An end-to-end semantic communication method and system for fishing vessel networking in marine environments

Abstract

This disclosure provides an end-to-end semantic communication method and system for fishing vessel networking in a marine environment, relating to the fields of wireless communication and artificial intelligence. The method includes: acquiring marine buoy data and converting it into a marine buoy data matrix; converting the preprocessed marine buoy data matrix into a marine buoy network data tensor and inputting it into a semantic communication network for a marine environment; repeatedly extracting feature maps of the marine buoy network data tensor using convolutional blocks; continuously adjusting the feature maps according to the signal-to-noise ratio using adaptive feature blocks to obtain feature vectors; compressing the feature vectors proportionally based on the data compression ratio to obtain semantic feature vectors; attenuating the semantic feature vectors through the overall channel response of the channel model to obtain the final output signal; and generating recovered data through the inverse reasoning process of the semantic decoder using the final output signal.

Description

Technical Field

[0001] This disclosure relates to the fields of wireless communication and artificial intelligence technologies, specifically to an end-to-end semantic communication method and system for fishing vessel networking in a marine environment. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] With the rapid development of the marine economy, the demand for efficient and reliable communication is increasing in scenarios such as fishing vessel operations, fishery resource management, and maritime emergency rescue. However, the maritime communication environment is uniquely complex, mainly manifested in problems such as long-distance transmission, significant multipath effects, severe Doppler shift, large signal attenuation, and interference from severe weather. These factors lead to traditional wireless communication technologies facing severe signal attenuation, high latency, and poor reliability in the maritime environment.

[0004] Semantic communication, as an emerging communication paradigm, significantly reduces communication overhead and improves transmission efficiency by extracting and transmitting semantic features of information rather than raw data. This gives semantic communication a clear advantage in environments with limited bandwidth and high latency sensitivity. Furthermore, as a novel communication technology that integrates communication and intelligence, semantic communication proposes a new approach that replaces modular, separate optimization with end-to-end integrated optimization. This enables a significant improvement in the overall performance of the communication system with a simpler network structure.

[0005] Although semantic communication has great potential for application in the marine environment, existing semantic communication technologies are mainly geared towards terrestrial environments and have not fully considered the special characteristics of the complex marine propagation environment, thus still having the following problems:

[0006] 1) Due to the strong reflectivity of seawater and the interference of weather conditions (such as wind, tides, ocean currents, etc.) on the signal, the signal will encounter severe attenuation and multipath effect during propagation, making it impossible to simulate the real marine environment and incorporate it into the learning paradigm.

[0007] 2) The complex marine environment leads to the variability of marine data, making it difficult to effectively process the complex semantic information in marine data, resulting in problems such as high latency and low transmission accuracy.

[0008] 3) In the marine environment, fishing fleet operations usually require multi-vehicle coordination. Traditional communication modes are difficult to meet networking requirements, and existing networking technologies lack support for semantic communication, resulting in low communication efficiency and insufficient resource utilization. Summary of the Invention

[0009] To address the aforementioned issues, this disclosure proposes an end-to-end semantic communication method and system for fishing vessel networking in a marine environment. It extracts semantic information from marine data, constructs a semantic communication environment model based on the Ricean channel model, simulates problems such as attenuation, latency, and multipath effects in marine communication, and incorporates these into the learning paradigm. Through multi-scale feature learning, the depth of the generated feature map is increased, significantly enhancing the model's ability to extract complex semantic features and achieving a significant improvement in the overall performance of the communication system.

[0010] According to some embodiments, the present disclosure adopts the following technical solutions:

[0011] An end-to-end semantic communication method for fishing vessel networking in a marine environment includes:

[0012] Acquire ocean buoy data, convert the ocean buoy data into matrix data according to time granularity and spatial granularity to obtain ocean buoy data matrix, and preprocess it;

[0013] The preprocessed ocean buoy data matrix is ​​transformed into an ocean buoy network data tensor and input into a semantic communication network for the marine environment to obtain the recovered data of end-to-end semantic communication.

