Complex sound field distribution prediction method, device and equipment in deep sea environment and medium

Abstract

The invention relates to a complex sound field distribution prediction method and device in a deep sea environment, equipment and a medium. The method comprises the following steps: constructing a sound propagation loss data set in a deep sea environment and taking the data set as a training set; performing Gaussian filtering smoothing processing to obtain global structure information and local detail information of sound propagation loss in the deep sea environment; a double-branch U-Net neural network model is constructed; preprocessing the sound source depth, the sound source frequency and the sound velocity profile in the training set after Gaussian smoothing; and inputting the training set after the sound source depth, the sound source frequency and the sound velocity profile are preprocessed into the model for iterative training, inputting the new unknown sound source depth, unknown sound source frequency and unknown sound velocity profile data into the trained model after the same preprocessing step, and predicting and outputting the sound propagation loss distribution in the corresponding deep sea environment. By adopting the method, efficient and high-precision prediction of sound propagation loss distribution in a complex sound field in a deep sea environment can be realized.

Description

Technical Field

[0001] The present application relates to the field of signal processing technology, and in particular to a method, device, equipment and medium for predicting complex sound field distribution in a deep-sea environment. Background Art

[0002] Sound waves are currently the only form of radiation that can propagate over long distances in seawater, and are also the primary means of underwater long-distance detection and information transmission. In deep-sea environments, sound waves often undergo multipath reflection, refraction, and scattering due to factors such as the undulating seafloor topography, dynamic changes in the seawater sound velocity profile, and sea surface fluctuations, forming obvious and complex interference fringes. For the simulation and prediction of such complex acoustic phenomena, traditional numerical methods usually rely on solving complex partial differential equations. The calculation process is lengthy and resource-intensive, making it difficult to meet the needs of deep-sea long-distance real-time applications.

[0003] With its powerful nonlinear fitting ability, the neural network model provides a feasible way to quickly predict complex acoustic phenomena such as interference fringes and multipath propagation. By training the sound field distribution under a variety of sound source conditions and environmental parameters, the model can significantly shorten the calculation time in the inference stage and better adapt to the dynamic changes of the ocean environment. However, when the existing neural network model copes with the prediction task of sound propagation loss distribution in deep-sea long-distance complex sound fields, due to the poor accuracy of sound propagation loss distribution feature extraction, there is still a problem of low prediction accuracy of sound propagation loss distribution. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, equipment and medium for predicting the distribution of complex sound fields in deep-sea environments in response to the above-mentioned technical problems, so as to achieve efficient and high-precision prediction of the distribution of sound propagation losses in complex sound fields in deep-sea environments.

[0005] A method for predicting complex sound field distribution in a deep sea environment, the method comprising: Generate a sound propagation loss dataset in a deep-sea environment based on different sound source depths, sound source frequencies, and sound velocity profiles, and use the dataset as a training set; The Gaussian filter is used to smooth the sound propagation loss in the deep sea environment in the training set to obtain the global structural information and local detail information of the sound propagation loss in the deep sea environment; A dual-branch U-Net neural network model for predicting sound propagation loss in deep-sea environments is constructed. The model consists of a global feature branch and a local detail branch, both of which are U-Net structures. The model learns the global structural information of the input data through the global feature branch, and extracts the local detail information of the input data through the local detail branch. Finally, by fusing the information of different scales learned by the two branches, the prediction result of sound propagation loss in deep-sea environments that takes into account both global structural information and local detail information is output. The sound source space coding algorithm is used to preprocess the sound source depth, sound source frequency and sound velocity profile in the training set after Gaussian smoothing. The preprocessed sound source depth, sound source frequency and sound velocity profile are encoded into a matrix with the same size as the predicted sound field. The training set of sound source depth, sound source frequency and sound speed profile after preprocessing is input into the double-branch U-Net neural network model for iterative training until a trained double-branch U-Net neural network model is obtained. The new unknown sound source depth, unknown sound source frequency and unknown sound speed profile data are subjected to the same preprocessing steps and then input into the trained double-branch U-Net neural network model to predict the sound propagation loss distribution in the corresponding deep-sea environment.

