Acoustic imaging microphone array optimization method and system based on graph neural network, and related device
By using a graph neural network-based acoustic imaging microphone array optimization method, the problems of large microphone array size and high power consumption are solved, achieving lightweighting and energy consumption optimization of the microphone array, thereby improving imaging efficiency and user experience.
Patent Information
- Application Number
- CN202511330109.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-17
- Publication Date
- 2026-02-10
AI Technical Summary
Microphone arrays in existing acoustic imaging detection instruments suffer from problems such as large size, high power consumption, and inconvenient deployment.
An acoustic imaging microphone array optimization method based on graph neural networks is adopted. By defining the sound source frequency and microphone layout as input variables, the sound field intensity distribution is obtained based on the beamforming algorithm. The acoustic imaging surrogate model is trained using graph neural networks, and the optimal arrangement of microphone coordinates is found under given imaging accuracy constraints by combining intelligent optimization algorithms.
While ensuring image quality, the microphone array was made lightweight and energy consumption was optimized, improving the imaging efficiency and user experience of the microphone array.
Smart Images
Figure CN121502838A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of acoustic imaging technology, and in particular to an acoustic imaging microphone array optimization method and system based on graph neural networks. Background Technology
[0002] Acoustic imaging technology is widely used in noise source localization, fault detection, and industrial inspection. Its core lies in the collaborative design of microphone array structure and beamforming algorithms. Traditional designs typically use a large number of uniformly arranged microphones to achieve sufficient resolution and sensitivity, leading to problems such as large size and high power consumption. In recent years, artificial intelligence methods have begun to show promise in acoustic reconstruction and optimization problems; however, introducing Graph Neural Networks (GNNs) into microphone array optimization is still an innovative approach with significant research and application value.
[0003] Microphone arrays in existing acoustic imaging detection instruments suffer from problems such as large size, high power consumption, and inconvenient deployment. Summary of the Invention
[0004] This application provides a method, system, and related apparatus for optimizing acoustic imaging microphone arrays based on graph neural networks, in order to improve the above-mentioned problems.
[0005] To achieve the above objectives, this application adopts the following technical solution:
[0006] In a first aspect, embodiments of this application propose an acoustic imaging microphone array optimization method based on graph neural networks, the method comprising:
[0007] Define the sound source frequency and microphone layout as input variables, wherein the microphone layout includes the distance between the sound source and the centroid of the microphone array, and the angle between the line connecting the sound source and the centroid and the microphone array;
[0008] Simulations were performed based on different microphone layouts, and sound field intensity distribution was obtained based on beamforming algorithms to generate training data.
[0009] Based on graph neural networks, an acoustic imaging proxy model is trained with microphone array graph structure and sound source characteristics as input and sound field intensity map as output.
[0010] The accuracy of the proxy model was evaluated using a validation set.
[0011] Based on the surrogate model, an intelligent optimization algorithm is introduced to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints.
[0012] In conjunction with the first aspect, in some implementations, simulations are performed based on different microphone layouts, and sound field intensity distributions are obtained based on beamforming algorithms to generate training data, including:
[0013] The sound intensity in a specific direction is calculated using a beamforming algorithm, and the following conditions are met:
[0014]
[0015] Where, x i (t) represents the signal received by the i-th microphone; w i is the weight; r is the location of the observation point; c is the speed of sound in the environment.
[0016] In conjunction with the first aspect, in some implementations, an acoustic imaging surrogate model is trained based on a graph neural network, using the microphone array graph structure and sound source characteristics as input and the sound field intensity map as output, including:
[0017] Input graph features, including: microphone node position m i The relative position of the sound source and the frequency of the sound source, where the relative position of the sound source includes the direction of the sound source and the relative distance between the sound source and the microphone array;
[0018] Edges are constructed using the radius-neighbor method, forming an adjacency matrix A;
[0019] Extract the spatial relationship features between microphones and use them in conjunction with an MLP (Multilayer Perceptron) to predict the sound field output;
[0020] Reconstruct the sound field intensity P(r) at each location on a two-dimensional plane.
[0021] In conjunction with the first aspect, in some implementations, an intelligent optimization algorithm is introduced based on a surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints, including:
[0022] The two-dimensional coordinates are discretized into an array of feasible point grids and encoded as node attributes of the structure graph.
