A method for reconstructing and simplifying 3D models for acoustical analysis

Abstract

The present disclosure relates to a computer-implemented method for generating an impulse response for a listening point comprised in at least one indoor volume as reconstructed according to a secondary computer-implemented method for reconstructing a 3D model of a building complex, wherein the computer-implemented method comprises receiving a 3D model of the at least one indoor volume, the position of at least one sound source in the at least one indoor volume, and acoustic properties of at least one boundary in the at least one indoor volume; and determining using a wave-based solver and / or a geometrical acoustics solver, an impulse response of a wave-based and / or a ray-based propagation of an impulse emitted at the at least one sound source in the at least one indoor volume and received at the listening point.

Description

[0001] P7391 PC00

[0002] 1

[0003] A Method for Reconstructing and Simplifying 3D Models for Acoustical Analysis The present disclosure relates to the field of 3D model processing and simplification, particularly for applications in acoustical simulation and analysis of building environments.

[0004] Background

[0005] In the modern construction and design industry, the creation and analysis of 3D models play a pivotal role in managing the lifecycle of buildings and infrastructure. These models are used to represent the geometric and functional aspects of structures, enabling architects, engineers, and other professionals to collaborate effectively.

[0006] Among the various tools available, Building Information Modeling (BIM) has emerged as a powerful framework, allowing detailed representations that integrate both geometry and metadata. BIM has transformed the industry by supporting tasks such as structural analysis, energy modeling, and facility management. However, extending its utility to certain specialized applications, such as acoustical simulation, remains a significant challenge.

[0007] A major difficulty lies in adapting 3D building models for acoustical analysis, a process that requires focusing on the air volumes and internal boundaries where sound propagates. While BIM models can provide highly detailed geometry and metadata, these models are often overloaded with excessive details irrelevant for acoustical purposes. For example, intricate surface textures, minor structural components, and detailed representations of furniture increase the complexity of the models, leading to inefficient processing and computational burden. Acoustic simulations, on the other hand, require simplified and computationally efficient representations that retain only acoustically relevant features. This mismatch between the richness of BIM data and the specific needs of acoustical simulations introduces significant inefficiencies in the workflow.

[0008] Moreover, the diversity of formats and data representations used in 3D modeling exacerbates the problem. While BIM models often conform to the Industry Foundation Classes (IFC) standard, other representations, such as point clouds, polygonal meshes, and wireframe models, may lack the standardization and metadata necessary for acoustical simulations. These inconsistencies across formats require extensive manual intervention to reconstruct and annotate the geometries appropriately.P7391 PC00

[0009] 2

[0010] Additionally, exported models often contain gaps, overlapping geometries, or non-watertight surfaces, further complicating the task of preparing them for simulation.

[0011] A related challenge is the lack of standardized acoustic metadata in BIM and related formats. While some material properties can be included in these models, detailed information relevant to sound propagation, such as absorption coefficients or reflection properties, is often absent. As a result, acoustic engineers must infer or manually input this data, introducing a high likelihood of error and reducing the reliability of simulation results. Furthermore, the process of manually simplifying and reconstructing these models for acoustical use is both time-consuming and error-prone, leading to delays and inconsistencies in project workflows.

[0012] Summary

[0013] It is therefore an objective of the present disclosure to provide an automated method for reconstructing and simplifying 3D models derived from diverse sources, including BIM data, for use in acoustical simulations. The disclosed method aims to overcome the limitations of existing approaches by efficiently processing complex geometries, extracting and annotating acoustically relevant spaces, and ensuring a watertight representation suitable for precise and reliable acoustic analysis.

[0014] The present disclosure relates to a computer-implemented method for reconstructing a 3D model of a building complex, wherein the 3D model of a building complex has a 3D model of a building complex volume, and wherein the 3D model of a building complex comprises at least one indoor volume and at least one outdoor volume, or at least two indoor volumes, wherein the at least one outdoor volume, when present, partially or entirely surrounds the at least one indoor volume, and wherein the at least one indoor volume comprises an inner volume, wherein the method comprises the steps of obtaining a 3D representation of the 3D model of a building complex; filling the inner volume with a plurality of spatial partitioning data structures (SPDSs), wherein the inner volume is delimited by at least one indoor boundary of the at least one indoor volume; applying at least one space proxy to the plurality of SPDSs, wherein each of the at least one space proxy represents a computational proxy of the at least one indoor volume and wherein the computational proxy forms a coherent patch of SPDSs, wherein the coherent patch of SPDSs comprises a plurality of SPDSs that are in contact with each other; and reconstructing the at least one indoor volume comprised in the 3D model of a building complex by assembling together the at least one indoorP7391 PC00

[0015] 3

[0016] boundary spatially correlated to the at least one space proxy, wherein the spatial correlation is based on proximity or geometric adjacency.

[0017] The method described offers several technical advantages. By using spatial partitioning data structures (SPDSs) and space proxies, the method enables efficient reconstruction of volumetric spaces, even when the 3D model of a building complex is incomplete or contains geometrical gaps. This represents a significant improvement over existing methods, which often require significant manual intervention to reconstruct indoor volumes. The use of SPDSs allows for efficient discretization of the indoor volume, ensuring precise spatial representation even when the input geometry is incomplete or contains gaps. This approach minimizes manual adjustments and reduces the computational complexity of subsequent simulations. The coherent patches of SPDSs provide a watertight and computationally manageable representation of spaces, facilitating the identification of boundaries critical for acoustical simulation or any other types of simulation requiring watertight 3D models. Additionally, the inclusion of spatial correlation based on proximity or geometric adjacency ensures that the reconstructed spaces are accurately aligned with the input geometry, minimizing errors and improving simulation reliability. These features collectively provide a practical and efficient solution for handling the complex and diverse data typically encountered in 3D building models.

[0018] The reconstructed indoor volume, which can be defined as a 3D reconstructed model, not only enables improved geometric representation but also provides a robust foundation for advanced acoustical simulations, such as generating impulse responses for specific listening points within the reconstructed spaces.

[0019] The present disclosure also relates to a computer-implemented method for generating an impulse response for a listening point comprised in at least one indoor volume as reconstructed according to the aforementioned method, wherein the method comprises receiving a 3D model of the at least one indoor volume, the position of at least one sound source in the at least one indoor volume, and acoustic properties of at least one boundary in the at least one indoor volume; determining using a wave-based solver and / or a geometrical acoustics solver, an impulse response of a wave-based and / or a ray-based propagation of an impulse emitted at the at least one sound source in the at least one indoor volume and received at the listening point.P7391 PC00

[0020] 4

[0021] The second method provides the ability to integrate the reconstructed 3D model into precise acoustical analysis workflows. By leveraging wave-based and ray-based solvers, the method supports accurate modeling of sound propagation within reconstructed indoor volumes, addressing the limitations of prior methods that often fail to correctly separate spaces within a 3D model of a building complex, thereby causing the simulation to be too computationally expensive.

