Acoustic wave environment imaging method based on acoustic array

Abstract

The application relates to the technical field of acoustic wave imaging, and discloses an acoustic wave environment imaging method based on an acoustic array. The method comprises the following steps: collecting multi-channel acoustic wave signals in a target space through a distributed acoustic array and extracting time-frequency domain characteristic parameters; constructing a space topology model containing a geometric constraint relationship between an acoustic wave reflection surface and an obstacle according to the parameters; calculating an acoustic wave energy attenuation gradient based on the model, dividing a dynamic partition of acoustic wave energy distribution, performing phase compensation on acoustic wave signals in the dynamic partition, and generating a compensated acoustic wave signal sequence; reconstructing a three-dimensional density field of acoustic wave energy in the target space and marking a non-continuous boundary; and finally, fusing the time-frequency domain characteristic parameters and the three-dimensional density field to generate a multi-modal imaging atlas of the acoustic wave environment. Through accurate processing of multiple links, the method realizes all-around and stereoscopic depiction of the acoustic wave environment, and provides a new technical path for fine perception of the acoustic wave environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of acoustic wave imaging, in particular to an acoustic wave environment imaging method based on an acoustic array. BACKGROUND

[0002] As a non-contact environmental perception means, acoustic wave imaging technology has been widely used in indoor space monitoring, industrial environment detection, security warning and other fields due to its strong adaptability to environment and controllable cost. The core logic is to reverse the environmental structure and energy distribution state in the target space through the collection and analysis of acoustic wave signals, and to provide intuitive basis for subsequent environmental evaluation and decision-making.

[0003] However, there are still many problems to be solved in the existing acoustic wave imaging technology. In the signal collection link, the traditional technology mostly uses single-point or small-scale centralized sensor array, which is difficult to fully capture the spatial distribution characteristics of acoustic wave signals in the target space, resulting in the lack of extracted time-frequency domain information and the inability to fully reflect the propagation law of acoustic waves. In the space modeling stage, the existing method often ignores the geometric constraint relationship between the acoustic wave reflection surface and the obstacle, and the constructed propagation path model deviates from the actual environment, which further affects the accuracy of subsequent energy calculation.

[0004] In terms of energy distribution description, the traditional technology mostly uses static partitioning to divide the acoustic wave energy range, which cannot adapt to the dynamic change characteristics of acoustic wave signals, resulting in lagging and limitation in energy distribution description. At the same time, the phase distortion of acoustic wave is easily affected by the propagation path in the propagation process, and the existing technology lacks a targeted compensation mechanism, which greatly reduces the accuracy of subsequent signal analysis and reconstruction. In the three-dimensional imaging link, the reconstruction of acoustic energy density field in the existing method mostly stays at the rough estimation level, which is difficult to accurately mark the discontinuous boundary and cannot clearly distinguish the key structures such as obstacles and reflection surfaces. In addition, the existing imaging technology often relies only on single-dimensional signal features to generate a map, lacks the fusion of multi-modal information, and the information dimension of the imaging result is single, which cannot fully meet the environmental perception needs in complex scenarios. SUMMARY

[0005] The purpose of the present application is to provide an acoustic wave environment imaging method based on an acoustic array to solve the problems raised in the background art.

[0006] To achieve the above purpose, the present application provides an acoustic wave environment imaging method based on an acoustic array, which comprises:

[0007] Collecting multi-channel acoustic wave signals in the target space through a distributed acoustic array, and extracting time-frequency domain feature parameters in the acoustic wave signals;

[0008] construct a spatial topology model of the sound wave propagation path according to the time-frequency domain feature parameters, the spatial topology model comprising geometric constraint relationships between sound wave reflection surfaces and obstacles;

[0009] calculate an energy attenuation gradient of the sound wave in the target space based on the spatial topology model, and divide a dynamic partition of the sound wave energy distribution;

[0010] perform phase compensation on the sound wave signals in the dynamic partition to generate a compensated sound wave signal sequence;

[0011] reconstruct a three-dimensional density field of the sound wave energy in the target space according to the compensated sound wave signal sequence, and mark a non-continuous boundary of the three-dimensional density field;

[0012] fuse the time-frequency domain feature parameters of the multi-channel sound wave signals and the three-dimensional density field to generate a multi-modal imaging atlas of the sound wave environment.

[0013] Preferably, the time-frequency domain feature parameters comprise:

[0014] a spectral entropy value of the sound wave signals in a preset time window;

[0015] a time delay difference of the sound wave signals arriving at each node of the distributed acoustic array;

[0016] a harmonic distortion coefficient of the sound wave signals in the frequency domain.

[0017] Preferably, the specific steps of constructing the spatial topology model of the sound wave propagation path are:

[0018] calculate a path length difference value of the sound wave signals arriving at each acoustic node according to the time delay difference in the time-frequency domain feature parameters;

[0019] back-calculate a spatial azimuth angle of the sound wave reflection surface based on the path length difference value, and establish geometric constraint relationships between the reflection surface and the obstacles;

[0020] map the geometric constraint relationships to a topology connection graph of the sound wave propagation path, the topology connection graph comprising a material attenuation coefficient of the reflection surface.

[0021] Preferably, the step of dividing the dynamic partition of the sound wave energy distribution comprises:

[0022] divide the target space into a high attenuation area, a medium attenuation area and a low attenuation area according to the energy attenuation gradient;

[0023] extract a scattering component of the sound wave signals in the high attenuation area, and calculate a spatial correlation of the scattering component;

[0024] dynamically adjust a partition boundary threshold value of the medium attenuation area and the low attenuation area based on the spatial correlation.

[0025] Preferably, the step of phase compensating the acoustic wave signals in the dynamic partition is:

[0026] identifying phase mutation points in the compensated acoustic wave signal sequence;

[0027] correcting phase offset of the acoustic wave signal sequence according to matching results of the phase mutation points and the non-continuous boundary of the three-dimensional density field;

[0028] superimposing the corrected phase offset to the original acoustic wave signal sequence to generate a phase-aligned acoustic wave signal sequence.

[0029] Preferably, the step of reconstructing the three-dimensional density field of acoustic wave energy includes:

[0030] dividing the compensated acoustic wave signal sequence into a plurality of subsequences according to time windows;

[0031] calculating energy projection vectors of each subsequence in three-dimensional space;

[0032] generating an initial grid of the density field according to spatial superposition results of the energy projection vectors;

[0033] topologically optimizing the initial grid based on the non-continuous boundary to obtain an optimized three-dimensional density field.