[0014] In this process, the ocean buoy network data tensor is input into a semantic communication network oriented towards the marine environment. The feature map of the ocean buoy network data tensor is extracted multiple times using convolutional blocks. Then, the feature map is continuously adjusted according to the signal-to-noise ratio using adaptive feature blocks to obtain the feature vector. The feature vector is then compressed proportionally based on the data compression rate to obtain the semantic feature vector. The semantic feature vector is attenuated through the overall channel response of the channel model to obtain the final output signal. The final output signal is then used to generate the recovered data through the inverse reasoning process of the semantic decoder.

[0015] According to some embodiments, the present disclosure adopts the following technical solutions:

[0016] An end-to-end semantic communication system for fishing vessel networking in a marine environment, characterized by comprising:

[0017] The data acquisition module is used to acquire ocean buoy data, convert the ocean buoy data into matrix data according to time and spatial granularity, obtain the ocean buoy data matrix, and preprocess it;

[0018] The semantic recovery module is used to convert the preprocessed ocean buoy data matrix into an ocean buoy network data tensor and input it into a semantic communication network for the marine environment to obtain the recovered data of end-to-end semantic communication.

[0019] In this process, the ocean buoy network data tensor is input into a semantic communication network oriented towards the marine environment. The feature map of the ocean buoy network data tensor is extracted multiple times using convolutional blocks. Then, the feature map is continuously adjusted according to the signal-to-noise ratio using adaptive feature blocks to obtain the feature vector. The feature vector is then compressed proportionally based on the data compression rate to obtain the semantic feature vector. The semantic feature vector is attenuated through the overall channel response of the channel model to obtain the final output signal. The final output signal is then used to generate the recovered data through the inverse reasoning process of the semantic decoder.

[0020] According to some embodiments, the present disclosure adopts the following technical solutions:

[0021] A computer program product includes a computer program that, when executed by a processor, implements the aforementioned end-to-end semantic communication method for fishing vessel networking in a marine environment.

[0022] According to some embodiments, the present disclosure adopts the following technical solutions:

[0023] A non-transitory computer-readable storage medium is provided for storing computer instructions, which, when executed by a processor, implement the aforementioned end-to-end semantic communication method for fishing vessel networking in a marine environment.

[0024] According to some embodiments, the present disclosure adopts the following technical solutions:

[0025] An electronic device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the described end-to-end semantic communication method for fishing vessel networking in a marine environment.

[0026] Compared with the prior art, the beneficial effects of this disclosure are as follows:

[0027] This disclosure presents an end-to-end semantic communication method for fishing vessel networking in a marine environment. Based on an end-to-end fishing vessel communication network using semantic communication, and combined with end-to-end optimization design of deep neural networks, a semantic communication network for the marine environment is constructed, significantly improving the reliability of maritime communication. Furthermore, the system is trained and learned in a learning paradigm based on the Ricean channel model to simulate multipath effects, signal attenuation, and latency at sea, enhancing the robustness of the system in complex marine environments. The network employs a semantic encoder and decoder combined with multi-scale feature learning technology, improving the feature extraction and recovery capabilities of marine buoy data, and effectively coping with complex marine environments.

[0028] This disclosure presents an end-to-end semantic communication method for fishing vessel networking in a maritime environment, comprising two main innovative improvements: First, it incorporates a Ricean fading channel learning paradigm, proposing a semantic communication network model for the maritime environment to extract and recover semantic information from maritime data. Based on this, it models the semantic communication environment using the Ricean channel model, simulating issues such as attenuation, latency, and multipath effects encountered in maritime communication, and incorporates these into the learning paradigm. Second, through multi-scale feature learning techniques, combined with multiple iterations of feature map extraction and increased depth of the generated feature maps, it enhances the model's ability to extract complex semantic features, significantly improving the overall performance of the communication system. This enables end-to-end semantic communication for fishing vessel networking in complex maritime propagation environments, achieving efficient, reliable, and low-latency maritime communication. Attached Figure Description

[0029] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0030] Figure 1 This is a schematic diagram of the marine propagation environment for point-to-point fishing vessel communication according to an embodiment of this disclosure;

[0031] Figure 2 This is a flowchart of the end-to-end fishing vessel networking semantic communication network model processing for complex maritime propagation environments according to an embodiment of this disclosure.

[0032] Figure 3 This is a diagram illustrating the data transmission recovery effect of an ocean buoy according to an embodiment of this disclosure.