[0022] In one embodiment, the method for reconstruction is configured to handle diverse 3D representations, including point cloud data, mesh models, polygonal models, voxelbased models, and wireframe models. This capability ensures flexibility in processing input data from various sources, making the method broadly applicable. In another embodiment, the geometry simplification process comprises bounding volume replacement, plane projection, decimation, and voxelization, allowing for computational efficiency without compromising acoustically relevant features. For example, bounding volume replacement simplifies detailed 3D objects while retaining their outer contours, and decimation reduces the number of surface elements while preserving reflective and absorptive properties. These geometry simplifications reduce the complexity of the 3D model, making it computationally feasible to perform acoustical simulations on large-scale building complexes without compromising the accuracy of sound propagation analysis.

[0023] The combination of the reconstruction method and the method for generating impulse responses provides a synergistic solution for acoustical simulation. The reconstruction method ensures accurate modeling of volumetric spaces, which serves as a robust foundation for further simulations to be run within the provided volume. The integration of spatial correlation techniques, geometry simplifications, and acoustic property extraction allows for seamless preparation and simulation of 3D models for acoustical applications. In one embodiment, the combination enables rendering of base audio signals by convolving them with the generated impulse responses, providing realistic audio rendering within the reconstructed spaces. This combination of methods is particularly advantageous for applications in architectural acoustics, virtual reality, and other fields requiring accurate sound simulation in complex 3D environments.

[0024] By integrating these features, the present disclosure provides a novel and efficient solution for reconstructing and simplifying 3D models, as well as for conducting precise acoustical simulations, addressing key challenges in handling complex, incomplete, orP7391 PC00

[0025] 5

[0026] excessively detailed input geometries. In addition to architectural acoustics and virtual reality, the disclosed methods may also be applied to gaming, sound system design, immersive audio experiences, and noise control studies, further demonstrating the versatility of the invention.

[0027] Description of the drawings

[0028] In the following, embodiments and examples will be described in greater detail with reference to the accompanying drawings:

[0029] Fig. 1 illustrates a 3D model of a building complex 100, showing both indoor and outdoor volumes, wherein the indoor volume 110 includes at least a first room 111 and a second room 112, and wherein the outdoor volume 113 partially or entirely surrounds the indoor volume,

[0030] Fig. 2 illustrates a 3D model of a building complex 200, showing the spatial partitioning of the indoor volume using spatial partitioning data structures (SPDSs) 202, wherein a coherent patch of SPDSs 201 is formed within the indoor volume, and wherein a third room 114 is comprised within the building complex,

[0031] Figs. 3A-B illustrate a detailed view of a coherent patch of SPDSs 201 and its relationship with the indoor boundary 302 and the outdoor boundary 301 ,

[0032] Fig. 4 illustrates the closest point query scheme used in the reconstruction of 3D models of building complexes, wherein the figure shows a set of SPDSs 401 , a first boundary 402, a second boundary 403, a correct closest point query 404 selecting the nearest boundary, and incorrect closest point queries 405 that incorrectly select the second boundary,

[0033] Fig. 5 illustrates multiple examples of 3D models of elements 501 , including furniture such as chairs and stools, and their corresponding plane projections 502 obtained using a plane projection module and / or a polygon replacement module, Fig. 6 illustrates examples of 3D models of furniture, including a chair with armrests 600, a table 610, and a piece of furniture 620, which shows how the 3D models are voxelized and then simplified using plane projection and / or polygon replacement based on the voxelization results,P7391 PC00

[0034] 6

[0035] Figs. 7A-B illustrate a 3D model of a vase with complex external contours and a simplified bounding volume representation of the vase,

[0036] Figs. 8A-B shows a 3D model of a cabinet with specific geometric details, including a front door, and the corresponding simplified bounding volume of the cabinet, Figs. 9A-B illustrate a 3D model of a window comprising a frame, and the simplified version of the window obtained using a plane projection module,

[0037] Figs. 10A-B show a 3D model of a chair (stool) with detailed geometries, and the simplified version obtained through a polygon replacement module after outer shell voxelization,

[0038] Figs. 11 A-B illustrate the decimation process, where a 3D model of an element undergoes simplification using a decimation module to reduce the number of surface elements.

[0039] Detailed description

[0040] The present disclosure relates to a computer-implemented method for reconstructing a 3D model of a building complex, wherein the 3D model of a building complex has a 3D model of a building complex volume, and wherein the 3D model of a building complex comprises at least one indoor volume and at least one outdoor volume, or at least two indoor volumes, wherein the at least one outdoor volume, when present, partially or entirely surrounds the at least one indoor volume, and wherein the at least one indoor volume comprises an inner volume, wherein the method can comprise the steps of obtaining a 3D representation of the 3D model of a building complex; filling the inner volume with a plurality of spatial partitioning data structures (SPDSs), wherein the inner volume is delimited by at least one indoor boundary of the at least one indoor volume; applying at least one space proxy to the plurality of SPDSs, wherein the at least one space proxy forms a coherent patch of SPDSs, wherein the coherent patch of SPDSs comprises a plurality of SPDSs that are in contact with each other; and preferably reconstructing the at least one indoor volume comprised in the 3D model of a building complex by assembling together the at least one indoor boundary spatially correlated to the at least one space proxy. The spatial correlation can be based on proximity or geometric adjacency.P7391 PC00

[0041] 7

[0042] In this aspect, the computer-implemented method may also be referred to simply as a method for reconstructing a 3D model of a building complex. For clarity, references to “the method” or “a method” throughout the description of this aspect imply a computer-implemented method.

[0043] In one embodiment, the proximity is defined as a shortest measurable distance between an SPDS and the at least one indoor boundary, satisfying a predetermined distance threshold. This definition can provide a quantifiable and computationally implementable criterion for determining whether an SPDS is spatially correlated to the indoor boundary. The distance may be calculated using a variety of metrics, such as the Euclidean distance between the center point of the SPDS and the nearest point on the at least one indoor boundary. In the case of voxel-based SPDSs, proximity may be determined by measuring the distance from the voxel centroid to the closest point on the surface of the at least one indoor boundary. This measurable distance can ensure that only SPDSs within a defined range of the boundary are considered for correlation, providing a clear and consistent method for associating SPDSs with their corresponding boundaries of the at least one indoor volume of the 3D model of a building complex.

[0044] The predetermined distance threshold may vary depending on the requirements of the specific application or the resolution of the 3D model. For example, a finer threshold may be used in scenarios requiring high precision, such as acoustical simulations where small variations in boundary positioning can significantly affect the results. Conversely, a coarser threshold may be suitable for applications with less stringent accuracy requirements, such as visualization or preliminary design studies. The threshold can also be adjusted dynamically based on the scale of the model or the level of detail required, providing flexibility to adapt the method to different use cases. By using proximity as a measurable criterion, the method can ensure a systematic approach to identifying SPDSs that are spatially relevant to the at least one indoor boundary.

[0045] In some embodiments, a predetermined distance threshold used for defining proximity between a spatial partitioning data structure (SPDS) and at least one indoor boundary may be defined with reference to a spatial resolution of the SPDSs used to fill an inner volume. Each SPDS may be associated with a characteristic resolution length, such as a voxel edge length in the case of voxel-based SPDSs, or another representative linearP7391 PC00

[0046] 8

[0047] dimension in the case of other types of SPDSs. In such embodiments, the predetermined distance threshold may be expressed as a multiple of this characteristic resolution length. This may reflect that discretization of the inner volume using SPDSs can approximate an original geometry within a tolerance related to the SPDS resolution, such that SPDSs representing an interior of an indoor volume may deviate from a corresponding boundary geometry by an amount related to the SPDS resolution.