[0034] Preferably, the specific steps of calculating the energy projection vectors are:

[0035] extracting the main frequency component of the acoustic wave signal in the subsequence;

[0036] determining the propagation direction vector of the acoustic wave energy according to the geometric constraint relationship between the main frequency component and the spatial topology model;

[0037] multiplying the propagation direction vector and the energy amplitude of the acoustic wave signal to obtain the energy projection vector.

[0038] Preferably, the step of marking the non-continuous boundary of the three-dimensional density field includes:

[0039] detecting grid nodes in the three-dimensional density field whose energy gradient changes exceed a preset threshold;

[0040] performing consistency checking on the energy gradient change direction of adjacent grid nodes;

[0041] marking the starting position and the ending position of the non-continuous boundary according to the checking results.

[0042] Preferably, the step of generating a multi-modal imaging atlas is:

[0043] mapping the time-frequency domain feature parameters to corresponding grid nodes of the three-dimensional density field;

[0044] allocating rendering channels of acoustic environment imaging according to the parameter types of the grid nodes;

[0045] The imaging data of different rendering channels are fused to generate a multi-modal imaging atlas.

[0046] Preferably, the specific steps of assigning rendering channels are as follows:

[0047] A first rendering channel is assigned to the spectral entropy value for representing the frequency domain distribution of the sound wave signal;

[0048] A second rendering channel is assigned to the harmonic distortion coefficient for representing the waveform distortion degree of the sound wave signal;

[0049] A third rendering channel is assigned to the energy value of the three-dimensional density field for representing the spatial density of the sound wave energy.

[0050] Compared with the prior art, the present application has the following advantages:

[0051] The multi-channel sound wave signals are collected by the distributed sound array, which can synchronously obtain sound wave information from multiple spatial positions compared with the traditional single-point or centralized collection method, and the extracted time-frequency domain feature parameters are more comprehensive and representative, which can completely capture the propagation characteristics and variation law of the sound wave signals in the target space, and provide rich and reliable basic data for subsequent modeling and imaging.

[0052] The spatial topology model constructed according to the time-frequency domain feature parameters integrates the geometric constraint relationship of the sound wave reflection surface and the obstacle, which can accurately restore the actual propagation path of the sound wave in the target space, avoids the propagation simulation deviation caused by ignoring the geometric constraint in the traditional model, and makes the modeling of the sound wave propagation process more suitable for the actual environment, thereby providing accurate model support for subsequent energy calculation and partition.

[0053] The sound wave energy attenuation gradient is calculated based on the spatial topology model and the dynamic partition is divided, which breaks through the limitation of the traditional static partition, can track the change trend of the sound wave energy in real time, accurately capture the dynamic distribution characteristics of the sound wave energy in different regions, and makes the energy partition result more reflect the real-time state of the sound wave environment, thereby effectively improving the adaptation ability to the dynamic sound wave scene.

[0054] The phase compensation is performed on the sound wave signals in the dynamic partition, which can effectively eliminate the phase distortion of the sound wave in the propagation process caused by the path difference, obstacle shielding and other factors, so that the compensated sound wave signal sequence is closer to the original propagation state, thereby providing high-quality signal source for subsequent three-dimensional density field reconstruction, and guaranteeing the accuracy of imaging from the source.

[0055] The three-dimensional density field of sound wave energy is reconstructed by the compensated signal sequence, and the non-continuous boundary is marked, so that the three-dimensional distribution form of the sound wave energy in the target space can be clearly presented, the positions and contours of key structures such as obstacles and reflecting surfaces can be accurately identified, the problem that the traditional reconstruction method cannot accurately distinguish the key structures in space is solved, and the spatial characteristics of the sound wave environment are more intuitively and clearly presented.

[0056] The time-frequency domain feature parameters of the fused multi-channel sound wave signals and the three-dimensional density field are used to generate a multi-modal imaging atlas, so that the time, frequency and spatial dimension information are organically combined, and the limitation of the traditional single-dimensional imaging is broken. The generated imaging atlas contains the time-frequency domain variation law of the sound wave signal and the three-dimensional characteristics of the spatial energy distribution, can comprehensively depict the overall state of the sound wave environment from multiple angles, and provides more abundant information support for the environmental perception needs in different scenes. BRIEF DESCRIPTION OF DRAWINGS

[0057] Figure 1 A working principle diagram of the sound wave environment imaging method based on an acoustic array is shown.

[0058] Figure 2 A flowchart for constructing a spatial topology model of a sound wave propagation path is shown.

[0059] Figure 3 A flowchart for reconstructing a three-dimensional density field of sound wave energy is shown. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present application.

[0061] Please refer to Figure 1The application provides a sound wave environment imaging method based on a sound array, which comprises the following steps: a distributed sound array is formed by a plurality of sound nodes arranged at different positions in a target space; the sound waves propagating in the space are synchronously collected by the distributed sound array; the collected multi-channel sound wave signals are sent to a digital signal processing unit after being converted from analog signals to digital signals; the sound wave signals of each channel are preprocessed by the digital signal processing unit, the preprocessing comprises noise reduction and signal normalization, and then time-frequency domain characteristic parameters representing the characteristics of the sound waves are extracted from the preprocessed sound wave signals; the time-frequency domain characteristic parameters are input into a space modeling module as basic data for constructing an environment model; the space modeling module deduces the geometric constraint relationship between the reflection surfaces and obstacles encountered by the sound waves in the space according to the physical law of sound wave propagation and the path information contained in the time-frequency domain characteristic parameters, and then constructs a space topology model describing the propagation path of the sound waves; the space topology model not only contains geometric information, but also integrates the attenuation characteristics of the reflection surface material to the sound waves; based on the established space topology model, an energy calculation module simulates the propagation process of the sound waves in the target space, calculates the attenuation gradient of the sound wave energy with the increase of the propagation distance and the reflection times, and divides the target space into dynamic partitions with different energy characteristics according to the spatial distribution difference of the energy attenuation gradient; a phase processing module compensates the phase inconsistency of the sound wave signals in the dynamic partitions, and generates a sound wave signal sequence with better phase consistency by considering the phase delay of the sound waves passing through different paths; an imaging reconstruction module reconstructs the density distribution of the sound wave energy in the three-dimensional space by using the compensated sound wave signal sequence, forms a three-dimensional density field, and identifies the non-continuous boundary with sharp energy change in the density field; finally, a multi-modal fusion module performs data layer fusion on the original time-frequency domain characteristic parameters and the reconstructed three-dimensional density field, assigns different visualization rendering channels to different characteristic parameters, and generates a sound wave environment multi-modal imaging atlas comprehensively reflecting the frequency domain characteristics, waveform quality and spatial energy distribution of the sound waves.