[0033] Figure 4 This is a diagram illustrating the effect of marine image data transmission restoration according to an embodiment of this disclosure;

[0034] Figure 5 This is a schematic diagram of semantic encoder compression according to an embodiment of the present disclosure. Detailed Implementation

[0035] The present disclosure will be further described below with reference to the accompanying drawings and embodiments.

[0036] It should be noted that the following detailed descriptions are illustrative and intended to provide further explanation of this disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains.

[0037] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this disclosure. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms “comprising” and / or “including” are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0038] Example 1

[0039] This disclosure discloses an end-to-end semantic communication method for fishing vessel networking in a marine environment, aiming to improve the reliability of communication networks in the complex and ever-changing marine environment. The steps include:

[0040] Step 1: Acquire ocean buoy data, convert the ocean buoy data into matrix data according to time and spatial granularity to obtain the ocean buoy data matrix, and preprocess it;

[0041] Step 2: Convert the preprocessed ocean buoy data matrix into an ocean buoy network data tensor and input it into a semantic communication network for the marine environment to obtain the recovered data of end-to-end semantic communication;

[0042] In this process, the ocean buoy network data tensor is input into a semantic communication network oriented towards the marine environment. The feature map of the ocean buoy network data tensor is extracted multiple times using convolutional blocks. Then, the feature map is continuously adjusted according to the signal-to-noise ratio using adaptive feature blocks to obtain the feature vector. The feature vector is then compressed proportionally based on the data compression rate to obtain the semantic feature vector. The semantic feature vector is attenuated through the overall channel response of the channel model to obtain the final output signal. The final output signal is then used to generate the recovered data through the inverse reasoning process of the semantic decoder.

[0043] As one embodiment, this disclosure presents an end-to-end semantic communication method for fishing vessel networking in a marine environment, comprising two main innovations: 1) Adding a Ricean fading channel learning paradigm to propose a semantic communication model for the marine environment; 2) Enhancing the model's ability to extract complex semantic features through multi-scale feature learning techniques combined with increasing the depth of the generated feature maps. The specific training implementation process is as follows:

[0044] Step 1: Obtain ocean buoy data. This can be done through open data platforms, such as NOAA or GOOS, which provide buoy location (latitude and longitude) and environmental data (such as temperature, salinity, and current velocity); or by obtaining RGB ocean image data through the Kaggle platform.

[0045] As one example, the essence of semantic feature extraction is the same for both image data and ocean buoy data. Both methods utilize convolutional neural networks to extract semantic features and retain only key information during channel transmission, thereby reducing redundancy and improving communication efficiency. Therefore, this application's embodiment uses ocean buoy data processing as an example to implement an end-to-end semantic communication method for fishing vessel networking in a marine environment. While it still retains basic image semantic transmission, the processing steps are largely the same and will not be elaborated upon here.

[0046] Specifically, current maritime communication systems mainly consist of key components such as coastal base stations, relay nodes, communication satellites, communication vessels, and unmanned aerial vehicles (UAVs). Coastal base stations, as an extension of the land-based communication network, are responsible for uploading marine monitoring data and transmitting operational scheduling commands, enabling efficient information exchange between ships and land. When vessels are far from shore base stations, communication satellites provide long-distance data transmission, ensuring that the fleet can maintain real-time connectivity with the command center. Relay nodes and UAVs play a supporting role in maritime communication, extending coverage and improving the system's flexibility and reliability. Under this premise, such as... Figure 1 As shown, this disclosure focuses on point-to-point communication between fishing vessels and proposes a fishing vessel networking system based on semantic communication. The system consists of a sending vessel and a receiving vessel. The sending vessel, as the communication core, is responsible for coordinating information exchange between vessels and receiving dispatch instructions from the shore command center via a coastal base station. The receiving vessel mainly undertakes fishing operation tasks, including sensor data acquisition, fishing, and patrol monitoring, and transmits relevant data to the sending vessel.

[0047] In this system, any two vessels can act as either the sending or receiving vessels, thereby enhancing the flexibility of the communication system, optimizing data transmission efficiency, and meeting the needs of marine fisheries for low-latency and high-reliability communication.

