data processing unit, an intelligent signal decoding unit, and a digital demodulation unit. The synchronization unit receives underwater acoustic communication signals passing through the underwater acoustic channel and performs data frame synchronization processing on the signals. The Doppler estimation and compensation unit receives output data from the synchronization unit and estimates and compensates for the Doppler frequency shift of the underwater acoustic channel. The intelligent modulation scheme identification unit is connected to the signal output of the Doppler estimation and compensation unit and is used to intelligently identify the modulation scheme of the underwater acoustic communication signal based on the bandpass underwater acoustic signal. The data processing unit is connected to the signal output of the intelligent modulation scheme identification unit. The intelligent signal decoding unit is connected to the signal output of both the data processing unit and the intelligent modulation scheme identification unit. It is used to intelligently decode the underwater acoustic communication frequency domain signal after the fast Fourier transform, based on the identified modulation scheme, to obtain communication data. The digital demodulation unit is connected to the signal output of the intelligent signal decoding unit and modulates the obtained communication data for the data receiving unit.

[0011] Channel information identification unit: Its input end is connected to the signal output end of the synchronization unit of the intelligent multi-mode receiver, and its output end is connected to the mode switching unit of the multi-mode transmitter.

[0012] The channel information identification unit is used to identify the signal-to-noise ratio and maximum channel delay of the underwater acoustic channel and transmit them to the mode switching unit. The mode switching unit adjusts and switches the modulation mode of the transmitted signal according to the signal-to-noise ratio and maximum channel delay. The intelligent modulation mode identification unit uses a lightweight neural network to intelligently identify the signal modulation mode of the underwater acoustic passband signal. The intelligent signal interpretation unit uses a deep learning neural network to intelligently interpret the underwater acoustic communication signal to recover the communication data.

[0013] In some embodiments of the present invention, the transmit signal modulation unit includes:

[0014] Digital modulation unit: connected to the output of the system switching unit;

[0015] The spread spectrum unit includes a first spread spectrum branch and a second spread spectrum branch. The first spread spectrum branch is provided with a switching assembly and a spread spectrum module. The output terminal of the digital modulation unit is connected to the first spread spectrum branch and the second spread spectrum branch respectively. The first spread spectrum branch and the second spread spectrum branch are connected and converge to form the output terminal of the spread spectrum unit.

[0016] The inverse fast Fourier transform unit includes a first transform branch and a second transform branch. The first transform branch is equipped with a switching component and an inverse Fourier transform module. The output terminal of the spread spectrum unit is connected to the first transform branch and the second transform branch respectively. The output terminals of the first transform branch and the second transform branch converge to form the output terminal of the inverse fast Fourier transform unit.

[0017] Framing unit: Used to add pilot sequences and cyclic prefixes to the output data of the inverse fast Fourier transform unit;

[0018] Frequency conversion modulation unit: used to perform frequency conversion processing on the output data of the framing unit;

[0019] The signal output terminal of the frequency conversion modulation unit outputs a signal to the underwater acoustic channel.

[0020] In some embodiments of the present invention, the multi-mode transmitter is configured to transmit one of the following modulation signal types: single-carrier cyclic prefix signal, single-carrier time-domain spread spectrum signal, orthogonal frequency division multiplexing signal, and multi-carrier frequency-domain spread spectrum signal.

[0021] When the multi-mode transmitter is configured to transmit a single-carrier cyclic prefix signal, the second spread spectrum branch is activated, and the second conversion branch is activated.

[0022] When the multi-mode transmitter is configured to transmit a single-carrier time-domain spread spectrum signal, the first spread spectrum branch is turned on and the second conversion branch is turned on.

[0023] When the multi-mode transmitter is configured to transmit orthogonal frequency division multiplexing signals, the second spread spectrum branch is turned on and the first conversion branch is turned on.

[0024] When the multi-mode transmitter is configured to transmit a multi-carrier frequency domain spread spectrum signal, the first spread spectrum branch is activated and the first conversion branch is activated.

[0025] In some embodiments of the present invention, the transmission signal of the multi-standard transmitter is defined as s(n), n = 1, 2, 3, ..., Q. S , n represents the sequence number of the transmitted signal, Q S The number of transmitted signals s(n):

[0026] The single-carrier cyclic prefix signal is represented as:

[0027] s(n) = x[k], k = 1, 2, 3, ..., Q S , n = k, where x[k] represents the digitally modulated transmitted data and k represents the sequence number of the transmitted data;

[0028] The single-carrier time-domain spread spectrum signal is represented as follows:

[0029] s(n) = x[k]c[i], i = 1, 2, 3, ..., N p n = (k-1)N p +i, where c[i] is the spreading sequence, i is the index of the spreading sequence, and N p It is the length of the spreading sequence;

[0030] The orthogonal frequency division multiplexed signal is represented as follows:

[0031] l = 1, 2, 3, ..., Q s n = l, k = 1, 2, 3, ..., NQ S Where N is fast

[0032] The number of subcarriers in the Fast Fourier Transform;

[0033] The multi-carrier frequency domain spread spectrum signal is represented as follows:

[0034] l = 1, 2, 3, ..., Q s n = l, where y(p) is the spread spectrum data sequence, y(p) = x[k]c[i]. i = 1, 2, 3, ..., N p p = (k-1)N p +i.

[0035] In some embodiments of the present invention, the data processing unit of the intelligent multi-standard receiver includes:

[0036] Frequency conversion demodulation unit: connected to the output of the intelligent modulation mode identification unit, used to set the frequency conversion strategy according to the modulation mode identified by the intelligent modulation mode identification unit, and to perform frequency conversion demodulation processing on the underwater acoustic communication signal;

[0037] Deframe unit: Connected to the output of the frequency conversion demodulation unit, it is used to deframe the underwater acoustic signal after frequency conversion demodulation based on the modulation method identified by the intelligent modulation method identification unit.

[0038] Fast Fourier Transform Unit: Connected to the output of the deframe unit, it is used to perform a Fast Fourier Transform on the deframed underwater acoustic communication signal according to the modulation mode identified by the intelligent modulation mode identification unit, to obtain the frequency domain signal of the underwater acoustic communication signal, including the pilot sequence frequency domain signal and the data sequence frequency domain signal of the underwater acoustic communication signal.

[0039] The intelligent signal interpretation unit is connected to the output of the fast Fourier transform unit and the output of the intelligent modulation mode identification unit. It is used to intelligently interpret the underwater acoustic communication frequency domain signal after the fast Fourier transform by combining the identified modulation mode to obtain communication data.