[0062] Example 1: see Figure 2, The distributed acoustic array is composed of multiple high-precision microphone nodes, which are deployed at the boundary or key internal positions of the target space according to a predetermined geometric pattern. Each microphone node is equipped with an independent signal conditioning circuit and an analog-to-digital converter to ensure the synchronous acquisition and digitalization accuracy of multi-channel sound wave signals. The collected multi-channel sound wave signals are transmitted to the central processing unit in pulse code modulation format. The central processing unit runs a digital signal processing algorithm to preprocess the original data. The preprocessing steps include applying a finite-length unit impulse response filter for band-pass filtering to remove environmental low-frequency noise and circuit high-frequency thermal noise, and using an amplitude normalization algorithm to eliminate signal amplitude inconsistency caused by differences in microphone node sensitivity. The preprocessed sound wave signals enter the feature extraction stage, which is implemented based on the short-time Fourier transform framework. The system divides the continuous sound wave signal stream into a series of overlapping time windows, with the length of each time window dynamically adjusted according to the main frequency of the sound wave to ensure that it contains enough waveform periods. For each signal segment within a time window, the fast Fourier transform algorithm converts it from the time domain to the frequency domain to obtain the power spectral density distribution. The calculation of spectral entropy value is based on the Shannon entropy formula to quantify the power spectral density distribution, and the quantization result reflects the complexity and uncertainty of the frequency components of the sound wave signal within a specific time window. The measurement of time delay difference relies on the generalized cross-correlation function. The algorithm performs cross-correlation operation on the same signal segment received by any two microphone nodes in the distributed acoustic array, and determines the relative time difference of the sound wave arriving at the two nodes by detecting the peak position of the cross-correlation function. The product of this time difference and the sound wave propagation speed is the path length difference. The analysis of harmonic distortion coefficient uses the total harmonic distortion calculation method. The algorithm first determines the fundamental frequency and its amplitude of the sound wave signal through spectral peak detection, then extracts the amplitude of the second, third, and up to the preset number of harmonic components, and defines the total harmonic distortion coefficient as the ratio of the root mean square value of the amplitudes of each harmonic to the amplitude of the fundamental wave.

[0063] The spatial topology modeling unit receives the time-frequency domain feature parameter dataset from the feature extraction unit, and the modeling process takes the path length difference value converted from the time delay difference as the core input data. The path length difference value forms an over-determined equation set, and the unknowns of the equation set are the spatial coordinates of the sound source point or the reflection point relative to the microphone node array. The iterative least squares algorithm is used to solve the equation set. In the initial stage of the algorithm, an initial position of the sound source is assumed. By calculating the residual sum of squares of the theoretical path difference and the actual measured path difference, the gradient descent method is used to continuously adjust the estimated value of the sound source position until the residual is minimized. The process of backstepping the spatial azimuth angle of the sound wave reflection surface combines the geometric acoustics principle and the mirror method. The algorithm equivalent the actual sound wave reflection path to the straight line propagation path from the virtual mirror point of the sound source to the microphone node. By analyzing the coordinates of the virtual mirror points of the reflection path received by multiple microphone nodes, the normal vector of the reflection surface is calculated using the plane fitting algorithm to determine its spatial azimuth angle. The establishment of the geometric constraint relationship between the reflection surface and the obstacle needs to introduce a spatial geometric reasoning engine. The engine regards the reflection surface as an infinitely extended plane and models the obstacle as a convex polyhedron in three-dimensional space. The derivation of the geometric constraint relationship is completed by calculating the intersection test of the reflection surface plane equation and the obstacle bounding box. The test results can show that there is a parallel, perpendicular or oblique space relationship between the reflection surface and the obstacle. For example, when the dot product of the normal vector of the reflection surface and the normal vector of a surface of the obstacle is close to zero, it is determined as a parallel relationship, and when the absolute value of the dot product is close to one, it is determined as a perpendicular relationship. The assignment of the material attenuation coefficient is realized by querying the pre-constructed acoustic material database. The database stores the sound wave absorption coefficient and reflection coefficient of common building materials such as concrete, glass and wood at different frequencies. The system retrieves the corresponding attenuation parameters from the database according to the material identifier of the reflection surface.

[0064] The construction of the topology connection graph adopts a graph theory data structure. The vertex set of the graph includes the sound source point, each microphone node, and the calculated reflection points. The edge set of the graph represents all possible propagation paths of sound waves between the vertices. Each edge is accompanied by a set of attribute values, which include the physical length of the path, the number of reflections experienced by the path, and the product of the attenuation coefficients of the materials on the path. The connectivity of the graph is verified by a sound ray tracing algorithm. The algorithm emits a large number of virtual sound rays from the sound source point and tracks the propagation trajectories of the sound rays between the reflection surfaces. Only those paths that eventually reach at least one microphone node are retained in the topology connection graph. The final topology connection graph is stored in the memory in the form of an adjacency matrix. The non-zero element values of the adjacency matrix represent the comprehensive attenuation weight of the path. The weight value is used for subsequent sound wave energy propagation simulation and attenuation gradient calculation. The entire implementation process relies on a high-precision time synchronization mechanism. The microphone nodes of the distributed sound array achieve microsecond-level synchronization of the sampling clock through a precise clock protocol. Time synchronization errors will directly lead to distortion of the time delay difference measurement and affect the accuracy of the spatial topology model. The delay of the signal processing link is strictly calibrated. The total delay from the analog signal input to the digital feature parameter output remains constant. The constant delay is conducive to the alignment and correlation of time series data. The feature extraction unit and the spatial topology modeling unit communicate through a high-speed data bus. The data bus uses a timestamp matching mechanism to ensure accurate association of each feature parameter set with the corresponding sound wave signal segment.

[0065] The updating mechanism of the spatial topology model supports dynamic environment adaptation. The system periodically re-executes the feature extraction and modeling process, and detects the movement or addition of obstacles in the environment by comparing the changes in the topology connection graph between consecutive periods. During dynamic updating, the algorithm maintains a historical model cache, compares the current model with the historical model, marks the changed graph vertices and edges, and evaluates their change significance. Only changes with significant changes exceeding the threshold will trigger a global update of the entire acoustic environment imaging model. This incremental updating strategy balances the computational efficiency and response speed to environmental changes. The accuracy verification of the model is integrated into the topology modeling unit. The cross-validation strategy is used to verify the method. The system randomly hides part of the microphone node measurement data, reconstructs the topology model using the remaining node data, and then compares the hidden node path difference predicted by the reconstructed model with the actual measurement value. The error statistics such as root mean square error and mean absolute error are calculated and recorded in real time. When the error exceeds the preset tolerance, the system will trigger the model reconstruction process instead of directly applying the current model. The quality monitoring of feature parameters runs through the entire process of this embodiment. The system sets reasonable value ranges and noise thresholds for spectral entropy, time delay difference, and harmonic distortion coefficient. Any abnormal parameter value that exceeds the range will be marked as invalid data and trigger data re-sampling or algorithm parameter adjustment. The monitoring logic establishes a control chart based on the statistical process control method, tracks the quality fluctuations of feature parameters in real time, and corrects systematic deviations in a timely manner.