[0048] In some embodiments, the predetermined distance threshold may be selected as a multiple of the SPDS resolution length, for example as one or more times the resolution length. Byway of example, when voxel-based SPDSs are used, the distance threshold may correspond to approximately one, two, or three voxel edge lengths. Smaller multiples may be associated with a more conservative boundary association, whereas larger multiples may allow for a more permissive association in cases where the geometry includes gaps, noise, or incomplete boundary representations. The selected multiple may be adjustable or configurable, for example depending on characteristics of the geometry or on an intended downstream application, such as acoustical simulation.

[0049] In some embodiments, the spatial resolution of the SPDSs may be selected with regard to a scale or volume of a 3D model of a building complex or of at least one indoor volume. For example, a relatively coarser SPDS resolution may be used for larger-scale building models, whereas a relatively finer SPDS resolution may be used for smaller indoor volumes. In some implementations, the SPDS resolution may fall within a range suitable for building-scale models, for example between approximately 0.03 meters and 2 meters, although other resolutions may also be used.

[0050] In further embodiments, the SPDS resolution may be determined based on a characteristic dimension derived from the inner volume, such as a cube root of the inner volume. In such cases, the predetermined distance threshold may be derived from the selected SPDS resolution in any of the manners described herein. This approach may allow the discretization resolution and the associated proximity threshold to vary with the size of the indoor volume, which may be useful for balancing spatial accuracy and computational performance across different building scales.

[0051] In another embodiment, the geometric adjacency is defined as a spatial relationship where an SPDS and the at least one indoor boundary are in direct contact or share at least one geometric feature, such as an edge, vertex, or face, within a predefinedP7391 PC00

[0052] 9

[0053] tolerance. The geometric adjacency between an SPDS and the at least one indoor boundary may be defined as a spatial relationship where the SPDS and the boundary are in direct contact or share at least one geometric feature, such as an edge, vertex, or face, within a predefined tolerance. This definition can establish a topological criterion for correlating SPDSs with the at least one indoor boundary, ensuring that the correlation accounts for shared or intersecting geometry. For example, an SPDS may be considered geometrically adjacent to at least one indoor boundary if its surface intersects or overlaps with the boundary surface. In the context of a voxel grid, adjacency may be determined by examining whether the voxel faces, edges, or corners directly align with or intersect the at least one boundary.

[0054] The predefined tolerance provides flexibility in handling geometries with minor misalignments or gaps. For instance, in cases where the at least one indoor boundary surface is not perfectly aligned with the SPDSs, a small tolerance can allow for slight discrepancies while still recognizing the adjacency relationship. This tolerance may be defined based on the resolution of the 3D model, the dimensions of the SPDSs, or the computational requirements of the application. In some implementations, adjacency may also be evaluated using hierarchical or multi-resolution approaches, such as octrees, which can identify adjacency relationships at varying levels of granularity.

[0055] By defining proximity and geometric adjacency with these criteria, the method enables robust and precise spatial correlation between SPDSs and indoor boundaries. These definitions ensure that the reconstructed 3D model accurately represents the relationships between the volumetric structures and the boundaries, which is particularly important for applications like acoustical simulations, where boundary interactions influence sound propagation. The use of clear and quantifiable criteria for proximity and adjacency enhances the reliability of the reconstruction process while maintaining flexibility to adapt to different geometries and computational scenarios. This approach ensures that the method can be applied effectively across a wide range of building models and use cases.

[0056] In the context of the present disclosure, the at least one indoor boundary can refer to a surface that separates two distinct indoor volumes within the 3D model of the building complex. This may correspond to partition walls, structural elements, or other enclosures that define separate rooms or functional areas within the indoor space. The outdoor boundary, on the other hand, may refer to the interface between the innerP7391 PC00

[0057] 10

[0058] volume and the outdoor volume, which represents the outermost enclosure of the building complex. The outdoor boundary may include exterior walls, facades, roofs, or other architectural components that define the transition between enclosed indoor spaces and the external environment.

[0059] The 3D model of a building complex can be obtained from a real building through various techniques and data sources, which may be chosen depending on the available resources, desired level of detail, and purpose of the reconstruction. In one approach, the 3D model may be generated using data acquired through laser scanning or LiDAR, where a scanner emits laser pulses and measures the reflected signals to create a point cloud that represents the spatial geometry of the building. Alternatively, photogrammetry can be used, wherein multiple photographs of the building are taken from different angles and processed through computational algorithms to generate a 3D representation. This technique may be particularly useful for capturing both the exterior and interior geometries of the building. In another method, the 3D model may be obtained from as-built architectural or engineering plans, which may include CAD drawings or Building Information Modeling (BIM) files such as those in the Industry Foundation Classes (I FC) format. These files can provide structured and metadata-rich representations of the building, which may include not only geometric details but also material properties and functional annotations.

[0060] In one embodiment, the 3D model of a building complex may be obtained by combining multiple data sources. For instance, LiDAR data capturing large-scale spatial information can be augmented with high-resolution photogrammetry to enhance surface details. Additionally, existing BIM or CAD models can be updated or corrected using real-world measurements from surveying tools. The integration of multiple data sources may allow for a comprehensive and accurate representation of the building, suitable for diverse applications, including acoustical analysis. In certain cases, simpler data sources, such as manual measurements or 2D blueprints, can be used as a basis for generating the 3D model, though these methods may require additional assumptions or extrapolation to create a complete volumetric representation.

[0061] Another option for acquiring a 3D model is photogrammetry, where overlapping 2D images of the building or its components, such as 3D models of elements comprised in the building, are captured using high-resolution cameras. These images are processed using computational algorithms that triangulate spatial coordinates, reconstructing a 3DP7391 PC00

[0062] 11

[0063] model that accurately represents the building complex or the element. This method is well-suited for capturing fine geometric details and surface textures and can be combined with LiDAR to improve precision. Structured light scanning may also be used, where patterns of light are projected onto the surfaces, and the deformations of the light patterns are analyzed to create 3D models. This approach is beneficial when capturing objects with intricate details, such as furniture or small architectural elements. For real-time or dynamic mapping, depth-sensing cameras, such as Time-of-Flight (ToF) sensors, can be employed. These cameras measure the time taken for light to reflect off surfaces, providing rapid, approximate 3D models of rooms or elements, such as objects. This technique is effective in scenarios that require real-time mapping, such as on construction sites. Additionally, Building Information Modeling (BIM) models may be used as an initial input, supplemented or validated through on-site measurements obtained using techniques like laser scanning or photogrammetry. This ensures that the 3D representation accurately reflects any structural changes or deviations from the initial design.

[0064] Manual measurements of building dimensions, taken using traditional surveying tools such as laser rangefinders or total stations, can also be digitized and used to construct 3D models through computer-aided design (CAD) systems. This method can remain useful in environments where automated scanning methods may not be feasible. For individual elements or objects within the building, such as furniture or equipment, preexisting 3D models from databases may be used. These models can come from libraries containing digitized scans or be generated through object-specific scanning using handheld devices like structured light scanners.