[0023] The acoustic intensity error in the imaging region is calculated based on the surrogate model, and the following conditions are met:
[0024] Fitness=∫r|I true (r)-I pred (r)| 2 dr+λ·N
[0025] Where N is the number of microphones, λ is a natural number used to balance control accuracy and array simplification, Itrue(r) is the true sound intensity, and Ipred(r) is the predicted sound intensity.
[0026] In conjunction with the first aspect, in some implementations, an intelligent optimization algorithm is introduced based on a surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints, including:
[0027] The optimal arrangement is verified to ensure that the accuracy of acoustic imaging meets the preset requirements under the optimal arrangement.
[0028] If the requirements are not met, the intelligent optimization algorithm will be re-performed until a microphone coordinate arrangement that meets the accuracy requirements is found.
[0029] In conjunction with the first aspect, in some implementations, the intelligent optimization algorithm employs a genetic algorithm to achieve global optimization of microphone coordinates, including:
[0030] The placement state of each microphone in the discretized grid is encoded into a binary gene sequence to form a chromosome representing the complete array layout;
[0031] Randomly generate N initial array layouts that satisfy the minimum microphone spacing constraint;
[0032] The acoustic field intensity map of each individual in the population is calculated using a surrogate model.
[0033] In conjunction with the first aspect, in some implementations, the placement state of each microphone in a discretized grid is encoded as a binary gene sequence to form a chromosome representing the complete array layout, including:
[0034] Establish a Cartesian coordinate system for the two-dimensional discretized grid, and define the grid point coordinates (u,v) where u∈[1,P],v∈[1,Q], and P×Q constitutes the total number of grids;
[0035] Assign a unique gene location index k to each grid point, satisfying k = (v-1) × P + u.
[0036] Secondly, embodiments of this application propose an acoustic imaging microphone array optimization system based on graph neural networks, which is configured as follows:
[0037] Define the sound source frequency and microphone layout as input variables, wherein the microphone layout includes the distance between the sound source and the centroid of the microphone array, and the angle between the line connecting the sound source and the centroid and the microphone array;
[0038] Simulations were performed based on different microphone layouts, and sound field intensity distribution was obtained based on beamforming algorithms to generate training data.
[0039] Based on graph neural networks, an acoustic imaging proxy model is trained with microphone array graph structure and sound source characteristics as input and sound field intensity map as output.
[0040] The accuracy of the proxy model was evaluated using a validation set.
[0041] Based on the surrogate model, an intelligent optimization algorithm is introduced to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints.
[0042] In conjunction with the second aspect, in some implementations, the system is configured as follows:
[0043] Simulations were performed based on different microphone layouts, and sound field intensity distribution was obtained using beamforming algorithms to generate training data, including:
[0044] The sound intensity in a specific direction is calculated using a beamforming algorithm, and the following conditions are met:
[0045]
[0046] Where, x i (t) represents the signal received by the i-th microphone; w i is the weight; r is the location of the observation point.
[0047] In conjunction with the second aspect, in some implementations, the system is configured to: train an acoustic imaging surrogate model based on a graph neural network, using the microphone array graph structure and sound source characteristics as input and a sound field intensity map as output, including:
[0048] Input graph features, including: microphone node position m i The relative position of the sound source and the frequency of the sound source, wherein the relative position of the sound source includes the direction of the sound source and the relative distance between the sound source and the microphone array;
[0049] Edges are constructed using the radius neighbor method, forming an adjacency matrix A;
[0050] Extract the spatial relationship features between microphones and use them in conjunction with MLP to predict the sound field output;
[0051] Reconstruct the sound field intensity P(r) at each location on a two-dimensional plane.
[0052] In conjunction with the second aspect, in some implementations, the system is configured to: introduce an intelligent optimization algorithm based on a surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints, including:
[0053] The two-dimensional coordinates are discretized into an array of feasible point grids and encoded as node attributes of the structure graph.
[0054] The acoustic intensity error in the imaging region is calculated based on the surrogate model, and the following conditions are met:
[0055] Fitness=∫ r |I true (r)-I pred (r)| 2 dr+λ·N
[0056] Where N is the number of microphones and λ is a natural number used to balance the control of precision and array simplification.
[0057] In conjunction with the second aspect, in some implementations, the system is configured to: introduce an intelligent optimization algorithm based on a surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints, including:
[0058] The optimal arrangement is verified to ensure that the accuracy of acoustic imaging meets the preset requirements under the optimal arrangement.