[0048] Furthermore, the high conductivity and dielectric constant of seawater make it a highly efficient medium for reflecting electromagnetic waves. According to electromagnetic theory, the influence of seawater on radio waves is mainly manifested in:

[0049] (1) Specular Reflection: For higher frequency (such as microwave, millimeter wave) signals, the sea surface can be approximated as a smooth surface, so that the signal undergoes specular reflection according to Snell's law, forming a predictable reflection path.

[0050] (2) Diffuse Reflection: Due to the fluctuations on the sea surface (such as ripples and waves), when a signal is incident on the sea surface, some of the energy will be scattered in multiple directions, resulting in signal power attenuation and random phase changes.

[0051] (3) Absorption: Some electromagnetic wave energy will be absorbed by seawater, resulting in energy loss. In particular, low-frequency wireless signals will be attenuated to a certain extent.

[0052] Figure 1 The reflective surface directly demonstrates the processes of specular reflection and diffuse reflection, while absorption is manifested as energy attenuation, i.e., a decrease in amplitude.

[0053] Step 2: Convert the ocean buoy data into matrix data according to the temporal and spatial granularities to obtain the ocean buoy data matrix, and preprocess it;

[0054] Specifically, the acquired raw ocean buoy data was processed into matrix data according to temporal and spatial granularities. The temporal granularity was in hours, and the spatial granularity was in coordinates of different work stations, further divided based on three data types: temperature, wave, and wind. After processing, the resulting ocean buoy data matrix was obtained.

[0055]

[0056] in, This represents the matrix of ocean buoy data acquired during the t-th time period. H Represents the height of the matrix. W Indicates the width of the matrix. C Represents different dimensions of representation for the same data type.

[0057] As one example, if it is image data, it is transmitted in RGB three-dimensional pixel matrix format.

[0058] Furthermore, the obtained ocean buoy data matrix is ​​preprocessed, including:

[0059] With each ocean buoy data matrix The data is normalized, missing values ​​are imputed, and outliers are handled on a per-unit basis to improve data quality and ensure model stability.

[0060] (1) Because ocean buoys may experience data loss due to equipment failure or communication interruption, the resulting NaN data (Not a Number, a special floating-point number representing invalid or missing values, usually caused by abnormal operation of the sensor data acquisition system) will lead to NaN values ​​in the neural network training gradient, thus preventing the normal updating of model parameters. Therefore, it is necessary to fill in the NaN values ​​in the matrix. The mathematical representation of the filling process is formula (1):

[0061] (1)

[0062] in: Here, represents the data point at time step t, position h,w, and channel c; N is the total number of non-NaN data points in the current column. To prevent data imputation failure due to large-area data loss, this disclosure re-detects NaN values ​​in the training data before model training, and uses multiple ocean buoy data matrices. The mean of the non-NaN value data in the corresponding channel is used as the secondary filling process.

[0063] (2) Using the Min-Max normalization method, the data is scaled to the range [0,1]. The calculation formula (2) is as follows:

[0064] (2)

[0065] in, and These are the minimum and maximum values ​​of the dataset, respectively. This is the normalized ocean buoy data matrix.

[0066] Step 3: Convert the preprocessed ocean buoy data matrix into an ocean buoy network data tensor and input it into a semantic communication network for the marine environment to obtain the recovered end-to-end semantic communication data, including:

[0067] Step 31: Construct and train a semantic communication network for the maritime environment;

[0068] Specifically, firstly, the ocean buoy data matrix Transformation Tensor, specifically the data tensor of an ocean buoy network. The dataset was divided into batches, and for the ocean buoy data matrix, samples were randomly drawn from it. One data segment was used as training data, and the remaining... The data segment serves as test data and as training input for the subsequent semantic communication network.

[0069] Furthermore, based on ocean buoy network data tensors The semantic communication network (MS) for the marine environment is constructed based on Channel State Information (CSI), Data Compression Ratio (CR), and the number of channels. The MS includes a speech encoder, a marine channel model, and a semantic decoder. Specifically, the marine channel model is constructed based on CSI, and the semantic encoder and decoder are constructed based on the data compression ratio (CR) and the number of channels. The semantic decoder is based on the marine buoy network data tensor. Update the model parameters.