[0006] In one embodiment, generating a sound propagation loss dataset in a deep sea environment based on different sound source depths, sound source frequencies and sound speed profiles includes: Different sound source depths, sound source frequencies, sound velocity profiles, terrain settings and geoacoustic parameter settings are used as input parameters, and the traditional numerical model Bellhop based on ray theory is used to generate a data set of sound propagation loss in deep sea environment.

[0007] In one embodiment, the conventional numerical model Bellhop based on ray theory is used to generate a sound propagation loss dataset in a deep sea environment, including: Bellhop uses ray theory to simulate the propagation of sound waves in the ocean environment and generates a data set of sound propagation losses in deep-sea environments, including intrinsic sound rays and underwater sound propagation losses. By sound pressure A reference signal at a receiver location relative to the source To calculate, specifically expressed as: ; ; in, represents the number of eigenvalues ​​that contribute to the sound field at a specific receiving position, By The sound pressure caused by the intrinsic sound rays, and , and Respectively represent the horizontal and vertical (ie, depth) coordinates of the receiving position.

[0008] In one embodiment, a sound source spatial coding algorithm is used to pre-process the sound source depth, sound source frequency and sound velocity profile in the training set after Gaussian smoothing, including: The study area Discretize into a zero matrix , the matrix elements represents the zero matrix Line The position of the column in the area The relative position within the study area The horizontal and vertical distances are and , and Respectively represent the number of rows and columns of the matrix; Known sound source , for the set underwater sound source depth , in the depth range Search and The closest vertical index , expressed as: ; in, ; Depth to sound source , sound source frequency Sonic velocity profile Normalization is performed separately; According to the characteristics of cylindrical coordinates, the sound source is set to appear only in the first column of the zero matrix. Therefore, the normalized sound source depth, sound source frequency and sound velocity profile are encoded in the zero matrix respectively. The first column of the vertical index , respectively get the matrix after the sound source depth encoding , the matrix after the sound source frequency encoding The matrix after encoding the sound velocity profile .

[0009] In one embodiment, the depth of the sound source , sound source frequency Sonic velocity profile Normalization is performed separately, including: Within the corresponding definition range, the sound source depth The frequency of the sound source Perform normalization to obtain the normalized sound source depth and the normalized sound source frequency , respectively expressed as: ; ; in, , and Respectively represent the minimum and maximum values ​​of the sound source depth; , and Respectively represent the minimum and maximum values ​​of the sound source frequency; Will Each sound velocity profile sample in the sound velocity profile Perform normalization processing to obtain the normalized sound velocity profile sample , expressed as: ; in, Indicates the serial number of the sound velocity profile sample, is the total number of sound velocity profiles, and They represent the minimum and maximum values ​​of the sound velocity profile respectively.

[0010] In one embodiment, the normalized sound source depth, sound source frequency and sound velocity profile are encoded in a zero matrix The first column of the vertical index , respectively get the matrix after the sound source depth encoding , the matrix after the sound source frequency encoding The matrix after encoding the sound velocity profile , expressed as: ; ; .

[0011] In one embodiment, the total loss function of the dual-branch U-Net neural network model is It is expressed as: ; ; ; in, represents the loss function of the global feature branch, represents the loss function of the local detail branch, and Both represent weights, which are used to adjust the proportion of global feature branch loss and local detail branch loss. Represents the global structural information learned by the global feature branch, The tag value representing the global structure information, Represents the local detail information extracted by the local detail branch, The label value representing the local detail information.