[0066] In embodiment 2, the energy calculation module receives the acoustic wave propagation path topology connection graph from the spatial topology modeling unit. The topology connection graph contains the propagation path geometry information and material attenuation coefficient from all sound source points to microphone nodes. The calculation of energy attenuation gradient is based on the propagation loss model of acoustic waves in free space and reflection process. The model considers two basic mechanisms: geometric diffusion attenuation and air absorption attenuation. Geometric diffusion attenuation follows the amplitude attenuation law of spherical or cylindrical waves. The air absorption attenuation coefficient is calculated according to the environmental temperature, humidity, and acoustic wave frequency through the international standard formula. For each acoustic wave propagation path, the algorithm integrates the path length, reflection times, and material attenuation coefficient of each reflection surface to calculate the total energy attenuation value of the acoustic wave from the sound source point along the path to any specified point in the target space. The target space is discretized into a three-dimensional grid, and the energy calculation module calculates the energy attenuation value of the acoustic wave through all effective paths to each grid node in parallel. The minimum path energy value is taken as the final energy estimate of the grid node.

[0067] Based on the distribution of energy attenuation values of the grid nodes, the energy calculation module identifies the spatial variation rate of the energy attenuation values in the three-dimensional grid space using a gradient descent algorithm. The algorithm calculates the energy attenuation difference between each grid node and its six neighboring nodes, and obtains the amplitude and direction of the energy attenuation gradient through three-dimensional difference operation. The dynamic partition of the sound wave energy distribution sets two threshold boundaries according to the amplitude of the energy attenuation gradient. The gradient amplitude greater than the high threshold is defined as the high attenuation area, the gradient amplitude between the high threshold and the low threshold is defined as the medium attenuation area, and the gradient amplitude less than the low threshold is defined as the low attenuation area. The initial values of the threshold boundaries are set according to historical experience data, and are dynamically adjusted during the system operation. The high attenuation area usually corresponds to the area where the sound wave needs to cross the physical barrier or experience multiple reflections, the medium attenuation area represents the area where there is single reflection or moderate distance propagation, and the low attenuation area is mainly the direct sound field area near the sound source.

[0068] In the high attenuation area, the signal processing algorithm needs to extract the scattering component of the sound wave signal. The extraction of the scattering component uses independent component analysis blind source separation technology. The algorithm regards the mixed signals received by multiple microphone nodes of the distributed acoustic array as the linear mixture of several independent source signals. The independent component analysis algorithm separates the source signals by maximizing the non-Gaussianity of the signal components. Those components in the separated signal components that do not have obvious directivity and have weak correlation are identified as scattering components. The calculation of the spatial correlation of the scattering components needs to analyze the statistical relationship between the scattering signals received by different microphone nodes. The spatial correlation calculation is based on the cross-correlation matrix of the scattering signal segment. The ratio of the maximum eigenvalue to the sum of the eigenvalues of the cross-correlation matrix is used as the quantitative indicator of the spatial correlation. A higher spatial correlation ratio indicates that the scattering energy mainly comes from a few dominant reflection paths, and a lower spatial correlation ratio reflects a more dispersed scattering energy distribution. Based on the calculation results of the spatial correlation, the system dynamically adjusts the partition boundary threshold values of the medium attenuation area and the low attenuation area. When the spatial correlation of the scattering components in the high attenuation area is high, the algorithm appropriately increases the boundary threshold values of the medium attenuation area and the low attenuation area, so that the range of the medium attenuation area expands towards the low attenuation area. The adjustment of the boundary threshold values uses a proportional-integral-derivative control strategy. The proportional term responds to the instantaneous change of the spatial correlation, the integral term accumulates the historical change trend, and the derivative term predicts the future change direction. The output of the controller acts on the offset of the partition boundary threshold value, realizing the smooth adaptive change of the partition boundary. The dynamic partition result is stored in the form of a three-dimensional mask matrix. The value of each element in the matrix identifies the partition category to which the corresponding grid node belongs. This mask matrix will be used to guide the subsequent differential signal processing.

[0069] The phase processing module receives the dynamic partition result and the corresponding acoustic signal data, and the phase compensation operation is performed on the signal in each partition respectively. The detection of phase mutation points adopts a phase difference method. The algorithm performs first-order difference operation on the instantaneous phase of the acoustic signal. The points whose difference values exceed the preset threshold are marked as phase mutation points. The detection of phase mutation points is performed jointly in time-frequency domain. The signal is first subjected to short-time Fourier transform to obtain a time-frequency spectrum. The phase difference operation is independently performed on each frequency band. The events that are simultaneously detected as phase mutations on multiple frequency bands are assigned a higher confidence level. The detected phase mutation points need to be matched with the non-continuous boundary of the three-dimensional density field. The matching process calculates the spatial distance between the sound propagation path corresponding to the timestamp of the phase mutation point and the non-continuous boundary of the three-dimensional density field. The matching pair with a distance less than a tolerance value is regarded as an effective matching. The phase offset of the acoustic signal sequence is corrected according to the matching result. The correction amount is calculated based on the geometric path difference of the acoustic propagation. For each matched phase mutation point, the algorithm recalculates the theoretical propagation time of the acoustic wave from the sound source to the microphone node according to the spatial position of the non-continuous boundary. The difference between the theoretical propagation time and the time when the mutation is actually detected is converted into a phase offset. The calculation of the phase offset considers the center frequency of the acoustic wave. The time difference is multiplied by the angular frequency to obtain the phase correction value in radians. The correction value is applied to the signal segment in a time window centered on the phase mutation point in the original acoustic signal sequence. The phase compensation is performed in the frequency domain. The algorithm converts the signal segment to the frequency domain, applies a linear phase rotation to each frequency component of the spectrum, and the rotation angle is determined by the calculated phase offset. After the frequency domain processing is completed, it is restored to the time domain through inverse Fourier transform.