[0040] In some embodiments of the present invention, the intelligent modulation scheme identification unit includes:

[0041] Signal conversion module: used to convert underwater acoustic communication signal time series signals into time-frequency image signals;

[0042] Signal recognition neural network: receives the time-frequency image signal and identifies the signal modulation mode based on the time-frequency image signal;

[0043] The signal recognition neural network mentioned above uses lightweight neural networks, including ShuffleNet, EfficientNet, and MobileNet.

[0044] In some embodiments of the present invention, the intelligent signal interpretation unit includes a deep learning network, which is a data-driven optimized DNN neural network, a data-driven LSTM neural network, or a model-driven ModelDriven-Net neural network. The input of the deep learning network is connected to the output of the fast Fourier transform unit to perform intelligent signal interpretation on the underwater acoustic communication frequency domain signal after the fast Fourier transform, thereby obtaining communication data.

[0045] The data-driven optimized DNN is configured to set the number of available network layers and network nodes according to the identified modulation scheme: if the identified signal modulation scheme is a single-carrier cyclic prefix signal or an orthogonal frequency division multiplexing signal, the DNN is configured to include a first number of network layers and a second number of network nodes; if the identified signal modulation scheme is a single-carrier time-domain spread spectrum signal or a multi-carrier frequency-domain spread spectrum signal, the DNN is configured to include a third number of network layers and a fourth number of network nodes; the third number is less than the first number, and the fourth number is less than the second number.

[0046] The data-driven LSTM neural network is configured as a two-layer LSTM network.

[0047] The ModelDriven-Net neural network is a deep learning network that combines intelligent signal interpretation networks with traditional channel estimation and channel equalization algorithms, including a channel estimation subnet and a channel equalization subnet.

[0048] The channel estimation subnet includes:

[0049] Least squares module: Obtains the pilot sequence signal y output by the Fast Fourier Transform unit of the intelligent multi-standard receiver. p And pilot sequence signals x transmitted by multi-standard transmitters p The channel impulse response is generated after least squares processing. Channel impulse response It is divided into real information and imaginary information;

[0050] Channel estimation neural network: input channel impulse response The real and imaginary parts of the signal are used to estimate the channel impulse response.

[0051] The channel equalization subnet includes:

[0052] Minimum mean square error module: Connects to the output of the channel estimation neural network to obtain an estimate of the output channel impulse response. And input the data sequence signal y output from the Fast Fourier Transform unit of the intelligent multi-standard receiver. D After minimum variance calculation, the minimum mean square error signal x is output. MMSE The output x MMSE The signal is divided into real part information and imaginary part information;

[0053] Signal equalization neural network: For single-carrier cyclic prefix signals and single-carrier time-domain spread spectrum signals, obtain the output x. MMSE The real and imaginary parts of the signal are used to output a data equalization signal. Signal, The data equalization signal is processed by hard decision to obtain data. For orthogonal frequency division multiplexing (OFDM) signals and multi-carrier frequency domain spread spectrum (MFSS) signals, the neural network simultaneously obtains estimates of the output channel impulse response. The data sequence signal y output by the Fast Fourier Transform unit of the intelligent multi-standard receiver D The minimum mean square error signal x MMSE Output data equalization signal Signal, The data equalization signal is processed by hard decision to obtain data.

[0054] The intelligent adaptive multi-mode underwater acoustic communication system provided by this invention has the following advantages: It is a deep learning-based intelligent adaptive multi-mode underwater acoustic communication system. The channel-adaptive multi-mode transmitter supports four signal modulation methods: single-carrier cyclic prefix (SC-CP), single-carrier time-domain spread spectrum (SC-TDSS), orthogonal frequency division multiplexing (OFDM), and multi-carrier frequency-domain spread spectrum (MC-FDSS). It adaptively switches the signal modulation method based on the channel signal-to-noise ratio and the maximum channel delay. The intelligent multi-mode receiver includes modules for intelligent modulation method identification based on a lightweight network, intelligent signal interpretation based on a deep learning network, synchronization, Doppler estimation and compensation, frequency conversion demodulation, deframe, and FFT. This achieves intelligent identification of underwater acoustic communication signal modulation methods and intelligent channel estimation and equalization processing for the four underwater acoustic communication signal modulation methods. It can adaptively adapt suitable underwater acoustic communication signal modulation methods for time-varying underwater acoustic channels in complex application scenarios. It also realizes intelligent identification and intelligent signal interpretation of various underwater acoustic communication modulation methods. Compared with traditional channel estimation and equalization signal processing methods, it improves the performance of signal interpretation and processing. It also solves the limitations of existing single-mode single-parameter underwater acoustic communication, single-mode multi-parameter underwater acoustic communication, multi-mode underwater acoustic communication based on traditional signal processing, and intelligent single-mode underwater acoustic communication when facing time-varying underwater acoustic channels in complex application scenarios. Attached Figure Description

[0055] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0056] Figure 1 This invention relates to an intelligent adaptive underwater acoustic communication system;

[0057] Figure 2 This is a data frame structure diagram of the intelligent adaptive underwater acoustic communication system of the present invention;

[0058] Figure 3a Time-frequency analysis diagram of a single-carrier cyclic prefix signal;

[0059] Figure 3b This is a time-frequency analysis diagram of a single-carrier time-domain spread spectrum signal.

[0060] Figure 3c This is a time-frequency analysis diagram of an orthogonal frequency division multiplexed signal;

[0061] Figure 3d Time-frequency analysis diagram of multi-carrier frequency domain spread spectrum signal;

[0062] Figure 4This is a schematic diagram of the channel estimation subnetwork structure;

[0063] Figure 5 Flowchart for training the channel estimation subnetwork;

[0064] Figure 6 A schematic diagram of one implementation scheme for a channel equalization subnet;

[0065] Figure 7 A schematic diagram of another implementation scheme for the channel equalization subnet;

[0066] Figure 8 A comparison chart showing the accuracy of different neural network models in recognizing signal modulation methods;

[0067] Figure 9a This is a graph showing the bit error rate of single-carrier cyclic prefix modulation.

[0068] Figure 9b Bit error rate diagram for single-carrier time-domain spread spectrum modulation;

[0069] Figure 9c This is a graph showing the bit error rate of orthogonal frequency division multiplexing (OFDM) modulation.