[0012] A device for predicting complex sound field distribution in a deep sea environment, the device comprising: A data set generation module is used to generate a data set of sound propagation loss in a deep-sea environment based on different sound source depths, sound source frequencies and sound speed profiles, and use the data set as a training set; Gaussian smoothing module, used to use Gaussian filter to smooth the sound propagation loss in the deep sea environment in the training set, and obtain the global structure information and local detail information of the sound propagation loss in the deep sea environment; A model building module is used to build a dual-branch U-Net neural network model for predicting sound propagation loss in deep-sea environments. The model consists of a global feature branch and a local detail branch, both of which are U-Net structures. The model learns the global structural information of the input data through the global feature branch, and extracts the local detail information of the input data through the local detail branch. Finally, by fusing the information of different scales learned by the two branches, the prediction result of sound propagation loss in deep-sea environments that takes into account both global structural information and local detail information is output; A data preprocessing module is used to preprocess the sound source depth, sound source frequency and sound speed profile in the training set after Gaussian smoothing by using a sound source space coding algorithm, and the preprocessed sound source depth, sound source frequency and sound speed profile are encoded into a matrix with the same size as the predicted sound field; The model training and prediction module is used to input the preprocessed training set of sound source depth, sound source frequency and sound speed profile into the double-branch U-Net neural network model for iterative training until a trained double-branch U-Net neural network model is obtained, and the new unknown sound source depth, unknown sound source frequency and unknown sound speed profile data are input into the trained double-branch U-Net neural network model after the same preprocessing steps, and the corresponding sound propagation loss distribution in the deep-sea environment is predicted and output.

[0013] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of the above-mentioned method for predicting complex sound field distribution in a deep sea environment.

[0014] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for predicting complex sound field distribution in a deep-sea environment.

[0015] The above-mentioned method, device, equipment and medium for predicting complex sound field distribution in deep sea environment have the following beneficial effects: 1. The sound propagation loss in the deep sea environment is decomposed into global structural information and local detail information using Gaussian filters, which improves the model's ability to learn complex interference fringe patterns, thereby reducing noise while retaining edges and details as much as possible.

[0016] 2. Construct a dual-branch U-Net neural network model to predict sound propagation loss in deep-sea environments. The model fuses the global structure information and local detail information learned by the global feature branch and the local detail branch respectively, and finally outputs the prediction result of sound propagation loss in deep-sea environments that takes into account both global structure information and local detail information, thereby achieving accurate prediction of the distribution of sound propagation loss in deep-sea environments.

[0017] 3. The sound source space coding algorithm is introduced to encode the sound source depth, sound source frequency and sound velocity profile into a matrix with the same size as the predicted sound field. The encoded matrix contains the sound source depth, sound source frequency, sound velocity profile and sound source spatial position information, and the encoded matrix can be stacked or combined in a specific way to form a multi-channel input with rich information, ensuring the consistency of the data format and the input of the dual-branch U-Net neural network model. At the same time, it enables the model to fully learn the relevant characteristics of sound propagation loss in deep-sea environment, thereby improving the accuracy of the prediction of sound propagation loss distribution in deep-sea environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of a flow chart of a method for predicting complex sound field distribution in a deep sea environment in one embodiment; Figure 2 A schematic diagram of generating a sound propagation loss data set in a deep sea environment in one embodiment; Figure 3 Schematic diagram of the architecture of a dual-branch U-Net neural network model in one embodiment; Figure 4 A schematic diagram of model training and prediction in one embodiment; Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0020] In one embodiment, Figure 1 As shown, a method for predicting complex sound field distribution in a deep sea environment is provided, comprising the following steps: Step S1, generating a sound propagation loss data set in a deep sea environment based on different sound source depths, sound source frequencies and sound speed profiles, and using the data set as a training set.

[0021] Among them, due to the lack of underwater sound field observation data, it is difficult to support the effective training of neural networks. This application uses traditional numerical calculation results as training label values. In the simulation of underwater sound fields and acoustic channels, Bellhop is a commonly used numerical calculation model. The Bellhop model is used to generate a data set of sound propagation loss in a deep sea environment, and this data set is used as a training set for model training. Figure 2 As shown in the figure, the process of generating the sound propagation loss dataset in the deep sea environment is as follows: different sound source depths, sound source frequencies, sound velocity profiles, terrain settings, and geoacoustic parameter settings are used as input parameters, and the traditional numerical model Bellhop based on ray theory is used to generate the sound propagation loss dataset in the deep sea environment. Among them, the sound velocity profile comes from the ocean temperature, salinity, and depth dataset, the terrain settings come from the terrain dataset, and the geoacoustic parameter settings come from the geoacoustic dataset.