[0070] The phase alignment of the sound wave signal sequence is generated by integrating the compensation results of all partitions and all time windows; the integration process adopts the overlap-add method, applies a window function to each compensated signal segment, and performs a weighted average of the signals in the overlapping part to eliminate the boundary effects caused by the segmentation process. The final phase-aligned sound wave signal sequence maintains the same sampling rate and length as the original sequence, but the phase continuity is significantly improved; the phase-aligned sequence is stored in a ring buffer for use by the subsequent three-dimensional density field reconstruction module. The entire phase compensation process is implemented on a field programmable gate array hardware platform, which uses parallel processing architecture to meet real-time requirements; the parameters of the phase compensation algorithm, such as the mutation detection threshold and the matching tolerance, can be adjusted online through a configuration interface to adapt to different acoustic environments and signal characteristics. The system continuously monitors the phase compensation effect, and the monitoring methods include calculating the coherence index and phase variance of the signal segments before and after compensation; the coherence index evaluates the consistency between different microphone node signals, and the phase variance reflects the stability of the phase within a single signal segment. The monitoring results are fed back to the parameter adjustment logic of the phase processing module to form a closed-loop control; when the monitoring index is below the expected level, the system automatically fine-tunes the phase compensation parameters or triggers the re-compensation process. This self-optimization mechanism ensures that the phase compensation operation can maintain good performance when the environmental conditions change, providing a fundamental guarantee for the accuracy of acoustic environment imaging.

[0071] Embodiment 3: refer to Figure 3 The imaging reconstruction module receives the phase-aligned sound wave signal sequence from the phase processing module, which is stored in the buffer memory as a discrete-time signal with a fixed sampling rate. The operation of dividing the sound wave signal sequence into multiple subsequences uses the overlapping sliding window mechanism, and the window length is dynamically adjusted according to the period of the main frequency of the sound wave signal to ensure that each window contains an integer multiple of the period; the window function is selected as the Hanning window to reduce spectral leakage, and the overlapping area is set to fifty percent of the window length to ensure time continuity. Each subsequence carries timestamp information and is associated with the potential sound source position in three-dimensional space, and the association relationship is calculated based on the time delay difference of the sound wave arriving at each node of the distributed acoustic array. Calculating the energy projection vector of each subsequence in three-dimensional space requires extracting the main frequency component of the sound wave signal; the identification of the main frequency component uses the autocorrelation function analysis method, which calculates the autocorrelation function of the subsequence signal and detects the position of the maximum peak, and the reciprocal of the main frequency component corresponds to the time delay of the maximum peak of the autocorrelation function. The amplitude of the main frequency component is obtained by calculating the short-time Fourier transform modulus of the subsequence signal at the corresponding frequency, and the frequency resolution is determined by the window length. Determining the propagation direction vector of the sound wave energy depends on the geometric constraint relationship provided by the spatial topology model, which includes the normal vector direction of the reflecting surface and the incident plane of the sound wave. The calculation of the propagation direction vector combines the law of specular reflection of sound rays, and the dot product operation of the incident direction and the normal vector of the reflecting surface determines the vector expression of the reflected direction.

[0072] The generation of energy projection vector is realized by vector construction, combining the scalar representing the intensity of sound wave with the unit vector of propagation direction. The mathematical expression of sound intensity projection vector is:

[0073]

[0074] wherein: represents the sound intensity projection vector, which is the directional representation of sound wave energy flow density in space, with the dimension of watt per square meter [W / m²]; represents the sound intensity amplitude, which is a scalar, calculated based on the square of the root mean square value of sound pressure signal, with the dimension of watt per square meter [W / m²]; represents the unit vector of propagation direction, which is a dimensionless quantity.

[0075] The spatial superposition of sound intensity projection vector is carried out in a three-dimensional grid space, and the resolution of the grid space is pre-set according to the target space size and accuracy requirements. Each grid cell maintains an accumulator, which performs summation operation on all sound intensity projection vectors projected into the cell; the projection judgment is based on the ray casting algorithm, which emits a ray from the potential position of the sound source along the propagation direction vector, and the grid cells crossed by the ray are marked as affected cells. The initial grid is generated using the inverse distance weighted interpolation method, and the accumulated energy value of each grid cell is used as the node value to perform surface fitting, generating a continuous three-dimensional energy distribution representation. Adaptive grid refinement technology needs to be introduced for topology optimization of the initial grid based on discontinuous boundaries; the technology automatically increases the grid density in the area where the discontinuous boundary is detected, and maintains a relatively coarse grid division in the area where the energy changes gently. The optimization process uses a grid refinement strategy based on error estimation, and the error estimator calculates the second-order derivative of the energy value in each grid cell, and the area with large second-order derivative is identified as the area where the energy distribution changes dramatically. The topology optimization algorithm inserts new grid nodes at the discontinuous boundary, and the energy value of the new node is calculated from the adjacent nodes by cubic spline interpolation, maintaining the continuity and smoothness of the energy distribution.

[0076] The optimized 3D density field is stored in a voxel grid, each voxel contains the energy intensity value and spatial coordinate information. The visualization of 3D density field uses direct volume rendering technique, which calculates the color and transparency of each pixel through ray casting algorithm to generate perspective images of 3D energy distribution. The 3D density field data is also output to the multi-modal fusion module for data fusion with frequency domain feature parameters. The entire reconstruction process contains multiple quality control steps, which are realized by calculating reconstruction error indicators. The error indicators include energy conservation check and boundary consistency check. The energy conservation check compares the total energy of the input signal with the volume integral energy of the 3D density field. The boundary consistency check verifies the alignment degree of the non-continuous boundary and the physical boundary in the acoustic image. When the error indicators exceed the allowable range, the system automatically triggers the re-execution of the reconstruction process and adjusts the related algorithm parameters such as grid resolution or interpolation method.

[0077] The computational efficiency of the reconstruction algorithm is optimized through parallel computing technology. The 3D grid space is divided into multiple sub-regions and assigned to different computing units for parallel processing. Graphics processor hardware acceleration is used for sound intensity projection vector calculation and grid interpolation operations to achieve real-time processing capability for large-scale 3D data sets. The system supports dynamic resolution adjustment, automatically balancing reconstruction accuracy and processing speed according to available computing resources. When computing resources are limited, a multi-resolution hierarchical processing strategy is used. The storage of 3D density field uses a hierarchical data structure, with the bottom layer storing full-resolution data and the upper layer storing down-sampled multi-resolution representations. The hierarchical structure supports fast data retrieval and visualization, and users can automatically switch to appropriate resolution data representations according to the viewing scale. Data compression algorithms are applied to the persistent storage of 3D density field. The compression algorithms are based on energy value prediction encoding and entropy encoding, which reduce storage space requirements while maintaining accuracy. The reliability of the reconstruction results is verified through cross-validation methods. The method randomly divides the distributed sound array nodes into training set and test set. After reconstructing the 3D density field using the training set node data, the energy values of the test set node positions are predicted and compared with the actual measured values. The verification results generate a consistency report, identifying areas with high reliability and areas with possible uncertainty in the 3D density field. The reliability information is stored as metadata along with the 3D density field for reference by subsequent processing modules.