[0070] Figure 9d Bit error rate diagram for multi-carrier frequency domain spread spectrum modulation;

[0071] Figure 10 A comparison chart of the performance and bit error rate of intelligent signal decoding under CH1 channel conditions;

[0072] Figure 11 A comparison chart of the performance and bit error rate of intelligent signal decoding under CH2 channel conditions;

[0073] Figure 12 A comparison chart of the performance and bit error rate of intelligent signal decoding under CH3 channel conditions. Detailed Implementation

[0074] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.

[0075] The prefixes such as "first" and "second" used in this application embodiment are merely for distinguishing different descriptive objects and do not limit the position, order, priority, quantity, or content of the described objects. The use of ordinal numbers and other prefixes used to distinguish descriptive objects in this application embodiment does not constitute a limitation on the described objects. The description of the described objects is given in the claims or the context of the embodiments, and should not constitute unnecessary restrictions due to the use of such prefixes. Furthermore, in the description of this embodiment, unless otherwise stated, "multiple" means two or more.

[0076] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B; the term "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone.

[0077] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0078] According to existing technologies, underwater acoustic communication technologies mainly include single-mode single-parameter underwater acoustic communication, single-mode multi-parameter underwater acoustic communication, multi-mode underwater acoustic communication, and intelligent single-mode underwater acoustic communication.

[0079] In terms of single-mode, single-parameter underwater acoustic communication, the performance of single-mode, single-parameter underwater acoustic communication systems (such as FSK, PSK, OFDM, etc.) that have been developed for a long time is often limited by the worst conditions of underwater acoustic channels in complex application scenarios. This limits their application in long-term real-time underwater observation, underwater network observation in complex environments, and large-scale dynamic networking, making it difficult to meet the demand for high-quality underwater acoustic communication that makes full use of channel bandwidth.

[0080] In terms of single-mode multi-parameter underwater acoustic communication, the main methods used are baseband adjustment coding and digital modulation to adapt to changes in the underwater acoustic channel. However, due to the limitation of a single carrier modulation method, the channel adaptability of single-mode multi-parameter adaptive underwater acoustic communication systems is limited.

[0081] In the field of multi-mode underwater acoustic communication, the receiver adopts channel estimation and channel equalization methods based on traditional signal processing technology. Due to the limitations of channel estimation and channel equalization performance, multi-mode adaptive underwater acoustic communication systems based on traditional signal processing technology are difficult to effectively adapt to complex time-varying underwater acoustic channels.

[0082] Intelligent single-mode underwater acoustic communication is currently mainly developed based on OFDM carrier modulation. However, it is also limited by the single carrier modulation method, which restricts the channel adaptability of intelligent single-mode underwater acoustic communication systems.

[0083] To address the above problems, this invention proposes an intelligent adaptive underwater acoustic communication system.

[0084] refer to Figure 1 The present invention provides an intelligent adaptive underwater acoustic communication system, the structure of which includes: a multi-mode transmitter, a channel information identification unit, and an intelligent multi-mode receiver.

[0085] Multi-standard transmitter: Includes a data transmission unit for transmitting communication data, a standard switching unit for switching the modulation mode of the transmitted signal, and a transmitted signal modulation unit for modulating the signal; the data transmission unit, standard switching unit, and transmitted signal modulation unit are connected sequentially. The standard switching unit acts as the receiving end of the transmitted data, outputting modulated signals of different standards after modulation of the transmitted signal. Its data frame format is referenced. Figure 2 .

[0086] Channel information identification unit: Its input end is connected to the signal output end of the synchronization unit of the intelligent multi-mode receiver, and its output end is connected to the mode switching unit of the multi-mode transmitter.

[0087] The channel information identification unit of the intelligent adaptive underwater acoustic communication system is used to identify the signal-to-noise ratio and maximum channel delay of the underwater acoustic channel and transmit them to the mode switching unit. The mode switching unit adjusts the signal modulation mode of the transmitted signal according to the signal-to-noise ratio and maximum channel delay.

[0088] To adapt to the time-varying underwater acoustic channels in different underwater acoustic communication scenarios, in some embodiments of this invention, the multi-mode transmitter is configured to transmit one of the following modulation signal types: single-carrier cyclic prefix signal (SC-CP), single-carrier time-domain spread spectrum signal (SC-TDSS), orthogonal frequency division multiplexing signal (OFDM), and multi-carrier frequency-domain spread spectrum signal (MC-FDSS). Table 1 provides a detailed design description of the four signal modulation methods.

[0089] Table 1 Design of Four Signal Modulation Methods

[0090] Signal modulation method Digital modulation Spread Spectrum Spread code length Number of signal subcarriers SC-CP QPSK no -- 1 SC-TDSS QPSK yes 16 1 OFDM QPSK no -- 1024 MC-FDSS QPSK yes 16 1024

[0091] In some embodiments of the present invention, the output modulation mode of the underwater acoustic signal is switched by changing the configuration of the transmitting signal modulation unit. To achieve the function of outputting multiple signal modulation modes, the transmitting signal modulation unit includes:

[0092] Digital modulation unit: connected to the output of the system switching unit;

[0093] The spread spectrum unit includes a first spread spectrum branch and a second spread spectrum branch. The first spread spectrum branch is equipped with a switching component and a spread spectrum module. The output terminal of the digital modulation unit is connected to the first spread spectrum branch and the second spread spectrum branch respectively. The first spread spectrum branch and the second spread spectrum branch are connected and converge to form the output terminal of the spread spectrum unit. It should be understood that the output signal of the digital modulation unit can be selected to undergo spread spectrum processing before being sent to the next stage.

[0094] The inverse fast Fourier transform unit includes a first transform branch and a second transform branch. The first transform branch is equipped with a switching component and an inverse Fourier transform module. The output of the spread spectrum unit is connected to the first transform branch and the second transform branch respectively. The outputs of the first transform branch and the second transform branch converge to form the output of the inverse fast Fourier transform unit. It should be understood that the output signal of the spread spectrum unit can be selected to undergo an inverse Fourier transform before being sent to the next stage.

[0095] Framing unit: Used to add pilot sequences and cyclic prefixes to the output data of the inverse fast Fourier transform unit;

[0096] Frequency conversion modulation unit: used to perform frequency conversion processing on the output data of the framing unit;

[0097] The signal output of the frequency conversion modulation unit is sent to the underwater acoustic channel.

[0098] As mentioned earlier, to achieve the output of one of the following: a single-carrier cyclic prefix signal, a single-carrier time-domain spread spectrum signal, an orthogonal frequency division multiplexing (OFDM) signal, or a multi-carrier frequency-domain spread spectrum signal, the following signal modulation scheme is designed. The transmit signal of the multi-mode transmitter is defined as S(n), where n = 1, 2, 3, ..., Q. s , n represents the sequence number of the transmitted signal, Q s Let s(n) be the number of transmitted signals.