[0065] For large-scale building exteriors, drones equipped with LiDAR or photogrammetry systems may be employed to capture roof structures, facades, and surrounding outdoor areas. This data can be integrated with internal scans to produce a complete 3D representation of the building complex. In certain cases, multi-sensor fusion may be applied, where data from different acquisition methods, such as LiDAR, photogrammetry, and depth cameras, are combined to create a highly accurate and comprehensive 3D model.

[0066] Drones can also be used to scan details of large indoor environments such as warehouses or concert halls.P7391 PC00

[0067] 12

[0068] These flexible acquisition techniques provide the necessary adaptability to handle various building environments and objects, whether small-scale furniture or expansive architectural structures. By accurately capturing real-world spatial and structural features, these inputs ensure that the 3D model of a building complex and / or the 3D model of an element can be accurately represented, enabling efficient computational processes.

[0069] In one embodiment, the computer-implemented method for reconstructing a 3D model of a building complex is configured to be suitable for acoustic simulation. In this context, the reconstructed 3D model is optimized to represent the air volumes and boundaries within the building complex that can be relevant for sound propagation, reflection, and absorption. The reconstructed 3D model may include features such as watertight indoor volumes and simplified geometries, ensuring that the spaces can be accurately analysed using wave-based or ray-based acoustical solvers. By focusing on the acoustically significant elements of the building geometry, the method may reduce computational complexity while maintaining the precision required for acoustical simulations. Variations of this implementation may further support customization of the reconstruction process based on specific frequency ranges or simulation goals, allowing for increased flexibility in handling different acoustical applications.

[0070] In another embodiment, the 3D representation of the building complex may be derived from various types of input data, including point cloud data, mesh models, polygonal models, wireframe models, voxel-based models, Non-Uniform Rational B-Splines (NURBS) models, or complete 3D scenes. These data sources can provide flexibility in adapting the method to diverse input formats, making it applicable across a wide range of use cases. For example, point cloud data obtained through laser scanning may represent the geometry as a dense set of points, while mesh or polygonal models may define surfaces and boundaries using vertices and edges. Wireframe models may provide a lightweight representation of the geometry, whereas voxel-based models discretize the volume into a structured grid of uniform elements. NURBS models, on the other hand, enable a smooth and parametric representation of surfaces, quite often used in architectural and design software. By supporting these formats, the method ensures compatibility with various tools and technologies used in the construction and design industries, enabling seamless integration with existing workflows.P7391 PC00

[0071] 13

[0072] In a further embodiment, the 3D representation of the building complex may be extracted from a Building Information Modeling (BIM) model, such as an Industry Foundation Classes (I FC) file. Alternatively, the method may support other file formats commonly used in design and visualization workflows, including Drawing Exchange Format (DXF), Universal Scene Description (USD), and GL Transmission Format (GLTF). These formats provide different levels of detail and metadata, potentially allowing the method to adapt to specific requirements of the reconstruction process. For example, IFC files may include rich metadata about the building components, such as material properties and spatial annotations, while DXF files primarily provide geometric information. USD and GLTF files may focus on the efficient representation and transmission of 3D models, enabling compatibility with visualization and virtual reality tools. Other types of file format can be supported, such as OBJ, 3DM or SKP. The .obj or OBJ file format, also known as Wavefront Object File, is a widely-used standard for representing 3D geometry. It can store information such as vertices, edges, faces, and textures of 3D objects or 3D models and is commonly used to exchange models between various software applications. The .3DM file format, also referred to as the Rhino 3D Model File, is associated with Rhinoceros (Rhino), a 3D modeling software frequently used in architecture and industrial design. The .3DM format supports precise representations of complex surfaces and geometries, often using Non-Uniform Rational B-Splines (NURBS), which allow for smooth curves and intricate designs. This format can be beneficial in contexts requiring high-precision 3D models. The .SKP file format is associated with SketchUp, a 3D modeling tool commonly used for conceptual designs and architectural visualization. The .SKP format stores 3D objects and related data, including textures and layers, making it effective for modeling large structures and architectural elements in an intuitive manner. Due to its simplicity, it can be employed in architectural projects, interior design, and construction. By accommodating these file types, the method may facilitate the processing of building models from various software platforms, improving interoperability and broadening the scope of applications.

[0073] In one embodiment, spatial partitioning data structures (SPDSs) may represent discrete units used to subdivide and organize the geometry of the inner volume of a 3D model. SPDSs can be computationally efficient structures designed to discretize complex spatial configurations, enabling precise representation and reconstruction of indoor volumes. The SPDSs may be adaptable to various geometrical configurations,P7391 PC00

[0074] 14

[0075] and their specific form can depend on the application, computational requirements, and type of input data being processed.

[0076] SPDSs may take several forms, including, but not limited to voxels, wherein SPDSs can be implemented as volumetric pixels, which divide the 3D space into a uniform grid of cubes. Each voxel may represent a small unit of volume, with a defined position and size. Voxels are particularly suitable for scenarios where the geometry needs to be discretized into manageable units for operations such as filling volumes or correlating boundaries. The uniform structure of voxels allows for straightforward computations, making them advantageous for applications like acoustical simulations or collision detection. SPDSs may also be structured as hierarchical spatial partitioning schemes, such as octrees. An octree divides the 3D space into progressively smaller regions, where each parent node is subdivided into eight child nodes. This structure can be advantageous for representing varying levels of detail, as denser regions of the model can be subdivided further, while less complex regions can remain coarser. Octrees may provide an efficient way to store and process spatial data, particularly in cases where the input geometry has uneven complexity or density. SPDSs may take the form of tetrahedral elements, which divide the volume into tetrahedra (four-sided pyramidal structures). Tetrahedral meshes may be beneficial for representing irregular or curved geometries, as they can adapt to complex surfaces and volumes more flexibly than regular grids. This type of SPDS may be especially relevant when higher accuracy is required in defining the boundaries of indoor volumes. In some cases, SPDSs may utilize adaptive grid structures, where the resolution of the grid varies based on the complexity of the geometry. For example, finer grids may be applied near intricate boundaries or critical areas, while coarser grids may be used in simpler regions.

[0077] Adaptive grids can balance computational efficiency with accuracy, making them suitable for large-scale building models. SPDSs can also be defined as convex polyhedra, where each unit represents a convex volume that can conform to the local geometry of the indoor space. This approach may be advantageous in cases where the model includes irregular or angular surfaces that are difficult to approximate with standard grid-based methods.

[0078] SPDSs may be used not only to discretize the inner volume but also to store associated data, such as spatial coordinates, material properties, or proximity relationships. For example, each SPDS may store information about its distance to the nearest boundary, its neighboring SPDSs, or its role in forming coherent patches. TheP7391 PC00

[0079] 15

[0080] coherent patches of SPDSs may then be used to represent contiguous indoor spaces, providing a structured and computationally manageable representation of the building’s interior.

[0081] By supporting multiple forms, SPDSs may adapt to a wide variety of geometries and use cases, ensuring that the method can handle diverse 3D input representations while maintaining flexibility and accuracy in reconstructing indoor volumes. This versatility may allow the method to be tailored to specific applications, ranging from acoustical simulations to structural analysis or energy modeling.