[0078] In the embodiment 4, the boundary detection module receives the optimized 3D density field data from the imaging reconstruction module, which is organized in a regular grid structure, and each grid node stores a normalized acoustic energy density value ranging from 0 to 1. The process of marking the discontinuous boundary starts from calculating the energy gradient field of the 3D density field, which is computed by the central difference algorithm to calculate the partial derivatives of each grid node in the three coordinate axes directions. For the internal grid nodes, the algorithm uses a six-neighborhood template to calculate the gradient components in X, Y, and Z directions, and for the boundary grid nodes, the forward or backward difference format is used to avoid out-of-bound access. Detecting the grid nodes with gradient changes exceeding the preset threshold requires setting double judgment conditions. The first condition judges whether the gradient amplitude exceeds the global threshold, and the gradient amplitude is calculated as the Euclidean norm of the three directional gradient components. The second condition judges the consistency of the gradient direction by analyzing the gradient vector angle of the adjacent grid nodes to filter out isolated noise points. The determination of the global threshold is based on the statistical distribution of the overall gradient amplitude of the 3D density field, and the system calculates the average value and standard deviation of the gradient amplitude, and sets the threshold as the average value plus three times the standard deviation to capture significant gradient changes. The local adaptive threshold supplements the weak boundaries that may be missed by the global threshold, and the local threshold is dynamically adjusted according to the gradient amplitude range in the small neighborhood around each grid node.

[0079] The consistency check of the energy gradient change direction of adjacent grid nodes uses the vector field analysis method, which calculates the average dot product value of the gradient vector of each grid node and the six grid nodes in the direct neighborhood. A dot product value close to one indicates a high consistency in the gradient direction, a dot product value close to negative one indicates an opposite gradient direction, and a dot product value close to zero indicates an orthogonal gradient direction. The consistency check sets an angle tolerance parameter, and only when more than half of the grid nodes in the neighborhood have a gradient vector angle with the center node less than fifteen degrees, it is considered to pass the consistency check. The grid nodes that do not pass the consistency check are considered as noise points even if the gradient amplitude exceeds the threshold. According to the check result, the starting position and the ending position of the discontinuous boundary need to apply the boundary tracking algorithm, which starts from any seed point that passes the threshold detection and consistency check, and grows along the plane perpendicular to the gradient direction. The boundary tracking algorithm uses a three-dimensional version of the Canny edge detection idea, which first performs non-maximum suppression to retain the grid nodes with locally maximum gradient amplitude, and then combines the strong edge nodes and the connected weak edge nodes into complete boundary contours through double-threshold connection. The post-processing of the boundary segments includes removing short and isolated segments, connecting disconnected boundary gaps, and smoothing the jagged fluctuations of the boundary curve.

[0080] The marking result is stored in the form of a set of spatial coordinates of the boundary mesh nodes, and each boundary segment is accompanied by topological connection information describing the adjacency relationship between segments. The system calculates geometric feature descriptors for each boundary segment, including parameters such as the average curvature, total length, bounding box size, normal vector direction, etc. of the segment. These geometric features are used in the subsequent multi-modal imaging atlas generation process for boundary type classification and rendering style assignment. The quality control of the boundary marking process is achieved through a multi-scale verification mechanism, which repeatedly executes the boundary detection algorithm on mesh representations at different resolutions. By comparing the consistency of the boundary positions detected at different scales, only those boundaries that consistently appear at multiple scales are confirmed as valid non-continuous boundaries. Multi-scale verification effectively suppresses pseudo-boundaries introduced by mesh discretization, improving the robustness and reliability of boundary marking. The marking result of the non-continuous boundaries of the three-dimensional density field is stored in a special boundary database, which uses a spatial indexing structure to support fast range queries and proximity searches. The boundary data and the original three-dimensional density field data are associated through spatial coordinates, allowing quick retrieval of the corresponding energy density value distribution characteristics based on boundary position. The database also records the detection confidence of the boundary, which is calculated based on the gradient amplitude, consistency verification score, and multi-scale verification consistency. The computational performance of the boundary marking algorithm is optimized through spatial blocking technology, which divides the three-dimensional mesh space into appropriate size sub-blocks, each of which performs boundary detection independently and then merges the global results. An overlap region is set at the boundary of the sub-block to avoid cutting the boundary segment, and the merging process is achieved through matching the boundary nodes in the overlap region to achieve seamless connection. Parallel computing architecture fully utilizes multi-core processors and graphics processors to accelerate vector operations and neighborhood queries in boundary detection. The system provides a visual configuration interface for boundary marking parameters, allowing users to interactively adjust parameters such as gradient threshold, consistency tolerance, and minimum boundary length and view the marking effect in real time. The parameter adjustment history and corresponding boundary detection results are automatically recorded to form a parameter-effect mapping relationship, which is analyzed by a machine learning algorithm to form an automatic parameter recommendation function, helping users quickly obtain satisfactory boundary marking results.

[0081] Referring to Table 1, the marking result of the non-continuous boundary is not only used for visual rendering, but also provides structural information for acoustic environment analysis. Hard boundaries correspond to the surfaces of physical obstacles, and soft boundaries correspond to areas of acoustic impedance change. Different types of boundaries are displayed in different colors and textures in the multi-modal imaging atlas. Boundary curvature information reveals the concave-convex characteristics of the surface, and boundary direction distribution reflects the structural orientation characteristics of the space.