[0099] The time-frequency diagrams of data frames for four signal modulation methods—Single-Carrier Cyclic Prefix (SC-CP), Single-Carrier Time-Domain Spread Spectrum (SC-TDSS), Orthogonal Frequency Division Multiplexing (OFDM), and Multi-Carrier Frequency-Domain Spread Spectrum (MC-FDSS)—are shown below. Figures 3a to 3b As shown.

[0100] SC-CP is a single-carrier modulation technique where data is modulated and mapped onto a single carrier. To combat frequency-selective fading, a cyclic prefix is ​​added at the beginning of each symbol period. At the receiver, the signal is first converted to the frequency domain using an FFT before intelligent signal decoding, making it well-suited for long-distance underwater acoustic communication.

[0101] A single-carrier cyclic prefix signal is represented as:

[0102] s(n) = x[k], k = 1, 2, 3, ..., Q S , n = k, where x[k] represents the digitally modulated transmitted data and k represents the sequence number of the transmitted data.

[0103] When a multi-mode transmitter is configured to transmit a single-carrier cyclic prefix signal, the second spread spectrum branch is activated, the second conversion branch is activated, and the framing unit adds a cyclic prefix to the signal.

[0104] Single-carrier time-domain spread spectrum modulation (SC-TDSS) is based on SC-CP and incorporates spread spectrum technology. The information signal is divided into multiple sub-channels, each with its own time-domain spread sequence. This spread sequence is multiplied by the data from each sub-channel to achieve time-domain spread of the signal. The spread spectrum technique used in this modulation reduces the data transmission rate, and the signal exhibits better anti-interference, anti-multipath, and anti-fading capabilities even with high latency and low signal-to-noise ratio. It is particularly suitable for complex and harsh underwater acoustic communication environments.

[0105] A single-carrier time-domain spread spectrum signal is represented as:

[0106] s(n) = x[k]c[i], i = 1, 2, 3, ..., N p n = (k-1)N p +i, where c[i] is the spreading sequence, i is the index of the spreading sequence, and N p It is the length of the spread spectrum sequence.

[0107] When a multi-mode transmitter is configured to transmit a single-carrier time-domain spread spectrum signal, the first spread spectrum branch is activated and the second transformation branch is activated to perform spread spectrum and inverse Fourier transform processing on the signal.

[0108] Orthogonal Frequency Division Multiplexing (OFDM) improves data transmission efficiency by transmitting multiple low-speed data streams simultaneously. It has a relatively fast transmission rate and good performance in combating multipath interference and frequency-selective fading, but it has high requirements for the channel signal-to-noise ratio.

[0109] Orthogonal frequency division multiplexing (OFDM) signals are represented as follows:

[0110] l = 1, 2, 3, ..., Q s n = l, k = 1, 2, 3, ..., NQ S Where N is fast

[0111] The number of subcarriers in the Fast Fourier Transform.

[0112] When a multi-mode transmitter is configured to transmit orthogonal frequency division multiplexing signals, the second spread spectrum branch is activated and the first conversion branch is activated.

[0113] MC-FDSS is a multi-carrier frequency domain spread spectrum technology. By adding spread spectrum technology to OFDM, data is distributed to different subcarriers, improving channel capacity and anti-interference capability, and making it more suitable for harsh channel environments.

[0114] When a multi-mode transmitter is configured to transmit a single-carrier frequency domain spread spectrum signal, the first spread spectrum branch is activated, the second conversion branch is activated, and the framing unit adds a cyclic prefix to the signal.

[0115] Multicarrier frequency domain spread spectrum signal is represented as:

[0116] l = 1, 2, 3, ..., Q s n = l, where y(p) is the spread spectrum data sequence, y(p) = x[k]c[i]. i = 1, 2, 3, ..., N p p = (k-1)N p +i.

[0117] The intelligent multi-standard receiver connects to the underwater acoustic channel, receives the output signal of the underwater acoustic channel, identifies the modulation mode of the output signal of the underwater acoustic channel, and analyzes the output signal of the underwater acoustic channel.

[0118] The intelligent multi-standard receiver mainly performs intelligent signal modulation mode identification and intelligent signal decoding processing.

[0119] In some embodiments of the present invention, the intelligent multi-standard receiver includes: a synchronization unit, a Doppler estimation and compensation unit, an intelligent modulation mode identification unit, a frequency conversion demodulation unit, a deframe unit, a fast Fourier transform unit, and an intelligent signal decoding unit.

[0120] Synchronization unit: Receives underwater acoustic communication signals passing through the underwater acoustic channel and performs data frame synchronization processing on the underwater acoustic communication signals to achieve underwater acoustic communication data frame synchronization;

[0121] Doppler estimation and compensation unit: Connected to the signal output of the synchronization unit, used to estimate and compensate for the Doppler frequency shift of the underwater acoustic channel;

[0122] Intelligent modulation mode identification unit: connected to the signal output of the Doppler estimation and compensation unit, used to intelligently identify the modulation mode of the underwater acoustic communication signal based on the bandpass underwater acoustic signal.

[0123] The intelligent modulation mode identification unit mainly completes the identification of signal modulation mode based on passband signal. Considering that the characteristics of time series signals are not obvious, this invention uses time-frequency analysis method to transform time series signals into time-frequency image signals for identification. Specifically, the time-frequency analysis method used is smooth pseudo-Wigner-Ville distribution.

[0124] The mathematical definition of a smooth pseudo-Wigner-Ville distribution can be expressed as:

[0125]

[0126] Among them, SPWD z (t,f) represents the smooth pseudo-Wigner-Ville distribution of the signal z(t), m(τ) is the time window function, g(ut) is the frequency window function, t and f are time and frequency respectively, and u and τ are integration variables.

[0127] Based on this, in some embodiments of the present invention, the step of intelligently identifying the modulation mode of the underwater acoustic communication signal according to the bandpass underwater acoustic signal includes:

[0128] Signal conversion module: used to convert underwater acoustic communication signal time series signals into time-frequency image signals;

[0129] Signal recognition neural network: Receives the time-frequency image signal and identifies the signal modulation mode based on the time-frequency image signal.