[0022] Specifically, Bellhop uses ray theory to simulate the propagation of sound waves in the ocean environment and generates a data set of sound propagation losses in the deep sea environment, including intrinsic sound rays and underwater sound propagation losses. Among them, the underwater sound propagation loss reflects the degree of energy attenuation during the sound wave propagation process. By sound pressure A reference signal at a receiver location relative to the source To calculate, specifically expressed as: ; ; in, represents the number of eigenvalues ​​that contribute to the sound field at a specific receiving position, By The complex sound pressure caused by the eigenvalue sound ray describes the The intrinsic sound rays are at the receiving position The intensity and phase information at , and Respectively represent the horizontal and vertical (ie, depth) coordinates of the receiving position.

[0023] Step S2, using a Gaussian filter to smooth the sound propagation loss in the deep sea environment in the training set, to obtain global structural information and local detail information of the sound propagation loss in the deep sea environment.

[0024] Among them, Gaussian filtering is a smoothing technology widely used in image processing and signal processing. It smoothes data by using a filter constructed using a Gaussian function to reduce noise while retaining edges and details as much as possible. Its mathematical expression is: ; in, is the spatial coordinate; is the standard deviation, which is used to control the width of the Gaussian function and determines the smoothness of the filter. Through Gaussian smoothing, the global features of the low-frequency part can be better preserved, while the high-frequency part focuses on the complex interference fringes.

[0025] Step S3, constructing a dual-branch U-Net neural network model for predicting sound propagation loss in a deep-sea environment, the model consists of a global feature branch and a local detail branch, both of which are U-Net structures; the model learns the global structural information of the input data through the global feature branch, and extracts the local detail information of the input data through the local detail branch. Finally, by fusing the information of different scales learned by the two branches, the prediction result of sound propagation loss in a deep-sea environment that takes into account both the global structural information and the local detail information is output.

[0026] Among them, the architecture of the dual-branch U-Net neural network model is as follows Figure 3 As shown in the figure, the U-Net in the architecture adopts a symmetrical autoencoder structure to extract and reconstruct image features by gradually compressing and restoring the spatial resolution of the image.

[0027] Furthermore, the total loss function of the two-branch U-Net neural network model is It is expressed as: ; ; ; in, represents the loss function of the global feature branch, represents the loss function of the local detail branch, and Both represent weights, which are used to adjust the proportion of global feature branch loss and local detail branch loss. Represents the global structural information learned by the global feature branch, The tag value representing the global structure information, Represents the local detail information extracted by the local detail branch, The label value representing the local detail information.

[0028] Step S4, using a sound source space coding algorithm, preprocessing the sound source depth, sound source frequency and sound speed profile in the training set after Gaussian smoothing, and the preprocessed sound source depth, sound source frequency and sound speed profile are encoded into a matrix with the same size as the predicted sound field.

[0029] In step S4, the present application considers that there are limitations in using the sound source depth, sound source frequency and sound speed profile as direct inputs to the neural network model. This input method fails to give full play to the advantages of the convolution kernel in spatial feature extraction. In order to overcome this limitation, the present application further introduces a sound source space coding algorithm, which encodes the sound source depth, sound source frequency and sound speed profile into a matrix with the same size as the predicted sound field, so that the encoded matrix can simultaneously contain the sound source depth, sound source frequency, sound speed profile and sound source spatial position information, thereby fully giving full play to the advantages of the convolution kernel in the dual-branch U-Net neural network model in spatial feature extraction, improving the accuracy of feature extraction related to sound propagation loss in deep-sea environments, and thereby improving the accuracy of model prediction. The sound source space coding algorithm specifically includes the following steps: (1) Regional discretization: Discretize the study area Discretize into a zero matrix , the matrix elements represents the zero matrix Line The position of the column in the area The relative position within the study area The horizontal and vertical distances are and , and Represent the number of rows and columns of the matrix respectively.

[0030] (2) Sound source depth positioning: known sound source , for the set underwater sound source depth , in the depth range Search and The closest vertical index , expressed as: ; in, .

[0031] (3) Normalization of sound source depth, frequency and sound velocity profile: , sound source frequency Sonic velocity profile Specifically, within the corresponding definition range, the sound source depth The frequency of the sound source Perform normalization to obtain the normalized sound source depth and the normalized sound source frequency , respectively expressed as: ; ; in, , and Respectively represent the minimum and maximum values ​​of the sound source depth; , and Respectively represent the minimum and maximum values ​​of the sound source frequency.