[0059] If the requirements are not met, the intelligent optimization algorithm will be re-performed until a microphone coordinate arrangement that meets the accuracy requirements is found.
[0060] In conjunction with the second aspect, in some implementations, the system is configured as follows:
[0061] The intelligent optimization algorithm uses a genetic algorithm to achieve global optimization of microphone coordinates, including:
[0062] The placement state of each microphone in the discretized grid is encoded into a binary gene sequence to form a chromosome representing the complete array layout;
[0063] Randomly generate N initial array layouts that satisfy the minimum microphone spacing constraint;
[0064] The acoustic field intensity map of each individual in the population is calculated using a surrogate model.
[0065] In conjunction with the second aspect, in some implementations, the system is configured to: encode the placement state of each microphone in a discretized grid into a binary gene sequence, forming a chromosome representing the complete array layout, including:
[0066] Establish a Cartesian coordinate system for the two-dimensional discretized grid, and define the grid point coordinates (u,v) where u∈[1,P],v∈[1,Q], and P×Q constitutes the total number of grids;
[0067] Assign a unique gene location index k to each grid point, satisfying k = (v-1) × P + u.
[0068] A third aspect of this invention provides an electronic device, which includes:
[0069] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method proposed in the first aspect of the present invention.
[0070] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention.
[0071] In summary, the above methods and systems have the following technical effects:
[0072] This invention proposes an acoustic imaging microphone array optimization method based on graph neural networks. First, sound source location, sound source frequency, and microphone layout are defined as input variables. Then, simulations are performed based on different microphone layouts, and sound field intensity distribution is obtained using a beamforming algorithm to generate training data. Next, an acoustic imaging surrogate model is trained based on a graph neural network, using the microphone array diagram and sound source characteristics as input and the sound field intensity map as output. The accuracy of the surrogate model is then evaluated using a validation set. Finally, an intelligent optimization algorithm is introduced based on the surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints. This invention proposes a microphone array optimization method based on graph neural networks and intelligent optimization algorithms, achieving lightweighting and energy consumption optimization of the microphone array structure while ensuring imaging quality. Attached Figure Description
[0073] Figure 1 This is a flowchart illustrating an acoustic imaging microphone array optimization method based on graph neural networks proposed in an embodiment of this application. Detailed Implementation
[0074] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0075] This application proposes an acoustic imaging microphone array optimization method based on graph neural networks. Please refer to [link to relevant documentation]. Figure 1 This includes the following steps:
[0076] S101: Define the sound source frequency and microphone layout as input variables, wherein the microphone layout includes the distance between the sound source and the centroid of the microphone array, and the angle between the line connecting the sound source and the centroid and the microphone array.
[0077] In this study, sound source location, sound source frequency, and microphone layout were considered as input variables. Specifically, sound source location refers to the specific coordinates of the sound source in space, including the coordinates of the sound source in the x, y, and z directions in three-dimensional space. Sound source frequency refers to the frequency of the sound emitted by the sound source, usually measured in Hertz (Hz), reflecting the speed of the sound source's vibration.
[0078] Furthermore, microphone layout refers to the configuration of the microphone array. This layout includes the distance between the sound source and the centroid of the microphone array, as well as the angle between the line connecting the sound source and the centroid and the microphone array. The relative distance between the sound source and the microphone array refers to the distance from the sound source to the center point of the microphone array; this distance can be a straight-line distance or an actual path distance. The microphone coordinates within the microphone array refer to the position coordinates of each microphone in three-dimensional space, typically including the coordinate values of each microphone in the x, y, and z directions. These input variables allow for more precise localization and frequency analysis of the sound source, thereby achieving more efficient acoustic signal processing.
[0079] For example, in this application, the microphone coordinates form an undirected graph G = (V, E), where node V represents the microphone position and edge E represents the connection relationship; the sound source position is: s = (x s ,y s ,z s Distance: d = ||sm i ||; Frequency: f∈[f min ,f max Wave speed: c, the speed of sound in the environment (usually 343 m / s).
[0080] S102: Simulations are performed based on different microphone layouts, and the sound field intensity distribution is obtained based on beamforming algorithms to generate training data.
[0081] This study employs various microphone layouts for simulation analysis. Detailed simulations of each layout allow for the evaluation of its performance under different environments. Specifically, beamforming algorithms will be used to process and analyze the audio signals received by each microphone array. Beamforming is an effective signal processing technique that can enhance and suppress sound from specific directions by adjusting the phase and gain of each microphone, thereby obtaining clearer and more accurate sound field information.