[0082] In one embodiment, the at least one space proxy may be a SPDS space representation or a SPDS-based proxy. The space proxy may thus be defined directly in terms of the spatial partitioning data structures (SPDSs) that may fill the inner volume of the at least one indoor volume. A SPDS space representation may comprise a group of contiguous SPDSs, such as voxels or other volume elements, that together approximate or delineate the interior of a space within the building complex. This grouping may be based on spatial criteria such as contiguity, proximity, and topological adjacency, allowing the formation of a coherent patch that can act as a computational stand-in for the actual indoor volume.

[0083] By defining the space proxy in terms of the SPDSs themselves, the method can allow for a highly adaptable and resolution-independent representation of indoor spaces. For example, in a voxel-based implementation, the space proxy may consist of a connected region of voxels within the filled inner volume, selected according to geometric or semantic criteria. These criteria may include overlap with predefined space annotations, enclosure by indoor boundaries, or consistency with structural segmentation.

[0084] This SPDS-based approach can enable a flexible and computation-friendly means of defining and manipulating spatial regions within the 3D model, particularly in cases where the original geometry is incomplete, noisy, or lacks explicit annotations. The SPDS space proxy may be used as a core input for further steps in the method, such as boundary correlation, intersection analysis, or acoustical simulation. Since SPDSs can be regular or hierarchical in nature (e.g., based on uniform grids or octrees), the SPDS-based space proxy allows the method to adapt to varying levels of geometric detail and computational requirements.P7391 PC00

[0085] 16

[0086] In addition, by anchoring the space proxy to the SPDS representation rather than to specific geometric surfaces or annotations, the method may operate robustly across different input formats and qualities of 3D representations. This enables practical use in scenarios involving as-built scans, partial models, or models reconstructed from real-world measurements, where traditional space definitions may be unavailable or unreliable. The use of a SPDS-based proxy may thus ensure a consistent and scalable representation of indoor spaces or volumes throughout the reconstruction process.

[0087] In another embodiment, the spatial correlation between the plurality of SPDSs and the at least one indoor boundary may be identified by generating at least one closest point query. This approach can involve analyzing the geometric relationships between SPDSs and the indoor boundaries, enabling precise identification of the boundaries. The closest point query may identify the nearest SPDS within the plurality, thereby correlating the spatial arrangement of the SPDSs with the corresponding indoor boundary. This process may provide a computationally efficient way of associating spatial data structures with the physical boundaries of the indoor volume, reducing manual intervention and potentially improving the accuracy of the reconstruction.

[0088] The at least one indoor volume may correspond to at least one room within the 3D model of the building complex. Rooms can represent functional indoor spaces, such as offices, living areas, or storage areas, that are delimited by walls, floors, ceilings, or other structural components, such as doors or windows. This configuration may allow the method to accurately model and reconstruct spaces that align with the intended use of the building. In some implementations, the method may be configured to handle variations in room shapes and sizes, enabling the reconstruction process to adapt to complex architectural designs.

[0089] In some cases, the at least one indoor volume may comprise at least two indoor volumes, such as two or more rooms. These indoor volumes may be adjacent, separated by partitions, such as walls, or interconnected through openings, such as doors or hallways. The ability to model and reconstruct multiple indoor volumes may facilitate applications that require detailed representation of the spatial relationships between rooms, such as acoustical simulations involving multi-room environments. By accommodating multiple indoor volumes, the method can potentially enhance its applicability to larger and more complex building complexes.P7391 PC00

[0090] 17

[0091] In one embodiment, the 3D model of a building complex may comprise at least one 3D model of an element, such as an object, arranged within the at least one indoor area. These elements may represent various features typically present in indoor environments, including furniture, appliances, fixtures, or other structural or non-structural components. The inclusion of such elements in the 3D model allows for a more comprehensive representation of the building complex, capturing both its architectural and functional characteristics. The elements may vary in complexity, ranging from simple geometric shapes, such as tables or chairs, to more intricate objects, such as light fixtures or machinery. This diversity may enable the method to reconstruct indoor spaces that are not only geometrically accurate but also reflective of their intended use.

[0092] In another embodiment, the method may further comprise applying a geometry simplification to the at least one 3D model of an element, thereby obtaining a simplified 3D model of the building complex. This simplification process may reduce the complexity of the elements without compromising their relevance for subsequent applications, such as acoustical simulations or visualization tasks. For instance, detailed objects like chairs or shelves may be replaced with bounding volumes, such as boxes or cylinders, that approximate their overall shape while removing intricate features that are computationally costly to process. This simplification can potentially enhance computational efficiency, especially in large-scale building models containing numerous elements.

[0093] The geometry simplification may be adaptable to different levels of abstraction, depending on the purpose of the reconstruction. In some scenarios, only non-essential details of the elements may be removed, preserving key features required for acoustical analysis or other specific simulations. In other cases, more aggressive simplification strategies may be applied, such as replacing high-polygon objects with low-polygon versions or applying decimation algorithms to reduce the number of surface elements. These approaches can provide flexibility, allowing the method to balance computational demands with the desired level of detail in the final 3D model. By integrating the simplification process, the method may also enable the 3D model to be more suitable for real-time rendering, such as in virtual reality or interactive visualization environments. Simplified elements may retain their spatial and functional context within the indoor area while avoiding excessive rendering loads. Furthermore,P7391 PC00

[0094] 18

[0095] the use of geometry simplification may improve the interoperability of the reconstructed 3D model with various simulation or design tools by ensuring that the model adheres to computational constraints or file size limitations imposed by these tools. This capability may extend the utility of the method to a wide range of practical applications, from architectural design and planning to acoustical performance analysis and immersive visualization.

[0096] In one embodiment, the geometry simplification extracts geometry metadata from the 3D representation of the building complex, such as at least one material property and / or at least one spatial annotation, such that the at least one 3D model of an element is simplified based on a category associated with the at least one 3D model of an element, and wherein the category is defined in the geometry metadata. This geometry metadata may provide useful information about the characteristics and attributes of the elements within the 3D model, enabling a more informed and context-aware simplification process. For example, material properties such as density, reflectivity, or acoustic absorption coefficients may inform decisions about how to simplify a particular object while retaining features that are relevant for downstream applications, such as acoustical simulations. Similarly, spatial annotations can describe the positioning, relationships, or functional roles of objects within the indoor volume, which may guide the simplification to preserve essential spatial characteristics.

[0097] Geometry simplification may be applied adaptively based on the nature of the object itself, as defined in the geometry metadata. For example, a chair, a sofa, or a fridge may require different levels of simplification depending on their usual geometries and relevance to the overall 3D model. Objects like chairs with complex details may be simplified into bounding volumes that approximate their outer contours, while a fridge with simpler geometry may retain more of its original features. This adaptive simplification process ensures that computational efficiency is achieved without unnecessarily compromising the fidelity of the model, particularly in areas critical for downstream applications like acoustical simulations.