[0032] Will Each sound velocity profile sample in the sound velocity profile Perform normalization processing to obtain the normalized sound velocity profile sample , expressed as: ; in, Indicates the serial number of the sound velocity profile sample, is the total number of sound velocity profiles, and They represent the minimum and maximum values ​​of the sound velocity profile respectively.

[0033] (4) Data embedding: According to the characteristics of cylindrical coordinates, the sound source is set to appear only in the first column of the zero matrix. Therefore, the normalized sound source depth, sound source frequency and sound velocity profile are encoded in the zero matrix respectively. The first column of the vertical index , respectively get the matrix after the sound source depth encoding , the matrix after the sound source frequency encoding The matrix after encoding the sound velocity profile , respectively expressed as: ; ; .

[0034] Through the above preprocessing process, the encoding can be obtained , and Stacking or combining in a specific way forms a multi-channel input with rich information, thereby ensuring the consistency of the data format with the input of the dual-branch U-Net neural network model. At the same time, it enables the model to fully learn the relevant characteristics of sound propagation loss in deep-sea environments and improve the accuracy of the prediction of sound propagation loss distribution in deep-sea environments.

[0035] Step S5, input the preprocessed training set of sound source depth, sound source frequency and sound speed profile into the double-branch U-Net neural network model for iterative training until a trained double-branch U-Net neural network model is obtained, and the new unknown sound source depth, unknown sound source frequency and unknown sound speed profile data are subjected to the same preprocessing steps and then input into the trained double-branch U-Net neural network model to predict the sound propagation loss distribution in the corresponding deep-sea environment.

[0036] Specifically, the model training and prediction process is as follows: Figure 4 As shown, including: Training phase: First, initialize the weights and biases of the dual-branch U-Net neural network model. Then input the pre-processed training set of sound source depth, sound source frequency and sound speed profile into the dual-branch U-Net neural network model. Through the forward propagation process, the input data passes through multiple layers of convolutional layers, batch normalization layers and nonlinear activation functions in turn, gradually extracting the global structure and local details, and generating preliminary prediction results of the sound propagation loss distribution in the deep-sea environment. In this process, the jump connection in the U-Net structure is used to fuse features at different levels to ensure that high-resolution details are retained. Subsequently, the error between the predicted result and the true label is calculated, and the total loss function is obtained using weighting. Quantization error.

[0037] The back propagation algorithm is then used to adjust the network weights and biases based on the gradient information of the loss function. The Adam optimizer is used in the optimization process to dynamically adjust the learning rate, accelerate convergence and avoid falling into the local optimum. The entire training process is repeated in multiple iterations, and the training data is batch processed in each iteration to improve computing efficiency.

[0038] Prediction phase: After training, the model parameters of the trained dual-branch U-Net neural network model are saved, and the new unknown sound source depth, unknown sound source frequency, and unknown sound velocity profile data are subjected to the same preprocessing steps and then input into the model. Through forward propagation, the model generates the final prediction results of the sound propagation loss distribution in the deep sea environment. The entire process is implemented in an end-to-end manner without the need for additional manual feature extraction steps, ensuring the consistency of the training and testing phases and the efficiency of the model.

[0039] In summary, the present application proposes a method for predicting the distribution of complex sound fields in deep-sea environments. It takes the preprocessed sound source depth, sound source frequency and sound velocity profile as input, and realizes end-to-end prediction of the distribution of sound propagation loss in complex sound fields in deep-sea environments through a dual-branch U-Net neural network model. Compared with traditional numerical models, the method proposed in the present application can significantly reduce the computational complexity and cost while maintaining the prediction accuracy.