[0082] Table 1: Geometric feature descriptors of non-continuous boundaries

[0083] Feature name Calculation method Physical meaning Data type Boundary segment length Euclidean distance accumulation between adjacent nodes on the boundary path Reflects the extension range of the boundary space Floating point number (meters) Mean curvature Arithmetic mean of the curvature of each point on the boundary path Characterizes the degree of bending of the boundary Floating point number (1 / meter) Normal vector consistency Standard deviation of the angle between the normal vector of each point on the segment and the average normal vector Describes the smoothness of the boundary surface Floating point number (degrees) Gradient amplitude mean Arithmetic mean of the gradient amplitude of the boundary nodes Indicates the strength of energy change at the boundary Floating point number Boundary dimension ratio Ratio of the longest side to the shortest side of the boundary bounding box Reflects the anisotropy of the boundary shape Floating point number

[0084] The integration of the boundary marking module with other modules of the system is achieved through a standardized data interface, which defines the format of boundary data, the coordinate system convention, and the updating mechanism. When the three-dimensional density field is updated, the boundary marking module automatically triggers recalculation, and the incremental updating algorithm only performs local boundary re-marking on the changed area of the density field to improve efficiency. The change history of the marking results is recorded for analyzing the dynamic evolution process of the acoustic environment.

[0085] In embodiment 5, the fusion rendering module receives the time-frequency domain feature parameters from the feature extraction unit and the three-dimensional density field data from the imaging reconstruction module. The time-frequency domain feature parameters include spectral entropy values and harmonic distortion coefficients associated with each grid node. The three-dimensional density field data contains spatial coordinates and energy density values of the grid nodes. The generation of the multi-modal imaging atlas begins with the data alignment step, which maps the time-frequency domain feature parameters to the corresponding grid nodes of the three-dimensional density field based on a unified space-time coordinate system. Each grid node is associated with a set of feature values, including energy density values, spectral entropy values, and harmonic distortion coefficients, forming a data point set with multiple attributes. Mapping the time-frequency domain feature parameters to the corresponding grid nodes of the three-dimensional density field requires addressing the data scale difference. The time-frequency domain feature parameters are derived from the results of sound wave signal processing, while the three-dimensional density field is derived from the spatial reconstruction algorithm. The data mapping uses a spatial interpolation method to resample the time-frequency domain feature parameters to the grid node positions of the three-dimensional density field. The interpolation method considers the physical characteristics of sound wave propagation. For each grid node, the system finds its adjacent sound wave sampling points and calculates the feature parameter values of the node based on the weighted average of the propagation path lengths. The mapping process ensures that each grid node obtains a complete feature vector, which serves as the data basis for multi-modal rendering.

[0086] To assign rendering channels for acoustic environment imaging based on the parameter types of the grid nodes, a color mapping function needs to be designed, which converts different types of feature parameters into visually distinguishable color components. The first rendering channel is assigned to the spectral entropy values, which uses a continuous color gradient scheme to represent the numerical range of spectral entropy values. Low spectral entropy values are mapped to blue tones, medium spectral entropy values are mapped to green tones, and high spectral entropy values are mapped to red tones. The color gradient is implemented through linear interpolation in the HSV color space. The second rendering channel is assigned to the harmonic distortion coefficients, which uses a brightness gradient scheme to represent the degree of waveform distortion. Low harmonic distortion coefficients are displayed as dark tones, and high harmonic distortion coefficients are displayed as bright tones. The brightness value has a linear relationship with the harmonic distortion coefficient. The third rendering channel is assigned to the energy values of the three-dimensional density field, which uses a transparency gradient scheme to represent the spatial density of sound wave energy. Low energy areas are displayed with high transparency, and high energy areas are displayed with low transparency. The transparency change follows an exponential decay curve to enhance the visual contrast.

[0087] The multi-modal imaging atlas is generated by fusing imaging data of different rendering channels using a pixel-level superimposition algorithm. The algorithm performs weighted mixing of color contributions of each grid node on three rendering channels. Color mixing is performed in the RGBA color space, and the final color value of each grid node is determined by the hue of the first rendering channel, the brightness of the second rendering channel, and the transparency of the third rendering channel. The mixing weights are dynamically adjusted according to the analysis needs of the observer, with the weight of the first rendering channel increased when focusing on frequency domain characteristic analysis, the weight of the second rendering channel increased when focusing on waveform quality evaluation, and the weight of the third rendering channel increased when focusing on energy distribution observation.

[0088] The visualization of the multi-modal imaging atlas is based on volume rendering technology, which generates a two-dimensional projection image of a three-dimensional scene through a ray casting algorithm. Each ray emitted from the viewpoint passes through the three-dimensional data field, sampling multiple points along the ray path, and the color and transparency of each sampling point are determined by the multi-modal color value of the grid node through trilinear interpolation. All sampling points on the ray are color-mixed in order from front to back or back to front, finally forming the color value of the pixel point. The rendering process is performed in real time, supporting viewpoint rotation, scaling and translation operations, and providing multi-angle observation capability for the sound wave environment. The system provides an adjustable interface for rendering parameters, and users can independently adjust parameters such as contrast, brightness, and color balance of the three rendering channels through slider controls. The parameter adjustment effect is fed back in real time on the display window, and users can find the best visualization configuration through interactive exploration. Common parameter combinations can be saved as preset schemes for quick loading and use in specific application scenarios. The preset schemes include frequency spectrum analysis mode, distortion detection mode, and energy distribution mode.

[0089] The multi-modal imaging atlas supports layered display function, and users can individually enable or disable the display of a certain rendering channel. Layered display facilitates the analysis of spatial correlation between different characteristic parameters, for example, the frequency spectrum entropy value distribution can be displayed separately, and then the energy distribution profile is superimposed to observe the corresponding relationship between frequency domain characteristics and energy distribution. The display control interface provides channel transparency adjustment to achieve smooth transition of display intensity of different channels. The atlas generation process includes an automatic optimization mechanism that analyzes the data distribution characteristics under the current view and automatically adjusts the rendering parameters to enhance the visualization effect. When a large dynamic range of data is detected, the system automatically enables logarithmic color mapping to enhance the display contrast of low value areas; when the data distribution is relatively concentrated, linear color mapping is used to maintain the direct correspondence between the value and the color. Automatic optimization reduces the workload of manual adjustment by users and improves the efficiency of atlas generation.

[0090] The output format of the multi-modal imaging atlas supports multiple standards, including two-dimensional image sequences, three-dimensional volume data files, interactive Web visualization formats, etc. Two-dimensional image sequences generate animation frames rotating around the model for one round, which are used to make demonstration videos; three-dimensional volume data files save the complete multi-modal data for further processing by professional analysis software; the interactive Web format is implemented based on WebGL technology, allowing three-dimensional viewing and basic analysis operations through the browser. The meta-information of the atlas data is recorded completely, including data acquisition time, spatial range, coordinate system definition, rendering parameter configuration, characteristic value statistical information, etc. The meta-information is embedded in the output file in JSON format, ensuring the traceability and repeatability of the data. Users can quickly understand the generation conditions and data characteristics of the atlas through the meta-information. Observers can simultaneously obtain multi-dimensional information such as the spatial distribution of acoustic energy, regional differences in spectral characteristics, and spatial patterns of waveform distortion from a single view. Fusion visualization enhances the understanding of complex acoustic environments and reveals the physical mechanisms of the interaction between sound waves and spatial structures. The multi-modal imaging atlas has application value in the fields of architectural acoustic design, noise source positioning, and abnormal acoustic event detection, etc. The atlas supports quantitative analysis functions, and users can measure the characteristic parameter values of specific regions in the three-dimensional scene, and generate statistical reports to support the professional decision-making process.