[0098] In a preferred embodiment, the category is an element name, preferably an object name, such as chair, table or cable. The method may further simplify elements based on categories associated with the 3D model of an element, wherein these categories are defined in the geometry metadata. These categories may include general classifications, such as furniture, fixtures, or utilities, and may further specify object names, such as chair, table, or cable. Simplification based on categories may allow theP7391 PC00

[0099] 19

[0100] method to adopt tailored approaches for different types of elements. For instance, furniture might be replaced with bounding volumes, whereas cables might be excluded entirely if their presence is not relevant for the intended simulation or visualization. By leveraging these categories, the method may achieve a balance between maintaining the fidelity of the 3D model and reducing its complexity, ensuring that the model remains suitable for its intended purpose.

[0101] In another embodiment, the IFC file comprises at least one space annotation, and wherein the at least one space annotation is extracted from an IFC class named IFCSpace, wherein the IFCSpace is defined in the IFC file. When the 3D representation is extracted from an Industry Foundation Classes (IFC) file, the simplification process may utilize at least one space annotation defined in the IFC file. These space annotations may be extracted from the IFC class named IFCSpace, which provides standardized descriptions of spaces within the building model. By using these annotations, the method may correlate reconstructed volumes or elements with their intended function or designation, such as living rooms, offices, kitchens, bedrooms, or corridors. This correlation may enhance the accuracy of the reconstruction and provide additional context for simulation or analysis tasks. For example, the spatial annotations may assist in ensuring that acoustical simulations are performed on well-defined spaces, reflecting the intended use of the building complex. The ability to extract and utilize metadata, categories, and annotations from the 3D representation, including IFC files, may make the method highly versatile and interoperable. It may allow the method to seamlessly integrate with existing BIM workflows, leveraging the rich metadata available in IFC models to automate and optimize the reconstruction and simplification processes. This approach may reduce manual effort while improving the accuracy and relevance of the reconstructed 3D model for its intended applications, whether in simulation, visualization, or architectural analysis.

[0102] In one embodiment, the geometry simplification comprises a bounding volume replacement module, wherein the bounding volume replacement module replaces the at least one 3D model of an element with a simplified 3D model of an element, wherein the simplified 3D model of an element approximates outer contours of the at least one 3D model of an element. The geometry simplification may include a bounding volume replacement module, which is configured to replace the at least one 3D model of anP7391 PC00

[0103] 20

[0104] element with a simplified 3D model of the element. The simplified 3D model may approximate the outer contours of the original element, such as the original 3D model of the element, providing a computationally efficient representation while retaining the spatial relevance of the element within the 3D model of the building complex. By focusing on the outer contours, the bounding volume replacement ensures that the essential geometry needed for applications such as acoustical simulations or visualization is preserved, while unnecessary internal or intricate details are removed. This approach may reduce the complexity of the model, making it suitable for processing and analysis within computational or rendering constraints.

[0105] The simplified 3D model of an element may take various forms, depending on the type and geometry of the original element. In one embodiment, the simplified model may be a bounding box, which provides a rectangular approximation of the outer dimensions of the element. This option may be particularly useful for elements with rectilinear shapes, such as cabinets or desks, where a bounding box closely approximates their geometry. For elements with more irregular or curved shapes, the simplified model may be represented as a convex hull volume, which encloses the element with the smallest convex shape. This may allow for a tighter approximation of the geometry compared to a bounding box, particularly for complex objects like furniture or appliances.

[0106] In a preferred embodiment, the simplified 3D model of an element is a bounding box, a convex hull volume, a cylindrical volume, a spherical volume, an ellipsoidal volume, an oriented bounding box, a capsule volume, or any combinations thereof.

[0107] Other forms of simplified 3D models may include cylindrical volumes, which are suitable for elongated or rounded objects, such as columns, pipes, or table legs. In some cases, spherical volumes may be used to approximate 3D objects with roughly uniform dimensions in all directions, such as decorative fixtures or small appliances. Ellipsoidal volumes may be employed for elements that are elongated in one or more directions, such as vases or certain types of lighting fixtures. Oriented bounding boxes may be advantageous for objects that are not aligned with the primary axes, as they provide a rotated bounding box that better fits the geometry of the element. For objects with elongated shapes and rounded ends, capsule volumes may be used as an alternative, combining a cylindrical middle section with hemispherical ends.

[0108] In another embodiment, combinations of these bounding volumes may be used to approximate more complex geometries. For example, a combination of a bounding boxP7391 PC00

[0109] 21

[0110] and a convex hull may be applied to an element composed of multiple distinct parts, such as a chair with a backrest and armrests. This flexibility can allow the method to adapt to a wide variety of element geometries, ensuring that the simplification process remains effective across different types of objects.

[0111] By implementing a bounding volume replacement module, the method can streamline the reconstruction of 3D models while ensuring that acoustically relevant or spatially significant features are preserved. This approach may significantly reduce computational demands during simulations or rendering, especially in large-scale models containing numerous elements, without sacrificing the overall fidelity or functionality of the model.

[0112] In one embodiment, the geometry simplification comprises a plane projection module, wherein the plane projection module projects the at least one 3D model of an element to a 2D projection plane, thereby replacing the at least one 3D model of an element by at least one 2D model of an element. The geometry simplification may include a plane projection module, which can be configured to project at least one 3D model of an element onto a 2D projection plane. This process can replace the original 3D model with a 2D model of the element, reducing the dimensional complexity of the representation. The projection to a 2D plane may preserve essential spatial characteristics while removing the third dimension, making the simplified representation computationally lighter and easier to process. This approach can be particularly useful for elements that are naturally flat or thin, such as wall-mounted panels, floor tiles, or glass panes, where the additional complexity of the third dimension may not provide a meaningful contribution to downstream applications such as acoustical simulations or visualization.

[0113] The plane projection module may determine whether an element is suitable for such simplification by evaluating its dimensional characteristics. In one embodiment, the at least one 3D model of an element is determined to be nearly two-dimensional when an aspect ratio threshold is satisfied, such that a ratio between the largest and smallest dimensions of the at least one 3D model of an element is greater than or equal to a plane projection threshold. The at least one 3D model of an element may be identified as nearly two-dimensional when an aspect ratio threshold is satisfied. For instance, the ratio between the largest and smallest dimensions of the element may be greater than or equal to a predefined plane projection threshold. This evaluation ensures that theP7391 PC00

[0114] 22

[0115] simplification process can be applied selectively, focusing on elements whose geometrical properties allow for accurate representation in two dimensions without loss of critical spatial information. For example, a long and thin object, such as a metal plate or a large picture frame, may meet the aspect ratio threshold and be projected to a 2D plane, while objects with more balanced dimensions, such as cubes or spheres, may not undergo such simplification.

[0116] The plane projection process may be adapted to different types of projection methods, such as orthogonal or perspective projections, depending on the intended use of the simplified model. Orthogonal projection may be used when maintaining uniform scale and accurate alignment is critical, while perspective projection may be used to approximate the appearance of the object when viewed from a particular vantage point. Additionally, the resulting 2D representation of the 2D model of an element may include metadata or annotations to ensure that the context of the original 3D element is preserved, such as its position, orientation, or material properties.

[0117] By implementing a plane projection module, the method can significantly reduce the computational load associated with processing 3D models, especially in cases where many nearly flat or thin elements are present within the building complex. This simplification process may be particularly advantageous for large-scale simulations or visualizations, enabling efficient processing while maintaining the overall fidelity of the reconstructed 3D model. Moreover, the selective application of plane projection based on aspect ratio thresholds ensures that the simplification is both targeted and effective, preserving the functional and spatial relevance of the simplified elements.