[0130] The time-frequency diagrams of data frames for four signal modulation methods—Single-Carrier Cyclic Prefix (SC-CP), Single-Carrier Time-Domain Spread Spectrum (SC-TDSS), Orthogonal Frequency Division Multiplexing (OFDM), and Multi-Carrier Frequency-Domain Spread Spectrum (MC-FDSS)—are shown below. Figures 3a to 3b As shown.

[0131] This invention employs a lightweight neural network that is easy to deploy in embedded systems for signal modulation pattern recognition. In different implementations, three typical lightweight neural networks—ShuffleNet, EfficientNet, and MobileNet—can be used, and their parameters are shown in Table 2.

[0132] The identification results of the intelligent modulation method identification unit are mainly used in two aspects: one is to select appropriate signal frequency conversion demodulation, deframe, and FFT processing, and the other is to select appropriate intelligent signal interpretation network models.

[0133] Table 2 Parameters of a lightweight neural network for signal modulation mode identification

[0134] Network Model Parameters (M) Floating-point operands (FLOPs) ShuffleNet 1.37 0.04 EfficientNet 5.3 1.0 MobileNet 3.5 0.31

[0135] The frequency conversion demodulation unit is connected to the output of the intelligent modulation mode identification unit. It is used to set the frequency conversion strategy according to the modulation mode identified by the intelligent modulation mode identification unit and to perform frequency conversion demodulation processing on the underwater acoustic communication signal.

[0136] The deframe unit is connected to the output of the frequency conversion demodulation unit and is used to deframe the underwater acoustic signal after frequency conversion demodulation based on the modulation method identified by the intelligent modulation method identification unit.

[0137] The Fast Fourier Transform (FFT) unit is connected to the output of the deframe unit and is used to perform a Fast Fourier Transform on the deframed underwater acoustic communication signal according to the modulation mode identified by the intelligent modulation mode identification unit, so as to obtain the frequency domain signal of the underwater acoustic communication signal, including the pilot sequence frequency domain signal and the data sequence frequency domain signal of the underwater acoustic communication signal.

[0138] The intelligent signal interpretation unit is connected to the output of the fast Fourier transform unit and the output of the intelligent modulation mode identification unit. It is used to interpret the underwater acoustic communication signal after the fast Fourier transform by combining the identified modulation mode and obtain communication data.

[0139] The input to the intelligent signal interpretation unit is the frequency domain signal of the pilot sequence and data sequence after fast Fourier transform processing, and the output is the baseband signal before digital demodulation. Signal interpretation is achieved based on deep learning networks, which employ data-driven deep learning networks and model-driven deep learning networks. Specifically, the data-driven deep learning network is an optimized DNN and LSTM neural network, and the model-driven deep learning network is a designed ModelDriven-Net neural network.

[0140] In some embodiments of this invention, the intelligent signal interpretation unit employs an LSTM network connected to the output of the Fast Fourier Transform unit to interpret the underwater acoustic communication signal after the Fast Fourier Transform. This invention uses a two-layer LSTM network. The network structure is shown in Table 3.

[0141] Table 3 LSTM Network Structure

[0142] Serial Number Layer (Name) type size 1 sequence Sequence Input 256 2 lstm_1 LSTM 128 3 lstm_2 LSTM 64 4 fc Fully connected 16 (spread spectrum) / 8 (non-spread spectrum) 5 regressionoutput Regression Output --

[0143] In some embodiments of the present invention, the intelligent signal interpretation unit adopts an optimized DNN network, which is connected to the output of the fast Fourier transform unit to perform signal analysis on the Fourier transformed underwater acoustic signal to obtain the analyzed signal. The DNN network is configured to set the number of available network layers and the number of network nodes according to the identified modulation method.

[0144] If the identified signal modulation method is a single-carrier cyclic prefix signal or an orthogonal frequency division multiplexing signal, the DNN network is configured to include a first number of network layers and a second number of network nodes.

[0145] If the signal modulation method is identified as a single-carrier time-domain spread spectrum signal or a multi-carrier time-domain spread spectrum signal, the DNN network is configured to include a third number of network layers and a fourth number of network nodes.

[0146] The third quantity is less than the first quantity, and the fourth quantity is less than the second quantity.

[0147] To optimize the structure of the DNN neural network, considering the characteristics of the signal modulation methods, a DNN network structure with a large number of layers and nodes was adopted for SC-CP and OFDM non-spread spectrum modulation signals. For SC-TDSS and MC-FDSS spread spectrum modulation signals, a DNN network with fewer layers and nodes was used to reduce the number of parameters. The specific DNN network structures used are shown in Tables 4 and 5.

[0148] Table 4. DNN network structure for non-spread spectrum modulated signals

[0149] Serial Number Layer (Name) type size 1 sequence Sequence Input 256 2 fc_1 Fully connected 500 3 relu_1 ReLU -- 4 fc_2 Fully connected 250 5 relu_2 ReLU -- 6 fc_3 Fully connected 120 7 relu_3 ReLU -- 8 fc_4 Fully connected 16 9 regressionoutput Regression Output --

[0150] Table 5. DNN network structure for spread spectrum modulation signals

[0151] Serial Number Layer (Name) type size 1 sequence Sequence Input 256 2 fc_1 Fully connected 500 3 relu_1 ReLU -- 4 fc_2 Fully connected 250 5 relu_2 ReLU -- 6 fc_3 Fully connected 8 7 regressionoutput Regression Output --

[0152] In some embodiments of the present invention, reference is made to Figures 4 to 7 The intelligent signal interpretation unit adopts a model-driven ModelDriven-Net neural network, which specifically includes a channel estimation subnet and a channel equalization subnet.

[0153] The channel estimation subnet includes: a least squares module (LS module) and a channel estimation neural network (BP neural network).

[0154] Least Squares Module (LS Module): Acquires the pilot sequence signal y output from the Fast Fourier Transform unit of the intelligent multi-mode receiver. p And pilot sequence signals x transmitted by multi-standard transmitters p The channel impulse response is generated after least squares processing. Channel impulse response It is divided into real information and imaginary information;

[0155] Backpropagation (BP) neural network for channel estimation: Input channel impulse response The real and imaginary parts of the signal are used to estimate the channel impulse response.

[0156] Specifically, the LS algorithm will receive the pilot signal y p With the pilot x of the transmitted signal p Performing least squares operations yields a coarse estimate of the channel impulse response. Simultaneously, the real and imaginary parts of the complex form of the channel impulse response are separated and input into a BP neural network for further optimization. After optimization, the real and imaginary parts are combined to obtain the channel impulse response.

[0157] The structure of the BP neural network for the channel estimation subnet is shown in Table 6.