[0091] The performance optimization of the rendering system ensures smooth interaction in large-scale data scenarios. The level-of-detail technique automatically adjusts the mesh resolution according to the view distance, and the view frustum culling avoids unnecessary rendering in invisible regions. The multi-thread architecture allocates data loading, feature calculation, and image rendering tasks to different threads for parallel execution, maintaining a responsive user interface. The graphics processor hardware acceleration enables efficient execution of the ray casting algorithm, supporting real-time rendering of datasets with millions of mesh nodes. The multi-modal imaging atlas establishes a bidirectional link with the original sound wave data, allowing users to click on regions of interest in the atlas, and the system automatically retrieves and plays the original sound wave signals at the corresponding locations. The bidirectional link function directly links visualization analysis and auditory perception, assisting users in verifying the auditory correspondence of the visualization results. The click query function displays detailed characteristic parameter values at the current location, which are superimposed in the three-dimensional scene in table form.

[0092] It should be noted that, in the present document, relational terms such as first and second and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0093] While embodiments of the application have been shown and described, it is to be understood that the application is not limited to the details of the embodiments described, since numerous changes, modifications, substitutions and variations can be made thereto without departing from the spirit and scope of the application as defined by the appended claims and their equivalents.

Claims

1. A method of acoustic wave environmental imaging based on an acoustic array, characterized by, The method comprises the following steps: acquiring a multi-channel sound wave signal in a target space through a distributed sound array, and extracting a time-frequency domain characteristic parameter in the sound wave signal; constructing a spatial topology model of a sound wave propagation path according to the time-frequency domain characteristic parameter, the spatial topology model comprising a geometric constraint relationship between a sound wave reflection surface and an obstacle; calculating an energy attenuation gradient of the sound wave in the target space based on the spatial topology model, and dividing a dynamic partition of sound wave energy distribution; performing phase compensation on the sound wave signal in the dynamic partition to generate a compensated sound wave signal sequence; reconstructing a three-dimensional density field of the sound wave energy in the target space according to the compensated sound wave signal sequence, and marking a non-continuous boundary of the three-dimensional density field; fusing the time-frequency domain characteristic parameter of the multi-channel sound wave signal and the three-dimensional density field to generate a multi-modal imaging atlas of a sound wave environment; the step of dividing the dynamic partition of the sound wave energy distribution comprises: dividing the target space into a high-attenuation region, a medium-attenuation region and a low-attenuation region according to the energy attenuation gradient; extracting a scattering component of the sound wave signal in the high-attenuation region, and calculating a spatial correlation of the scattering component; dynamically adjusting a partition boundary threshold of the medium-attenuation region and the low-attenuation region based on the spatial correlation; the step of marking the non-continuous boundary of the three-dimensional density field comprises: detecting a grid node in the three-dimensional density field whose energy gradient change exceeds a preset threshold; performing consistency verification on the energy gradient change direction of adjacent grid nodes; marking a starting position and an ending position of the non-continuous boundary according to a verification result.

2. The acoustic array-based acoustic environmental imaging method of claim 1, wherein, the time-frequency domain characteristic parameter comprises: a spectral entropy value of the sound wave signal in a preset time window; a time delay difference of the sound wave signal arriving at each node of the distributed sound array; a harmonic distortion coefficient of the sound wave signal in the frequency domain.

3. The acoustic array-based acoustic environmental imaging method of claim 1, wherein, the specific steps of constructing the spatial topology model of the sound wave propagation path are: calculating a path length difference value of the sound wave signal arriving at each sound node according to the time delay difference in the time-frequency domain characteristic parameter; back-calculating a spatial azimuth of the sound wave reflection surface based on the path length difference value, and establishing a geometric constraint relationship between the reflection surface and the obstacle; mapping the geometric constraint relationship into a topology connection diagram of the sound wave propagation path, the topology connection diagram comprising a material attenuation coefficient of the reflection surface.

4. The acoustic array-based acoustic environmental imaging method of claim 1, wherein, the step of performing phase compensation on the sound wave signal in the dynamic partition is: identifying a phase mutation point in the compensated sound wave signal sequence; correcting a phase offset of the sound wave signal sequence according to a matching result of the phase mutation point and the non-continuous boundary of the three-dimensional density field; superimposing the corrected phase offset to an original sound wave signal sequence to generate a phase-aligned sound wave signal sequence.

5. The acoustic array-based acoustic environmental imaging method of claim 1, wherein, the steps of reconstructing the three-dimensional density field of the sound wave energy comprise: dividing the compensated sound wave signal sequence into a plurality of subsequences according to a time window; calculating an energy projection vector of each subsequence in a three-dimensional space; generating an initial grid of the density field according to a spatial superposition result of the energy projection vector; topologically optimizing the initial grid based on the non-continuous boundary to obtain an optimized three-dimensional density field.

6. The acoustic array-based acoustic environmental imaging method of claim 5, wherein, the specific steps of calculating the energy projection vector are: extracting a main frequency component of the sound wave signal in the subsequence; determining a propagation direction vector of the sound wave energy according to the geometric constraint relationship between the main frequency component and the spatial topology model; multiplying the propagation direction vector and an energy amplitude of the sound wave signal to obtain the energy projection vector.

7. The acoustic array-based acoustic environmental imaging method of claim 1, wherein, The steps of generating the multi-modal imaging atlas are: mapping the time-frequency domain feature parameters to corresponding grid nodes of the three-dimensional density field; allocating rendering channels of sound wave environment imaging according to the parameter types of the grid nodes; fusing imaging data of different rendering channels to generate the multi-modal imaging atlas.

8. The acoustic array-based acoustic environmental imaging method of claim 7, wherein, The specific steps of allocating the rendering channels of sound wave environment imaging are: allocating a first rendering channel to the spectral entropy value for representing the frequency domain distribution of the sound wave signal; allocating a second rendering channel to the harmonic distortion coefficient for representing the waveform distortion degree of the sound wave signal; allocating a third rendering channel to the energy value of the three-dimensional density field for representing the spatial density of the sound wave energy.

Images