[0082] After processing with beamforming algorithms, a detailed distribution of sound field intensity can be obtained. This distribution map can intuitively show key information such as the location and intensity of the sound source in space.
[0083] In summary, through simulations based on different microphone layouts and in-depth research on beamforming algorithms, high-quality training data can be generated, laying a solid foundation for the development of acoustic signal processing technology.
[0084] For example, in this embodiment, a beamforming algorithm is used to calculate the sound intensity in a specific direction, and the following conditions are met:
[0085]
[0086] Where, x i (t) represents the signal received by the i-th microphone; w i is the weight; r is the location of the observation point.
[0087] Understandably, the beamforming part doesn't offer much innovation; it uses traditional beamforming algorithms to obtain the sound intensity cloud distribution from a microphone array. The innovation of this invention lies in the construction of the graph neural network and its optimization for the microphone array. Traditional patents use neural networks (and not even graph neural networks; CNNs and LSTMs are more common) for speech recognition, while this invention uses a neural network (GNN) for acoustic localization, further combined with a genetic algorithm for microphone array optimization.
[0088] S103: Based on a graph neural network, an acoustic imaging surrogate model is trained with the microphone array graph structure and sound source characteristics as input and the sound field intensity map as output.
[0089] Understandably, traditional GNNs are often used for tasks such as social networks, molecular structures, and traffic maps. However, in this application, microphone arrays inherently possess spatial topological properties, making them naturally suitable for graph modeling. Nodes represent microphones, and edges represent their spatial connectivity and signal coordination relationships. This "spatial-acoustic" graph model is tailor-made for imaging tasks. Unlike traditional GNN tasks, the input features of each node in this application include not only location coordinates but also: the direction of the sound source, the distance to the sound source, and the dominant frequency of the ideal sound source used to optimize the microphone array.
[0090] Specifically, in this embodiment, the input graph features include: microphone node position m. i The relative position and frequency of the sound source are determined, where the relative position includes the direction of the sound source and the relative distance between the sound source and the microphone array. Edges are constructed using a radius-neighbor method to form an adjacency matrix A; spatial relationship features between microphones are extracted and used in conjunction with an MLP to predict the sound field output; the sound field intensity I(r) at each location is reconstructed on a two-dimensional plane. The loss function designed in this application comprehensively considers the structural similarity to the simulated sound field map (using image similarity metrics such as SSIM) and the difference from the actual measured values.
[0091] S104: Evaluate the accuracy of the surrogate model using the validation set.
[0092] Evaluating the accuracy of a surrogate model using a validation set ensures that its performance in real-world applications meets expectations. A validation set is a dataset independent of the training and test sets, used to evaluate the model's performance during training. By testing the model on the validation set, its performance on unseen data can be observed, thus determining its generalization ability. This method helps to fine-tune the model's parameters and structure to achieve optimal prediction accuracy. Through repeated validation and tuning, the surrogate model can be ensured to accurately complete tasks in real-world applications, improving the overall system's reliability and efficiency.
[0093] S105: Based on the surrogate model, an intelligent optimization algorithm is introduced to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints.
[0094] Given imaging accuracy constraints, by introducing intelligent optimization algorithms combined with surrogate models, the optimal arrangement of microphone coordinates can be effectively found. This method leverages the rapid computational power of the surrogate model, enabling the evaluation of the performance of different microphone arrangements in a short time, thus significantly improving the efficiency of the optimization process. Intelligent optimization algorithms, such as genetic algorithms, particle swarm optimization, or simulated annealing, can find the optimal microphone layout in a broad search space, ensuring optimal acoustic performance while meeting imaging accuracy requirements. This approach not only improves the imaging quality of microphone arrays but also achieves higher efficiency and a better user experience in practical applications.
[0095] For example, in this embodiment, the placement state of each microphone in the discretized grid is encoded as a binary gene sequence. A Cartesian coordinate system is established for the two-dimensional discretized grid, and the grid point coordinates (u,v) are defined, where u∈[1,P], v∈[1,Q], and P×Q constitutes the total number of grids. A unique gene position index k is assigned to each grid point, satisfying k=(v-1)×P+u. Chromosomes representing the complete array layout are formed. N sets of initial array layouts that satisfy the minimum microphone spacing constraint are randomly generated. The sound field intensity map of each individual in the population is calculated through a surrogate model.