[0158] Table 6 Estimated Subnet Network Structure

[0159] Serial Number Layer (Name) type size 1 sequence Sequence Input 128 2 fc_1 Fully connected 256 3 relu_1 ReLU -- 4 fc_2 Fully connected 128 5 regressionoutput Regression Output --

[0160] The training process of the channel estimation subnet is as follows: Figure 5 As shown, the channel estimation subnet extracts data from the noisy underwater acoustic channel output and uses it as the training and validation sets, while using the unnoisy underwater acoustic channel output as the labels for the training and validation sets.

[0161] The channel equalization subnet includes: the minimum mean square error module (MMSE module) and the signal equalization neural network module.

[0162] Minimum mean square error module: Connects to the output of the channel estimation neural network to obtain an estimate of the output channel impulse response. And input the data sequence signal y output from the Fast Fourier Transform unit of the intelligent multi-standard receiver. D After minimum variance calculation, the minimum mean square error signal x is output. MMSE The output x MMSE The signal is divided into real part information and imaginary part information;

[0163] Signal equalization neural network: For single-carrier cyclic prefix signals and single-carrier time-domain spread spectrum signals, obtain the output x. MMSE The real and imaginary parts of the signal are used to output an equalized communication data signal. Signal, The communication data equalization signal is processed by hard decision to obtain the communication data before digital demodulation. For orthogonal frequency division multiplexing (OFDM) signals and multi-carrier frequency domain spread spectrum (MFSS) signals, the neural network simultaneously obtains estimates of the output channel impulse response. The data sequence signal y output by the Fast Fourier Transform unit of the intelligent multi-standard receiver D The minimum mean square error signal x MMsE Output communication data equalization signal Signal, The communication data equalization signal is processed by hard decision to obtain the communication data before digital demodulation.

[0164] The signal equalization neural network uses BP network and DNN network.

[0165] Table 7 Single-Carrier Channel Equalized Subnet Structure

[0166]

[0167]

[0168] Table 8 Multicarrier signal channel equalization subnet

[0169] Serial Number Layer (Name) type size 1 imageinput Image input 6×64×1 2 fc_1 Fully connected 250 3 relu_1 ReLU -- 4 fc_2 Fully connected 120 5 relu_2 ReLU -- 6 fc_3 Fully connected Non-spread spectrum: 16 / Spread spectrum: 8 7 regressionoutput Regression Output --

[0170] The system simulation analysis process and results of the intelligent adaptive underwater acoustic communication system provided by this invention are as follows.

[0171] (1) Underwater acoustic channel modeling

[0172] The system simulation is based on the Millica underwater acoustic channel model. Only large-scale factors are considered during model building. Three typical channel models are established by changing the water depth, transmission distance, and relative position, as shown in Table 9. The three channels are represented by CH1, CH2, and CH3, respectively. Relative position I indicates that both the transceiver and receiver are at a distance of H / 2 from the bottom of the water. Relative position II indicates that the transmitter is at a distance of 3H / 4 from the bottom of the water, and the receiver is at a distance of H / 4 from the bottom of the water.

[0173] Due to the combined influence of various complex factors such as water movement, temperature and salinity changes, dynamic characteristics of the sea surface and seabed, signal frequency dependence, refraction and scattering effects, nonlinear propagation effects, and changes in environmental noise and interference sources, the maximum time delay and number of multipath paths of the channel will vary under the same underwater acoustic channel conditions. However, the overall characteristics of the channel are basically the same. Therefore, 3600 communication scenarios were selected for each channel.

[0174] Table 9 Communication scenario parameters for underwater acoustic channel modeling

[0175] Channel water depth H Transmission distance D relative position Number of channels Maximum delay CH1 100m 1000m Ⅰ 3600 30ms CH2 100m 5000m Ⅱ 3600 48ms CH3 500m 1000m Ⅱ 3600 21ms

[0176] (2) Dataset Construction

[0177] 1) Dataset Construction for Intelligent Signal Modulation Mode Recognition

[0178] 2000 images were generated for training the neural network under channel CH1 signal-to-noise ratios of -5dB, 0dB, 5dB, 10dB, and 15dB, and 200 images were used for testing at each signal-to-noise ratio.

[0179] The dataset construction and parameter settings for network training are shown in Table 10.

[0180] Table 10 Communication scenario parameters for underwater acoustic channel modeling

[0181]

[0182] 2) Dataset Construction for Intelligent Signal Interpretation

[0183] A) Dataset Construction for Data-Driven Networks

[0184] The data-driven network employs DNN and LSTM. The input to the dataset is a signal processed by FFT, with the real and imaginary parts of the signal separated before being fed into the neural network. The labels are the bit signals transmitted from the transmitter. The dataset was generated under a signal-to-noise ratio of 25dB, producing 50,000 frames, divided into a 35,000-frame training set and a 15,000-frame validation set. The Adam optimization algorithm was used, with a validation evaluation frequency of 500. The number of training epochs was modified to 13, with 3,500 iterations per epoch. A test set of 2,000 frames was generated every 2dB within the -10 to 10dB range to validate the model's performance.

[0185] B) Model-driven network dataset construction

[0186] The channel estimation subnet was trained under a signal-to-noise ratio of 25 dB. The input dataset consisted of the channel impulse response coarsely estimated by the traditional LS algorithm, labeled with the perfect channel impulse response without noise. The training and validation sets were divided in a 7:3 ratio, with 35,000 and 15,000 frames respectively. The Adam optimization algorithm was used, and the validation set evaluation frequency was 500. Compared to data-driven neural networks, model-driven neural networks are simpler to train, require less computation, and train faster. Therefore, the number of training rounds was changed to 10, with 3500 iterations per round. L2 regularization was used to prevent overfitting, with an L2 regularization value set to 0.002. After training, the parameters of the channel estimation subnet were fixed.

[0187] The channel equalization subnet was also trained at a signal-to-noise ratio of 25dB, and the dataset was again divided in a 7:3 ratio. The network training process was more complex than before; 10 training epochs were not optimal. Here, the network was trained for 20 epochs, with 1750 iterations per epoch. All other parameters remained the same. After training, a test set of 2000 frames was generated every 2dB within the range of -10 to 10dB to validate the model's performance.

[0188] (3) Network training parameters

[0189] The network training parameters for data-driven neural networks and model-driven neural networks are shown in Tables 10 and 11.