[0118] In one embodiment, the geometry simplification comprises a decimation module, wherein the decimation module executes a decimation algorithm, wherein the decimation algorithm removes at least one surface element of the at least one 3D model of an element, thereby reducing the number of the at least one surface element of the at least one 3D model of an element. The geometry simplification may include a decimation module, which can be configured to execute a decimation algorithm for simplifying the 3D model of at least one element by reducing the number of surface elements. The decimation algorithm may remove selected surface elements, such as polygons, vertices, or edges, while preserving the overall shape and structure of the element. This process can provide a computationally efficient representation of theP7391 PC00

[0119] 23

[0120] element by reducing its geometric complexity without compromising its relevance to downstream applications, such as acoustical simulations or visualizations.

[0121] The decimation module may be designed or implemented to selectively remove surface elements based on specific criteria, such as curvature, edge length, or adjacency relationships. For example, flat or nearly flat surfaces may undergo more aggressive decimation, as their visual or functional representation can often be maintained with fewer polygons. In contrast, highly detailed or curved surfaces may undergo minimal decimation to preserve critical geometric features. The decimation algorithm may use techniques such as edge collapse, vertex removal, or triangle simplification to achieve the desired reduction in complexity.

[0122] In some cases, the decimation algorithm may consider additional factors, such as the acoustic relevance of the surface elements. For example, surfaces that are significant for sound reflection or absorption may retain higher levels of detail to ensure accurate simulation results, while less critical surfaces may be simplified more extensively. This context-aware simplification can balance computational efficiency with the fidelity required for specific applications.

[0123] The decimation module may be further adapted to dynamically adjust the level of simplification based on the overall complexity of the 3D model and the computational resources available. For instance, a large and intricate 3D model of a building complex may undergo more aggressive decimation to ensure efficient processing, while a smaller or simpler model may retain a higher degree of detail. Additionally, user-defined parameters, such as a maximum allowable polygon count or a target level of detail, may guide the decimation process, providing flexibility to tailor the simplification to specific project requirements.

[0124] By implementing a decimation module, the method can significantly reduce the computational demands associated with large and complex 3D models while maintaining their functional and visual relevance. This simplification process may enable faster processing and simulation, enhance compatibility with resource-constrained environments, and streamline workflows for applications such as acoustical analysis, rendering, or immersive visualization. The ability to selectively reduce surface complexity ensures that the reconstructed 3D model remains accurate and fit for its intended use, even after significant simplification.P7391 PC00

[0125] 24

[0126] In one embodiment, the geometry simplification comprises an outer shell voxelization module, wherein the outer shell voxelization module comprises identifying at least one outer shell of the at least one 3D model of an element, creating a set of SPDSs matching the at least one outer shell, and wherein the set of SPDSs defines at least one 3D model of an element outer boundary, and returning a triangulated representation of the at least one 3D model of an element outer boundary, thereby returning a representation of outer faces of the set of SPDSs. The outer shell voxelization module can be configured to identify the outer shell of at least one 3D model of an element and create a set of spatial partitioning data structures (SPDSs) that correspond to this outer shell. The outer shell can represent the external boundary of the 3D model of the element, encompassing its visible or acoustically relevant surfaces. By focusing on the outer shell, the module can streamline the representation of the element, reducing internal complexity while retaining the features critical for simulations or visualizations.

[0127] The outer shell voxelization module may first analyze the geometry of the 3D model of the element to detect its outer boundary surfaces. This detection may involve examining the spatial relationships and connectivity between surface elements to identify the external-facing parts of the model. Once the outer shell is identified, the module may generate a set of SPDSs, such as voxels or other volumetric units, that closely match the geometry of the outer shell. These SPDSs may discretize the outer boundary into manageable computational units, allowing for efficient representation and processing of the element's geometry.

[0128] To enhance the accuracy and usability of the representation, the module may return a triangulated representation of the outer shell. This triangulated representation may define the outer faces of the set of SPDSs, providing a structured and simplified geometric model suitable for downstream applications. The triangulation process may preserve the overall contours and structure of the original outer shell, ensuring that the simplified representation remains faithful to the original geometry.

[0129] The outer shell voxelization module may offer flexibility in its operation, supporting different levels of detail based on the specific requirements of the application. For instance, in scenarios where computational efficiency is prioritized, the voxelization process may use larger SPDSs to approximate the outer shell, resulting in a coarser representation. Conversely, for applications requiring high precision, such as acousticalP7391 PC00

[0130] 25

[0131] simulations or detailed visualizations, smaller SPDSs may be used to create a finer and more accurate representation of the outer shell.

[0132] The outer boundary of the at least one 3D model of an element often plays a critical role in sound reflection, absorption, and transmission, making it a key factor in acoustical analyses. The simplified outer shell representation may retain these acoustically relevant features while reducing the overall complexity of the model, enabling faster and more efficient simulations.

[0133] The implementation of the outer shell voxelization module may provide significant benefits in handling large and complex 3D models, such as complex 3D model of elements. By isolating and simplifying the outer shell, the method can reduce computational demands, improve processing speed, and streamline workflows for a wide range of applications. This approach ensures that the resulting 3D model of a building complex remains both accurate and efficient, making it suitable for diverse use cases, from acoustical analysis to architectural visualization.

[0134] In one embodiment, the geometry simplification comprises an exclusion module, wherein the exclusion module removes the at least one 3D model of an element if the at least one 3D model of an element is a 1 D object such as at least one cable, at least one wire, and / or at least one railing. The exclusion module can be configured to remove at least one 3D model of an element if the element is identified as a onedimensional (1 D) object. Examples of such 1 D objects may include but not be limited to cables, wires, and railings. These objects can contribute minimally to the overall geometry of the 3D model, particularly in applications like acoustical simulations, where their presence may not significantly impact the accuracy of the results. The exclusion module may identify and eliminate these elements from the model, thereby reducing computational complexity and improving processing efficiency.

[0135] The exclusion module may analyze the geometric properties of each element within the 3D model to determine whether it qualifies as a 1 D object. This determination can be based on parameters such as aspect ratio, cross-sectional size, or other geometric characteristics indicative of one-dimensionality. By removing elements that fall within this category, the exclusion module may streamline the 3D model, focusing computational resources on the geometries that are more relevant to the intended application.P7391 PC00

[0136] 26

[0137] In another embodiment, the 1 D object has a 1 D object length equal to or below a length threshold. The exclusion of 1 D objects may be further refined based on their length. The exclusion module may compare the length of a 1 D object to a predefined length threshold, and only remove those objects whose length is equal to or below this threshold. For instance, short cables or wires below a certain length may be excluded as they are unlikely to impact acoustical simulations or other applications, while longer elements such as railings or significant conduits may be retained. This length-based filtering ensures that the exclusion process is context-sensitive, removing only the objects that are deemed unnecessary for the specific purpose of the model.

[0138] The length threshold may be user-defined or automatically adjusted based on the requirements of the application. For example, in a large-scale building model, the threshold may be set higher to exclude minor 1 D elements, whereas in a smaller or more detailed model, the threshold may be lower to preserve more of the geometry. This flexibility allows the exclusion module to adapt to different scales and levels of detail, ensuring that the resulting 3D model is both efficient and fit for purpose.