[0096] Specifically, the two-dimensional coordinates are discretized into an array of feasible point grids and encoded as node attributes of the structure graph;
[0097] The acoustic intensity error in the imaging region is calculated based on the surrogate model, and the following conditions are met:
[0098] Fitness=∫ r |I true (r)-I pred (r)| 2 dr+λ·N
[0099] Where N is the number of microphones, λ is a natural number used to control the trade-off between accuracy and array simplification, Itrue(r) is the true sound intensity, and Ipred(r) is the predicted sound intensity.
[0100] As is understandable, an individual is a combination of microphone coordinate configurations. Unlike traditional binary / real number encoding, this application discretizes two-dimensional / three-dimensional coordinates into an array of feasible point grids, encoding them as structural graph node attributes, which facilitates subsequent...
[0101] The GNN model is graph-structure compatible; a multi-objective optimization mechanism (such as NSGA-II) is incorporated to handle the trade-offs between "imaging quality, minimum number of microphones, and minimum microphone spacing". Constraints on the practical feasibility of acoustic engineering, such as array boundaries and minimum microphone spacing, are set.
[0102] This application proposes an acoustic imaging microphone array optimization method based on graph neural networks. First, sound source location, sound source frequency, and microphone layout are defined as input variables. Then, simulations are performed based on different microphone layouts, and sound field intensity distribution is obtained using a beamforming algorithm to generate training data. Next, an acoustic imaging surrogate model is trained based on a graph neural network, using the microphone array diagram structure and sound source characteristics as input and the sound field intensity map as output. The accuracy of the surrogate model is then evaluated using a validation set. Finally, an intelligent optimization algorithm is introduced based on the surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints. This invention proposes a microphone array optimization method based on graph neural networks and intelligent optimization algorithms, achieving lightweighting and energy consumption optimization of the microphone array structure while ensuring imaging quality.
[0103] Based on the same inventive concept, this application also proposes an acoustic imaging microphone array optimization system based on graph neural networks, which is configured as follows:
[0104] Define the sound source frequency and microphone layout as input variables, wherein the microphone layout includes the distance between the sound source and the centroid of the microphone array, and the angle between the line connecting the sound source and the centroid and the microphone array;
[0105] Simulations were performed based on different microphone layouts, and sound field intensity distribution was obtained based on beamforming algorithms to generate training data.
[0106] Based on graph neural networks, an acoustic imaging proxy model is trained with microphone array graph structure and sound source characteristics as input and sound field intensity map as output.
[0107] The accuracy of the proxy model was evaluated using a validation set.
[0108] Based on the surrogate model, an intelligent optimization algorithm is introduced to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints.
[0109] In some implementations, the system is configured as follows:
[0110] Simulations were performed based on different microphone layouts, and sound field intensity distribution was obtained using beamforming algorithms to generate training data, including:
[0111] The sound intensity in a specific direction is calculated using a beamforming algorithm, and the following conditions are met:
[0112]
[0113] Where, x i (t) represents the signal received by the i-th microphone; w i is the weight; r is the location of the observation point.
[0114] In some implementations, the system is configured to: train an acoustic imaging surrogate model based on a graph neural network, using the microphone array graph structure and sound source characteristics as input and a sound field intensity map as output, including:
[0115] Input graph features, including: microphone node position m i The relative position of the sound source and the frequency of the sound source, wherein the relative position of the sound source includes the direction of the sound source and the relative distance between the sound source and the microphone array;
[0116] Edges are constructed using the radius neighbor method, forming an adjacency matrix A;
[0117] Extract the spatial relationship features between microphones and use them in conjunction with MLP to predict the sound field output;
[0118] Reconstruct the sound field intensity P(r) at each location on a two-dimensional plane.
[0119] In some implementations, the system is configured to: introduce an intelligent optimization algorithm based on a surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints, including:
[0120] The two-dimensional coordinates are discretized into an array of feasible point grids and encoded as node attributes of the structure graph.
[0121] The acoustic intensity error in the imaging region is calculated based on the surrogate model, and the following conditions are met:
[0122] Fitness=∫ r |I true (r)-I pred (r)| 2 dr+λ·N
[0123] In some implementations, the system is configured to: introduce an intelligent optimization algorithm based on a surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints, including:
[0124] The optimal arrangement is verified to ensure that the accuracy of acoustic imaging meets the preset requirements under the optimal arrangement.