[0190] Table 10 Training parameters for data-driven neural networks

[0191] Serial Number Main parameters Parameter settings 1 Training signal-to-noise ratio 25dB 2 Optimization Algorithm Adam 3 Training dataset 35000 4 Validation dataset 15000 5 Validation set evaluation frequency 500 6 Maximum number of training rounds 13 7 Number of iterations per round 3500 8 Learning rate Initial value: 0.001, multiplied by 0.8 after two rounds. 9 L2 regularization 0.002

[0192] Table 11 Training parameters of the model-driven neural network

[0193]

[0194]

[0195] (4) System simulation test results and analysis

[0196] 1) Test results of intelligent signal modulation mode recognition

[0197] Figure 8 The image shows a comparison of the recognition accuracy of three neural networks. The results show that as the signal-to-noise ratio increases, the signal recognition rate also increases. Under the condition of a signal-to-noise ratio of 0dB, the recognition rates of the three networks all reached over 90%.

[0198] 2) Performance test results of intelligent signal interpretation

[0199] After training the neural network according to the network structure and training parameters, the performance of the system in intelligent signal interpretation under CH1 channel conditions was simulated and tested. The simulation results are as follows: Figure 8 As shown in the results, the signal processing module of the intelligent algorithm performs significantly better than the traditional LS+ZF and LS+MMSE algorithms, while the performance of the three intelligent methods is relatively similar. For SC-CP, SC-TDSS, and OFDM signal modulation methods, the model-driven underwater acoustic communication intelligent receiver exhibits better performance; for MC-FDSS signal modulation method, the model-driven underwater acoustic communication intelligent receiver performs similarly to the LSTM underwater acoustic communication intelligent receiver, but is generally slightly worse than the DNN.

[0200] The intelligent signal interpretation results of different signal modulation methods are as follows: Figures 9a to 9d As shown, the four signal modulation methods—SC-CP, SC-TDSS, OFDM, and MC-FDSS—have different channel adaptability capabilities, with bit error rates below 10%. -4 Under CH1 channel conditions, OFDM can be used in scenarios with an SNR of 8dB or higher, SC-CP can be used in scenarios with an SNR of 5dB or higher, MC-FDSS can be used in scenarios with an SNR of -3dB or higher, and SC-TDSS can be used in scenarios with an SNR of -7dB or higher, demonstrating the channel adaptability of intelligent multi-mode communication methods.

[0201] To verify the generalization performance of the three neural networks, the intelligent signal interpretation performance of the system under CH2 and CH3 channel conditions was further simulated and tested. The test results are as follows: Figure 11 , Figure 12 As shown in the figure, the results indicate that the three intelligent signal interpretation methods exhibit similar communication performance in the CH2 and CH3 channels as in the CH1 channel, demonstrating good generalization performance.

[0202] (5) Network computational complexity analysis

[0203] Table 12 shows a statistical comparison of the network parameters of the three neural networks. It can be seen that the model-driven neural network has fewer parameters. Considering that the performance of the three networks is comparable, the model-driven neural network is more suitable for deployment in embedded systems with limited computing resources.

[0204] Table 12 Comparison of parameter quantities for three types of neural networks

[0205]

[0206]

[0207] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent adaptive underwater acoustic communication system, characterized in that, include: Multi-standard transmitter: includes a data transmission unit for transmitting communication data, a standard switching unit for switching the modulation mode of the transmitted signal, and a transmitted signal modulation unit for modulating the signal; the data transmission unit, the standard switching unit, and the transmitted signal modulation unit are connected in sequence; the output of the multi-standard transmitter is connected to an underwater acoustic channel; Intelligent multi-standard receiver: includes a synchronization unit, a Doppler estimation and compensation unit, an intelligent modulation mode identification unit, a data processing unit, an intelligent signal decoding unit, and a digital demodulation unit; The synchronization unit receives underwater acoustic communication signals passing through the underwater acoustic channel and performs data frame synchronization processing on the underwater acoustic communication signals; the Doppler estimation and compensation unit receives the output data of the synchronization unit and estimates and compensates for the Doppler frequency shift of the underwater acoustic channel. The intelligent modulation scheme identification unit is connected to the signal output of the Doppler estimation and compensation unit, and is used to intelligently identify the modulation scheme of the underwater acoustic communication signal based on the bandpass underwater acoustic signal; the data processing unit is connected to the signal output of the intelligent modulation scheme identification unit; the intelligent signal decoding unit is connected to the signal output of the data processing unit, and... The output of the intelligent modulation mode identification unit is used to intelligently interpret the underwater acoustic communication frequency domain signal after the fast Fourier transform by combining the identified modulation mode to obtain communication data; the digital demodulation unit is connected to the signal output of the intelligent signal interpretation unit and modulates the communication data to the data receiving unit. Channel information identification unit: Its input end is connected to the signal output end of the synchronization unit of the intelligent multi-mode receiver, and its output end is connected to the mode switching unit of the multi-mode transmitter. The channel information identification unit is used to identify the signal-to-noise ratio and maximum channel delay of the underwater acoustic channel and transmit them to the mode switching unit. The mode switching unit adjusts and switches the modulation mode of the transmitted signal according to the signal-to-noise ratio and maximum channel delay. The intelligent modulation mode identification unit uses a lightweight neural network to intelligently identify the signal modulation mode of the underwater acoustic passband signal. The intelligent signal interpretation unit uses a deep learning neural network to intelligently interpret the underwater acoustic communication signal to recover the communication data.

2. The intelligent adaptive underwater acoustic communication system as described in claim 1, characterized in that, The transmitted signal modulation unit includes: Digital modulation unit: connected to the output of the system switching unit; The spread spectrum unit includes a first spread spectrum branch and a second spread spectrum branch. The first spread spectrum branch is provided with a switching assembly and a spread spectrum module. The output terminal of the digital modulation unit is connected to the first spread spectrum branch and the second spread spectrum branch respectively. The first spread spectrum branch and the second spread spectrum branch are connected and converge to form the output terminal of the spread spectrum unit. The inverse fast Fourier transform unit includes a first transform branch and a second transform branch. The first transform branch is equipped with a switching component and an inverse Fourier transform module. The output terminal of the spread spectrum unit is connected to the first transform branch and the second transform branch respectively. The output terminals of the first transform branch and the second transform branch converge to form the output terminal of the inverse fast Fourier transform unit. Framing unit: Used to add pilot sequences and cyclic prefixes to the output data of the inverse fast Fourier transform unit; Frequency conversion modulation unit: used to perform frequency conversion processing on the output data of the framing unit; The signal output terminal of the frequency conversion modulation unit outputs a signal to the underwater acoustic channel.