[0070] Finally, the training process employs the gradient descent algorithm to train the semantic communication network for the maritime environment. The gradient of each parameter (i.e., the weights in the model) with respect to the loss is calculated based on backpropagation until the loop ends.

[0071] Step 32: Convert the normalized ocean buoy data matrix into a tensor to obtain the ocean buoy network data tensor. After inputting it into the semantic communication network for the marine environment, the specific processing procedure of the semantic communication network for the marine environment is as follows:

[0072] First, the ocean buoy network data tensor is fed into the semantic encoder, and the ocean buoy network data tensor is extracted based on convolutional blocks (conv). The spatial patterns or local correlations are normalized by a generalized division normalization layer (GDN) to remove redundant information from the input data and reduce overfitting of certain features. Then, a PReLU layer is used to non-linearly activate the feature matrix to prevent neurons from "dying" during training. The calculation process is shown in formula (3):

[0073] (3)

[0074] Furthermore, the features of the output feature map are extracted a second time. At the same time, in order to effectively alleviate the gradient vanishing problem in deep networks, a residual structure is constructed, as shown in formula (4):

[0075] (4)

[0076] Furthermore, based on the calculated mean of the features and combined with the signal-to-noise ratio (SNR), an adaptive feature block AF is constructed. The adaptive feature block can adjust the features according to the SNR, thereby helping the model to better handle noise or other environmental changes, as shown in Equation (5):

[0077] (5)

[0078] The feature matrix F output by equation (5) is fed back into equation (3) as input, and the process of (3), (4), and (5) is repeated N times to obtain the output feature vector. Finally, based on the data compression ratio CR, the feature vectors are... A semantic encoder is designed to compress data at a ratio of CR, and its output semantic information is a one-dimensional tensor. The compression process is as follows: Figure 5 As shown, assuming the encoder outputs 1024 bits of semantic information and the compression ratio CR=0.1, we discard the last 90% of the data. The sender actually sends the first 1024 * 0.1 = 102 (rounded) bits of data. This yields the semantic feature output of the semantic encoder, i.e., the semantic feature vector. .

[0079] Secondly, the semantic feature vectors are attenuated using the overall channel response of the maritime channel model to obtain the final output signal, including:

[0080] First, a channel model is constructed based on Channel State Information (CSI). The Ricean fading channel was chosen to simulate the marine environment. The Ricean fading channel is a common fading model in wireless communication. The model assumes the existence of a dominant direct path signal, as well as multiple path signals generated by reflection, refraction, or scattering from other objects.

[0081] Because the sea surface is a strong reflective surface, maritime communication typically generates multiple transmission paths. Point-to-point fishing vessel communication at sea usually has a strong LOS (Lowest Optical Length) direct path signal. Therefore, the Ricean channel model is very suitable for simulating signal propagation characteristics in maritime communication. By adjusting the K value, the Ricean model can better adapt to different maritime propagation conditions. The specific process is as follows:

[0082] Determining the channel gain coefficient based on signal-to-noise ratio (SNR) ,have The fading of the analog signal in the Ricean channel divides the signal into two parts: a real part and an imaginary part, forming a complex signal. The complex form of the LOS direct path signal is obtained by calculation, where K is the power ratio of the direct path to the multipath scattering path in Rician fading, L is the characteristic length, and the LOS (direct path) components are shown in formula (6):

[0083] (6)

[0084] The signal components of the scattering path are random signals with a Gaussian distribution, as shown in formula (7):

[0085] (7)

[0086] The signals from each scattering path are subjected to random time delay processing, and the signals from all scattering paths are summed to obtain the final multipath scattering signal. The LOS direct path signal and the multipath scattering signal are added together to obtain the total channel response. Semantic feature vector Through the overall channel response Attenuation processing is performed to obtain the output signal after passing through the fading channel. .

[0087] Based on the set noise power P, Gaussian noise is generated and added to the signal to obtain the received signal. The received signal is then normalized to remove the influence of channel gain, resulting in the final output signal. As shown in formula (8):

[0088] (8)

[0089] Finally, the final output signal is used to generate recovered data through the inverse reasoning process of the semantic decoder, including:

[0090] To recover the compressed output signal information from a semantic communication network oriented towards a maritime environment, a semantic decoder is constructed based on a deconvolution layer. Feature extraction is performed on the semantically compressed data passing through the channel based on deconvolution blocks, the output feature map is normalized by a generalized division normalization layer (GDN), and then the feature matrix is ​​nonlinearly activated by a PReLU layer, as shown in Equation (9).