[0040] In one embodiment, a device for predicting complex sound field distribution in a deep sea environment is provided, comprising: A data set generation module is used to generate a data set of sound propagation loss in a deep-sea environment based on different sound source depths, sound source frequencies and sound speed profiles, and use the data set as a training set; Gaussian smoothing module, used to use Gaussian filter to smooth the sound propagation loss in the deep sea environment in the training set, and obtain the global structure information and local detail information of the sound propagation loss in the deep sea environment; A model building module is used to build a dual-branch U-Net neural network model for predicting sound propagation loss in deep-sea environments. The model consists of a global feature branch and a local detail branch, both of which are U-Net structures. The model learns the global structural information of the input data through the global feature branch, and extracts the local detail information of the input data through the local detail branch. Finally, by fusing the information of different scales learned by the two branches, the prediction result of sound propagation loss in deep-sea environments that takes into account both global structural information and local detail information is output; A data preprocessing module is used to preprocess the sound source depth, sound source frequency and sound speed profile in the training set after Gaussian smoothing by using a sound source space coding algorithm, and the preprocessed sound source depth, sound source frequency and sound speed profile are encoded into a matrix with the same size as the predicted sound field; The model training and prediction module is used to input the preprocessed training set of sound source depth, sound source frequency and sound speed profile into the double-branch U-Net neural network model for iterative training until a trained double-branch U-Net neural network model is obtained, and the new unknown sound source depth, unknown sound source frequency and unknown sound speed profile data are input into the trained double-branch U-Net neural network model after the same preprocessing steps, and the corresponding sound propagation loss distribution in the deep-sea environment is predicted and output.

[0041] For the specific limitations of the device for predicting the distribution of complex sound fields in a deep-sea environment, please refer to the limitations of the method for predicting the distribution of complex sound fields in a deep-sea environment mentioned above, which will not be repeated here. Each module in the above-mentioned device for predicting the distribution of complex sound fields in a deep-sea environment can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above-mentioned modules.

[0042] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 5As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting the distribution of complex sound fields in a deep-sea environment is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0043] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0044] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of the above-mentioned method for predicting complex sound field distribution in a deep sea environment.

[0045] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned method for predicting the distribution of complex sound fields in a deep-sea environment are implemented.

[0046] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0047] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0048] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for predicting complex sound field distribution in a deep sea environment, characterized in that: The method comprises: Generate a sound propagation loss dataset in a deep-sea environment based on different sound source depths, sound source frequencies, and sound velocity profiles, and use the dataset as a training set; Using a Gaussian filter to smooth the sound propagation loss in the deep sea environment in the training set, so as to obtain global structural information and local detail information of the sound propagation loss in the deep sea environment; A dual-branch U-Net neural network model for predicting sound propagation loss in deep-sea environments is constructed. The model consists of a global feature branch and a local detail branch, both of which are U-Net structures. The model learns the global structural information of the input data through the global feature branch, and extracts the local detail information of the input data through the local detail branch. Finally, by fusing the information of different scales learned by the two branches, the prediction result of sound propagation loss in deep-sea environments that takes into account both global structural information and local detail information is output. The sound source space coding algorithm is used to preprocess the sound source depth, sound source frequency and sound velocity profile in the training set after Gaussian smoothing. The preprocessed sound source depth, sound source frequency and sound velocity profile are encoded into a matrix with the same size as the predicted sound field. The training set after preprocessing of the sound source depth, sound source frequency and sound speed profile is input into the dual-branch U-Net neural network model for iterative training until a trained dual-branch U-Net neural network model is obtained. The new unknown sound source depth, unknown sound source frequency and unknown sound speed profile data are subjected to the same preprocessing steps and then input into the trained dual-branch U-Net neural network model to predict the sound propagation loss distribution in the corresponding deep-sea environment.

2. The method according to claim 1, characterized in that: Generate a deep-sea sound propagation loss dataset based on different sound source depths, sound source frequencies, and sound velocity profiles, including: Different sound source depths, sound source frequencies, sound velocity profiles, terrain settings and geoacoustic parameter settings are used as input parameters, and the traditional numerical model Bellhop based on ray theory is used to generate a data set of sound propagation loss in deep sea environment.

3. The method according to claim 2, characterized in that Using the traditional numerical model Bellhop based on ray theory, a data set of sound propagation loss in deep sea environment is generated, including: Bellhop uses ray theory to simulate the propagation of sound waves in the ocean environment and generates a data set of sound propagation losses in deep-sea environments, including intrinsic sound rays and underwater sound propagation losses. By sound pressure A reference signal at a receiver location relative to the source To calculate, specifically expressed as: ； ； in, represents the number of eigenvalues ​​that contribute to the sound field at a specific receiving position, By The sound pressure caused by the intrinsic sound rays, and , and Respectively represent the horizontal and vertical coordinates of the receiving position.