3. The intelligent adaptive underwater acoustic communication system as described in claim 2, characterized in that, The multi-mode transmitter is configured to transmit one of the following modulation signal types: single-carrier cyclic prefix signal, single-carrier time-domain spread spectrum signal, orthogonal frequency division multiplexing signal, and multi-carrier frequency-domain spread spectrum signal. When the multi-mode transmitter is configured to transmit a single-carrier cyclic prefix signal, the second spread spectrum branch is activated, and the second conversion branch is activated. When the multi-mode transmitter is configured to transmit a single-carrier time-domain spread spectrum signal, the first spread spectrum branch is turned on and the second conversion branch is turned on. When the multi-mode transmitter is configured to transmit orthogonal frequency division multiplexing signals, the second spread spectrum branch is turned on and the first conversion branch is turned on. When the multi-mode transmitter is configured to transmit a multi-carrier frequency domain spread spectrum signal, the first spread spectrum branch is activated and the first conversion branch is activated.

4. The intelligent adaptive underwater acoustic communication system as described in claim 3, characterized in that, Define the transmit signal of the multi-mode transmitter before frequency conversion modulation as s(n), n=1,2,3,…,Q S , n represents the sequence number of the transmitted signal, Q S The number of transmitted signals s(n): The single-carrier cyclic prefix signal is represented as: s(n) = x[k], k = 1, 2, 3, ..., Q S , n = k, where x[k] represents the digitally modulated transmitted data and k represents the sequence number of the transmitted data; The single-carrier time-domain spread spectrum signal is represented as follows: Where c[i] is the spreading sequence, i is the sequence number of the spreading sequence, and N is the number of the spreading sequence. p It is the length of the spreading sequence; The orthogonal frequency division multiplexed signal is represented as follows: Where N is the number of subcarriers in the Fast Fourier Transform; The multi-carrier frequency domain spread spectrum signal is represented as follows: Where y(p) is the spread spectrum data sequence, 5. The intelligent adaptive underwater acoustic communication system as described in any one of claims 1 to 4, characterized in that, The data processing unit of the intelligent multi-standard receiver includes: Frequency conversion demodulation unit: connected to the output of the intelligent modulation mode identification unit, used to set the frequency conversion strategy according to the modulation mode identified by the intelligent modulation mode identification unit, and to perform frequency conversion demodulation processing on the underwater acoustic communication signal; Deframe unit: Connected to the output of the frequency conversion demodulation unit, it is used to deframe the underwater acoustic signal after frequency conversion demodulation based on the modulation method identified by the intelligent modulation method identification unit. Fast Fourier Transform Unit: Connected to the output of the deframe unit, it is used to perform a Fast Fourier Transform on the deframed underwater acoustic communication signal according to the modulation mode identified by the intelligent modulation mode identification unit, to obtain the frequency domain signal of the underwater acoustic communication signal, including the pilot sequence frequency domain signal and the data sequence frequency domain signal of the underwater acoustic communication signal. The intelligent signal interpretation unit is connected to the output of the fast Fourier transform unit and the output of the intelligent modulation mode identification unit. It is used to intelligently interpret the underwater acoustic communication frequency domain signal after the fast Fourier transform by combining the identified modulation mode to obtain communication data.

6. The intelligent adaptive underwater acoustic communication system as described in claim 5, characterized in that, The intelligent modulation scheme identification unit includes: Signal conversion module: used to convert underwater acoustic communication signal time series signals into time-frequency image signals; Signal recognition neural network: receives the time-frequency image signal and identifies the signal modulation mode based on the time-frequency image signal; The signal recognition neural network mentioned above uses lightweight neural networks, including ShuffleNet, EfficientNet, and MobileNet.

7. The intelligent adaptive underwater acoustic communication system as described in claim 5, characterized in that, The intelligent signal interpretation unit includes a deep learning network, which is a data-driven optimized DNN neural network, a data-driven LSTM neural network, or a model-driven ModelDriven-Net neural network. The input of the deep learning network is connected to the output of the fast Fourier transform unit to perform intelligent signal interpretation on the underwater acoustic communication frequency domain signal after the fast Fourier transform, thereby obtaining communication data. The data-driven optimized DNN is configured to set the number of available network layers and network nodes according to the identified modulation scheme: if the identified signal modulation scheme is a single-carrier cyclic prefix signal or an orthogonal frequency division multiplexing signal, the DNN is configured to include a first number of network layers and a second number of network nodes; if the identified signal modulation scheme is a single-carrier time-domain spread spectrum signal or a multi-carrier frequency-domain spread spectrum signal, the DNN is configured to include a third number of network layers and a fourth number of network nodes; the third number is less than the first number, and the fourth number is less than the second number. The data-driven LSTM neural network is configured as a two-layer LSTM network. The ModelDriven-Net neural network is a deep learning network that combines intelligent signal interpretation networks with traditional channel estimation and channel equalization algorithms, including a channel estimation subnet and a channel equalization subnet. The channel estimation subnet includes: Least squares module: Obtains the pilot sequence signal y output by the Fast Fourier Transform unit of the intelligent multi-standard receiver. p And pilot sequence signals x transmitted by multi-standard transmitters p The channel impulse response is generated after least squares processing. Channel impulse response It is divided into real information and imaginary information; Channel estimation neural network: input channel impulse response The real and imaginary parts of the signal are used to estimate the channel impulse response. The channel equalization subnet includes: Minimum mean square error module: Connects to the output of the channel estimation neural network to obtain an estimate of the output channel impulse response. And input the data sequence signal y output from the Fast Fourier Transform unit of the intelligent multi-standard receiver. D After minimum variance calculation, the minimum mean square error signal x is output. MMSE The output x MMSE The signal is divided into real part information and imaginary part information; Signal equalization neural network: For single-carrier cyclic prefix signals and single-carrier time-domain spread spectrum signals, obtain the output x. MMSE The real and imaginary parts of the signal are used to output an equalized communication data signal. Signal, The communication data equalization signal is processed by hard decision to obtain the communication data before digital demodulation. For orthogonal frequency division multiplexing (OFDM) signals and multi-carrier frequency domain spread spectrum (MFSS) signals, the neural network simultaneously obtains estimates of the output channel impulse response. The data sequence signal y output by the Fast Fourier Transform unit of the intelligent multi-standard receiver D The minimum mean square error signal x MMSE Output data equalization signal Signal, The communication data equalization signal is processed by hard decision to obtain the communication data before digital demodulation.

Images