[0091] (9)

[0092] The residual structure is reconstructed as shown in formula (10):

[0093] (10)

[0094] Corresponding to the encoder part, an adaptive feature block AF is constructed to perform inverse reasoning on the encoded data, as shown in formula (11):

[0095] (11)

[0096] The recovery feature matrix of equation (11) The data is fed back into formula (9) as input, and the process (9), (10), and (11) is repeated N times to obtain the recovered data. .

[0097] Example 2

[0098] One embodiment of this disclosure provides an end-to-end semantic communication system for fishing vessel networking in a marine environment, comprising:

[0099] The data acquisition module is used to acquire ocean buoy data, convert the ocean buoy data into matrix data according to time and spatial granularity, obtain the ocean buoy data matrix, and preprocess it;

[0100] The semantic recovery module is used to convert the preprocessed ocean buoy data matrix into an ocean buoy network data tensor and input it into a semantic communication network for the marine environment to obtain the recovered data of end-to-end semantic communication.

[0101] In this process, the ocean buoy network data tensor is input into a semantic communication network oriented towards the marine environment. The feature map of the ocean buoy network data tensor is extracted multiple times using convolutional blocks. Then, the feature map is continuously adjusted according to the signal-to-noise ratio using adaptive feature blocks to obtain the feature vector. The feature vector is then compressed proportionally based on the data compression rate to obtain the semantic feature vector. The semantic feature vector is attenuated through the overall channel response of the channel model to obtain the final output signal. The final output signal is then used to generate the recovered data through the inverse reasoning process of the semantic decoder.

[0102] Example 3

[0103] One embodiment of this disclosure provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned end-to-end semantic communication method for fishing vessel networking in a marine environment.

[0104] Example 4

[0105] One embodiment of this disclosure provides a non-transitory computer-readable storage medium for storing computer instructions. When these computer instructions are executed by a processor, they implement the aforementioned end-to-end semantic communication method for fishing vessel networking in a marine environment.

[0106] Example 5

[0107] One embodiment of this disclosure provides an electronic device, including: a processor, a memory, and a computer program; wherein the processor is connected to the memory, and the computer program is stored in the memory. When the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to implement the described end-to-end semantic communication method for fishing vessel networking in a marine environment.

[0108] This disclosure is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0110] While the specific embodiments of this disclosure have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of this disclosure. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of this disclosure are still within the scope of protection of this disclosure.

Claims

1. An end-to-end semantic communication method for fishing vessel networking in a marine environment, characterized in that, include: Acquire ocean buoy data, convert the ocean buoy data into matrix data according to time granularity and spatial granularity to obtain ocean buoy data matrix, and preprocess it; The preprocessed ocean buoy data matrix is ​​transformed into an ocean buoy network data tensor and input into a semantic communication network for the marine environment to obtain the recovered data of end-to-end semantic communication. In this process, after the ocean buoy network data tensor is input into the semantic communication network for the marine environment, the feature map of the ocean buoy network data tensor is extracted multiple times using convolutional blocks. Then, the feature map is continuously adjusted according to the signal-to-noise ratio by adaptive feature blocks to obtain the feature vector. The feature vector is then compressed according to the data compression ratio to obtain the semantic feature vector. The semantic feature vector is attenuated by the total channel response of the channel model to obtain the final output signal. The final output signal is then used to generate the recovered data through the inverse reasoning process of the semantic decoder. A channel model is constructed based on Channel State Information (CSI). A Ricean channel is selected to simulate the marine environment. The Ricean channel is a fading model, which assumes the existence of a dominant direct path signal and multiple path signals generated by reflection, refraction, or scattering from other objects. The semantic feature vector is attenuated through the overall channel response of the channel model, including: determining the channel gain coefficient based on the signal-to-noise ratio (SNR); simulating signal fading in the Ricean channel; dividing the signal into real and imaginary parts to form a complex signal; calculating the complex form of the direct path signal; performing random time delay processing on the signal of each scattering path; accumulating the signals of all scattering paths to obtain the final multipath scattering signal; adding the direct path signal and the multipath scattering signal to obtain the overall channel response; attenuating the semantic feature vector through the overall channel response; and then normalizing it to obtain the final output signal.