[0125] If the requirements are not met, the intelligent optimization algorithm will be re-performed until a microphone coordinate arrangement that meets the accuracy requirements is found.
[0126] In some implementations, the system is configured as follows:
[0127] The intelligent optimization algorithm uses a genetic algorithm to achieve global optimization of microphone coordinates, including:
[0128] The placement state of each microphone in the discretized grid is encoded into a binary gene sequence to form a chromosome representing the complete array layout;
[0129] Randomly generate N initial array layouts that satisfy the minimum microphone spacing constraint;
[0130] The acoustic field intensity map of each individual in the population is calculated using a surrogate model.
[0131] In some implementations, the system is configured to: encode the placement state of each microphone in a discretized grid as a binary gene sequence, forming a chromosome representing the complete array layout, including:
[0132] Establish a Cartesian coordinate system for the two-dimensional discretized grid, and define the grid point coordinates (u,v) where u∈[1,P],v∈[1,Q], and P×Q constitutes the total number of grids;
[0133] Assign a unique gene location index k to each grid point, satisfying k = (v-1) × P + u.
[0134] This application proposes an acoustic imaging microphone array optimization system based on graph neural networks. First, sound source location, sound source frequency, and microphone layout are defined as input variables. Then, simulations are performed based on different microphone layouts, and sound field intensity distribution is obtained using a beamforming algorithm to generate training data. Next, based on a graph neural network, an acoustic imaging surrogate model is trained using the microphone array diagram structure and sound source characteristics as input and the sound field intensity map as output. The accuracy of the surrogate model is then evaluated using a validation set. Finally, an intelligent optimization algorithm is introduced based on the surrogate model to find the optimal arrangement of microphone coordinates under given imaging accuracy constraints. This invention proposes a microphone array optimization system based on graph neural networks and intelligent optimization algorithms, achieving lightweighting and energy consumption optimization of the microphone array structure while ensuring imaging quality.
[0135] Based on the same inventive concept, embodiments of this application also propose an electronic device, which includes:
[0136] At least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the graph neural network-based acoustic imaging microphone array optimization method of the present application embodiments.
[0137] Furthermore, to achieve the above objectives, embodiments of this application also propose a computer-readable storage medium storing a computer program that, when executed by a processor, implements the graph neural network-based acoustic imaging microphone array optimization method of this application.
[0138] The following is a detailed introduction to the various components of the electronic device:
[0139] In this context, the processor is the control center of the electronic device. It can be a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).
[0140] Alternatively, the processor can perform various functions of the electronic device by running or executing software programs stored in memory and by calling data stored in memory.
[0141] The memory is used to store the software program that executes the solution of the present invention, and the execution is controlled by the processor. The specific implementation method can be referred to the above method embodiment, which will not be repeated here.
[0142] Optionally, the memory can be read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory can be integrated with the processor or exist independently and coupled to the processor through an interface circuit of an electronic device; this embodiment of the invention does not specifically limit this.
[0143] A transceiver is used to communicate with network devices or with terminal devices.
[0144] Optionally, the transceiver may include a receiver and a transmitter. The receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.
[0145] Optionally, the transceiver can be integrated with the processor or exist independently and coupled to the processor through the router's interface circuit. This embodiment of the invention does not specifically limit this.
[0146] Furthermore, the technical effects of the electronic device can be referred to the technical effects of the data transmission method in the above method embodiments, and will not be repeated here.
[0147] It should be understood that the processor in the embodiments of the present invention can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.
[0148] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).
[0149] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.
[0150] It should be understood that the term "and / or" in this article 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, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.
[0151] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.
[0152] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0153] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
Claims
1. A method for optimizing acoustic imaging microphone arrays based on graph neural networks, characterized in that, The method includes: Define the sound source frequency and microphone layout as input variables, wherein the microphone layout includes the distance between the sound source and the centroid of the microphone array, and the angle between the line connecting the sound source and the centroid and the microphone array; Simulations were performed based on different microphone layouts, and sound field intensity distribution was obtained based on beamforming algorithms to generate training data. Based on graph neural networks, an acoustic imaging proxy model is trained with the array graph structure and sound source characteristics of the microphone as input and the sound field intensity map as output. The accuracy of the proxy model was evaluated using a validation set. An intelligent optimization algorithm is introduced based on a surrogate model to find the optimal arrangement of the microphone coordinates under given imaging accuracy constraints.