4. The method according to claim 1, characterized in that: The sound source spatial coding algorithm is used to preprocess the sound source depth, sound source frequency and sound velocity profile in the training set after Gaussian smoothing, including: The study area Discretize into a zero matrix , the matrix elements Represents the zero matrix Line The position of the column in the area The relative position within the study area The horizontal and vertical distances are and , and Respectively represent the number of rows and columns of the matrix; Known sound source , for the set underwater sound source depth , in the depth range Search and The closest vertical index , expressed as: ； in, ; Depth to sound source , sound source frequency Sonic velocity profile Normalization is performed separately; According to the characteristics of cylindrical coordinates, the sound source is set to appear only in the first column of the zero matrix. Therefore, the normalized sound source depth, sound source frequency and sound velocity profile are encoded in the zero matrix respectively. The first column of the vertical index , respectively get the matrix after the sound source depth encoding , the matrix after the sound source frequency encoding The matrix after encoding the sound velocity profile .

5. The method according to claim 4, characterized in that Depth to sound source , sound source frequency Sonic velocity profile Normalization is performed separately, including: Within the corresponding definition range, the sound source depth The frequency of the sound source Perform normalization to obtain the normalized sound source depth and the normalized sound source frequency , respectively expressed as: ； ； in, , and Respectively represent the minimum and maximum values ​​of the sound source depth; , and Respectively represent the minimum and maximum values ​​of the sound source frequency; Will Each sound velocity profile sample in the sound velocity profile Perform normalization processing to obtain the normalized sound velocity profile sample , expressed as: ； in, Indicates the serial number of the sound velocity profile sample, is the total number of sound velocity profiles, and They represent the minimum and maximum values ​​of the sound velocity profile respectively.

6. The method according to claim 5, characterized in that The normalized sound source depth, sound source frequency and sound velocity profile are encoded in the zero matrix respectively. The first column of the vertical index , respectively get the matrix after the sound source depth encoding , the matrix after the sound source frequency encoding The matrix after encoding the sound velocity profile , expressed as: ； ； 。 7. The method according to claim 1, characterized in that The total loss function of the dual-branch U-Net neural network model It is expressed as: ； ； ； in, represents the loss function of the global feature branch, represents the loss function of the local detail branch, and Both represent weights, which are used to adjust the proportion of global feature branch loss and local detail branch loss. Represents the global structural information learned by the global feature branch, The tag value representing the global structure information, Represents the local detail information extracted by the local detail branch, The label value representing the local detail information.

8. A device for predicting complex sound field distribution in a deep sea environment, characterized in that: The device comprises: A data set generation module is used to generate a data set of sound propagation loss in a deep-sea environment based on different sound source depths, sound source frequencies and sound speed profiles, and use the data set as a training set; A Gaussian smoothing module is used to use a Gaussian filter to smooth the sound propagation loss in the deep sea environment in the training set, so as to obtain global structural information and local detail information of the sound propagation loss in the deep sea environment; A model building module is used to build a dual-branch U-Net neural network model for predicting sound propagation loss in deep-sea environments. The model consists of a global feature branch and a local detail branch, both of which are U-Net structures. The model learns the global structural information of the input data through the global feature branch, and extracts the local detail information of the input data through the local detail branch. Finally, by fusing the information of different scales learned by the two branches, the prediction result of sound propagation loss in deep-sea environments that takes into account both global structural information and local detail information is output; A data preprocessing module is used to preprocess the sound source depth, sound source frequency and sound speed profile in the training set after Gaussian smoothing by using a sound source space coding algorithm, and the preprocessed sound source depth, sound source frequency and sound speed profile are encoded into a matrix with the same size as the predicted sound field; The model training and prediction module is used to input the training set of sound source depth, sound source frequency and sound speed profile after preprocessing into the double-branch U-Net neural network model for iterative training until a trained double-branch U-Net neural network model is obtained, and the new unknown sound source depth, unknown sound source frequency and unknown sound speed profile data are subjected to the same preprocessing steps and then input into the trained double-branch U-Net neural network model to predict the sound propagation loss distribution in the corresponding deep-sea environment.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images