2. The end-to-end semantic communication method for fishing vessel networking in a marine environment as described in claim 1, characterized in that, The ocean buoy data is transformed into matrix data according to time granularity and spatial granularity to obtain the ocean buoy data matrix. This includes processing the original ocean buoy data into matrix data according to time granularity and spatial granularity. The time granularity is in hours and the spatial granularity is in the coordinates of different work stations. The data is divided based on three data types: temperature, wave, and wind.

3. The end-to-end semantic communication method for fishing vessel networking in a marine environment as described in claim 1, characterized in that, The preprocessing process includes: normalization, missing value imputation, and outlier handling for each ocean buoy data matrix. Among them, the Min-Max normalization method is used to scale the ocean buoy data matrix to the range of [0,1].

4. The end-to-end semantic communication method for fishing vessel networking in a marine environment as described in claim 3, characterized in that, Transform the normalized ocean buoy data matrix The tensor, obtained from the ocean buoy network data tensor, is input into the semantic communication network for the marine environment. It first enters the semantic encoder, which extracts the spatial patterns or local correlations of the ocean buoy data matrix based on convolutional blocks to obtain a feature map. Then, the output feature map is normalized by a generalized division normalization layer, and then the features of the output feature map are extracted a second time after nonlinear activation of the feature matrix by a PReLU layer. Based on the calculated mean of the features and combined with the signal-to-noise ratio, an adaptive feature block is constructed. The adaptive feature block adjusts the feature map according to the signal-to-noise ratio and re-extracts the feature map. After N extraction processes, a feature vector is obtained, where N is the total number of non-NaN data in the ocean buoy data matrix. Finally, the feature vector is compressed according to the data compression ratio CR to obtain the semantic feature vector.

5. An end-to-end semantic communication system for fishing vessel networking in a marine environment, characterized in that, include: The data acquisition module is used to acquire ocean buoy data, convert the ocean buoy data into matrix data according to time and spatial granularity, obtain the ocean buoy data matrix, and preprocess it; The semantic recovery module is used to convert the preprocessed ocean buoy data matrix into an ocean buoy network data tensor and input it into a semantic communication network for the marine environment to obtain the recovered data of end-to-end semantic communication. In this process, after the ocean buoy network data tensor is input into the semantic communication network for the marine environment, the feature map of the ocean buoy network data tensor is extracted multiple times using convolutional blocks. Then, the feature map is continuously adjusted according to the signal-to-noise ratio by adaptive feature blocks to obtain the feature vector. The feature vector is then compressed according to the data compression ratio to obtain the semantic feature vector. The semantic feature vector is attenuated by the total channel response of the channel model to obtain the final output signal. The final output signal is then used to generate the recovered data through the inverse reasoning process of the semantic decoder. A channel model is constructed based on Channel State Information (CSI). A Ricean channel is selected to simulate the marine environment. The Ricean channel is a fading model, which assumes the existence of a dominant direct path signal and multiple path signals generated by reflection, refraction, or scattering from other objects. The semantic feature vector is attenuated through the overall channel response of the channel model, including: determining the channel gain coefficient based on the signal-to-noise ratio (SNR); simulating signal fading in the Ricean channel; dividing the signal into real and imaginary parts to form a complex signal; calculating the complex form of the direct path signal; performing random time delay processing on the signal of each scattering path; accumulating the signals of all scattering paths to obtain the final multipath scattering signal; adding the direct path signal and the multipath scattering signal to obtain the overall channel response; attenuating the semantic feature vector through the overall channel response; and then normalizing it to obtain the final output signal.

6. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the end-to-end semantic communication method for fishing vessel networking in a marine environment as described in any one of claims 1-4.

7. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium is used to store computer instructions, which, when executed by a processor, implement an end-to-end semantic communication method for fishing vessel networking in a marine environment as described in any one of claims 1-4.

8. An electronic device, characterized in that, include: The device includes a processor, a memory, and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory to enable the electronic device to perform an end-to-end semantic communication method for fishing vessel networking in a marine environment as described in any one of claims 1-4.

Images