2. The method for optimizing an acoustic imaging microphone array based on a graph neural network according to claim 1, characterized in that, Simulations were performed based on different microphone layouts, and sound field intensity distributions were obtained using beamforming algorithms to generate training data, including: The sound intensity in a specific direction is calculated using a beamforming algorithm, and the following conditions are met: Where, x i (t) represents the signal received by the i-th microphone; w i is the weight; r is the location of the observation point.
3. The method for optimizing an acoustic imaging microphone array based on a graph neural network according to claim 1, characterized in that, Based on a graph neural network, using the microphone array graph structure and sound source characteristics as input and the sound field intensity map as output, an acoustic imaging proxy model is trained, including: Input graph features, the graph features including: microphone node position m i The relative position of the sound source and the frequency of the sound source, wherein the relative position of the sound source includes the direction of the sound source and the relative distance between the sound source and the microphone array; Edges are constructed using the radius neighbor method, forming an adjacency matrix A; Extract the spatial relationship features between microphones and use them in conjunction with MLP to predict the sound field output; Reconstruct the sound field intensity P(r) at each location on a two-dimensional plane.
4. The method for optimizing an acoustic imaging microphone array based on a graph neural network according to claim 3, characterized in that, Based on a surrogate model, an intelligent optimization algorithm is introduced to find the optimal arrangement of the microphone coordinates under given imaging accuracy constraints, including: The two-dimensional coordinates are discretized into an array of feasible point grids and encoded as node attributes of the structure graph. The acoustic intensity error in the imaging region is calculated based on the surrogate model, and the following conditions are met: Fitness=∫ r |I true (r)-I pred (r)| 2 dr+λ·N Where N is the number of microphones, and λ is a natural number used to control the trade-off between accuracy and array simplification.
5. The method for optimizing an acoustic imaging microphone array based on a graph neural network according to claim 4, characterized in that, Based on a surrogate model, an intelligent optimization algorithm is introduced to find the optimal arrangement of the microphone coordinates under given imaging accuracy constraints, including: The optimal arrangement is verified to ensure that the accuracy of acoustic imaging meets the preset requirements under the optimal arrangement; If the requirements are not met, the intelligent optimization algorithm will be re-performed until a microphone coordinate arrangement that meets the accuracy requirements is found.
6. The method for optimizing an acoustic imaging microphone array based on a graph neural network according to claim 1, characterized in that, The intelligent optimization algorithm uses a genetic algorithm to achieve global optimization of microphone coordinates, including: The placement state of each microphone in the discretized grid is encoded into a binary gene sequence to form a chromosome representing the complete array layout; Randomly generate N initial array layouts that satisfy the minimum microphone spacing constraint; The acoustic field intensity map of each individual in the population is calculated using a surrogate model.
7. The method for optimizing an acoustic imaging microphone array based on a graph neural network according to claim 6, characterized in that, The placement state of each microphone in the discretized grid is encoded as a binary gene sequence, forming a chromosome representing the complete array layout, including: Establish a Cartesian coordinate system for the two-dimensional discretized grid, and define the grid point coordinates (u,v) where u∈[1,P],v∈[1,Q], and P×Q constitutes the total number of grids; Assign a unique gene location index k to each grid point, satisfying k = (v-1) × P + u.
8. An acoustic imaging microphone array optimization system based on graph neural networks, characterized in that, The system is configured as follows: Define the sound source frequency and microphone layout as input variables, wherein the microphone layout includes the distance between the sound source and the centroid of the microphone array, and the angle between the line connecting the sound source and the centroid and the microphone array; Simulations were performed based on different microphone layouts, and sound field intensity distribution was obtained based on beamforming algorithms to generate training data. Based on a graph neural network, an acoustic imaging proxy model is trained with the microphone array graph structure and sound source characteristics as input and the sound field intensity map as output. The accuracy of the proxy model was evaluated using a validation set. An intelligent optimization algorithm is introduced based on a surrogate model to find the optimal arrangement of the microphone coordinates under given imaging accuracy constraints.
9. An electronic device, characterized in that, include: At least one processor; And, a memory communicatively connected to at least one of the processors; The memory stores instructions that can be executed by at least one of the processors, which, when executed by at least one of the processors, enable the at least one of the processors to perform the method as claimed in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the method as claimed in any one of claims 1-7.