[0139] By implementing an exclusion module, the method may provide significant benefits in handling complex 3D models that include numerous minor elements. The removal of 1 D objects that do not contribute meaningfully to the simulation or visualization can reduce file size, processing time, and computational load, making the method particularly advantageous for large-scale or resource-intensive applications. This approach ensures that the 3D model remains streamlined and focused, retaining only the elements that are relevant to the intended analysis or use case.

[0140] In one embodiment, the geometry simplification comprises a polygon replacement module, wherein the polygon replacement module comprises filling the at least one 3D model of an element with a secondary set of SPDSs, comparing a first and a second subsidiary set of SPDSs of the secondary set of SPDSs, wherein the first and the second subsidiary sets are neighbouring sets of SPDSs, thereby defining a dimensionality of the at least one 3D model of an element, extracting feature edges of the at least one 3D model of an element based on the dimensionality, and replacing each of the feature edges by a polygon representation. The polygon replacement module can be configured to process the geometry of at least one 3D model of an element by utilizing a secondary set of spatial partitioning data structures (SPDSs). The polygon replacement module may operate by filling the 3D model of an element withP7391 PC00

[0141] 27

[0142] the secondary set of SPDSs, enabling a structured representation of the element's geometry. By organizing the geometry into SPDSs, the module may facilitate efficient analysis and processing of the element’s structural characteristics.

[0143] The module may further compare a first subsidiary set of SPDSs to a second subsidiary set of SPDSs, where the two subsidiary sets are neighboring in the spatial structure. This comparison may enable the identification of key geometric features, such as transitions or changes in surface orientation, which can define the dimensionality of the 3D model of the element. For example, a significant difference in alignment or orientation between neighboring SPDSs may indicate the presence of a feature edge, such as a corner or boundary line.

[0144] In the context of the present disclosure, the term dimensionality can refer to a local geometric classification of a region within a 3D model of an element, determined by analyzing spatial relationships between neighbouring spatial partitioning data structures (SPDSs). Dimensionality can describe whether a given part of the model behaves locally as a point (0D), a line (1 D), a surface (2D), or a volume (3D). The determination is made by comparing subsidiary sets of SPDSs, subsets of the secondary set of SPDSs, based on their connectivity and distribution.

[0145] For example, when two neighbouring subsidiary sets of SPDSs share a minimal interface, such as a single line or edge, the local geometry may be identified as onedimensional, suggesting that the region resembles a beam or rod. If they share a broader, flat interface, such as a common face, the region may be classified as two-dimensional, indicating a planar surface. When the neighbouring sets are surrounded by and connected through multiple spatial faces in three dimensions, the region may be considered three-dimensional, corresponding to a solid body.

[0146] This classification can be used to guide the simplification process. By understanding the local dimensionality of the 3D model, the method can identify feature edges, transitions between different dimensional regions or geometric discontinuities, and apply appropriate polygonal substitutions that preserve key structural characteristics while reducing geometric complexity.

[0147] In a preferred embodiment, the feature edges are defined by an element boundary between two faces of the at least 3D model of an element.P7391 PC00

[0148] 28

[0149] Once the dimensionality of the element is defined, the module may extract feature edges based on the comparison as defined herein. These feature edges can represent important boundaries or transitions within the geometry, such as the edges formed between adjacent surfaces or faces. Feature edges may be defined by at least one element boundary between two faces of the at least one 3D model of an element, enabling the module to identify the structural outline or contour of the 3D model with precision. This process may allow the module to isolate the most geometrically significant aspects of the element while omitting less relevant details.

[0150] After extracting the feature edges, the module may replace them with polygon representations, thereby simplifying the geometry of the element. The resulting polygon representation may maintain the structural integrity of the 3D model of an element by preserving the outer contours and key features of the element while significantly reducing the complexity of its internal structure. This approach can balance computational efficiency with geometric accuracy, ensuring that the simplified model remains suitable for applications such as acoustical simulations, visualization, or other computational analyses.

[0151] In some implementations, the polygon replacement module may be configured to adapt the level of simplification based on the specific requirements of the application. For instance, elements with complex or intricate geometries may retain a higher number of feature edges, resulting in a more detailed polygon representation. Conversely, elements with simpler or repetitive geometries may undergo more aggressive simplification, leading to a reduced polygon count. This adaptability ensures that the method can be applied effectively across a wide range of elements with varying degrees of complexity.

[0152] By incorporating a polygon replacement module, the method may enhance the ability to handle large and complex 3D models efficiently, especially if these 3D models are numerous in the 3D model of a building complex. The extraction and simplification of feature edges can significantly reduce computational demands while preserving the essential spatial and structural characteristics of the elements. This approach may streamline workflows for large-scale projects, enabling faster processing and analysis without compromising the overall fidelity of the reconstructed 3D model.

[0153] In one embodiment, the method further comprises applying an intersection solver to the simplified 3D model of a building complex, wherein the intersection solver determinesP7391 PC00

[0154] 29

[0155] or establishes a plurality of coherent shells. The intersection solver can analyze and resolve geometric inconsistencies, such as overlapping or intersecting surfaces, to establish a plurality of coherent shells within the 3D model, and preferably within the at least 3D model of an element. Coherent shells may refer to unified, continuous boundaries that are free of gaps, overlaps, or other discontinuities, ensuring that the geometry is watertight and well-suited for downstream applications such as acoustical simulations, visualization, or structural analysis. By identifying and addressing these inconsistencies, the intersection solver can refine the 3D model or the at least one 3D model of an element to produce a more robust and computationally efficient representation.

[0156] In a preferred embodiment, the intersection solver analyzes a plurality of triangles comprised in the 3D model of a building complex, preferably comprised in the at least one indoor boundary and / or in the at least one 3D model of an element, and wherein the intersection solver reconstructs edges and vertices based on an intersection solver predetermined threshold, such that the plurality of triangles forms a plurality of coherent shells.

[0157] The intersection solver may operate by analyzing a plurality of triangles within the 3D model. These triangles can include those forming part of the indoor boundary or part of the elements present in the building complex, such as the at least one 3D model of an element. By examining the edges and vertices of these triangles, the solver may detect and address intersections or misalignments based on a predetermined threshold. For instance, the solver can identify areas where edges of neighboring triangles overlap or fail to connect properly and reconstruct the geometry to resolve these issues. The predetermined threshold may define acceptable tolerances for vertex alignment or edge proximity, ensuring that the reconstructed geometry adheres to specific accuracy requirements. The output of this process may result in a plurality of coherent shells, where the triangles form continuous and well-defined boundaries.

[0158] In another embodiment, the plurality of coherent shells are inner shells and / or outer shells, wherein the inner shells are at least one primary boundary of the at least one 3D model of an element, and the outer shells are the at least one indoor boundary. Inner shells may correspond to the primary boundaries of individual elements within the 3D model, such as the surfaces defining the geometry of furniture, fixtures, or appliances, which can be identified as the at least one 3D model of an element or elements. TheseP7391 PC00

[0159] 30

[0160] inner shells may play a significant role in applications like acoustical simulations, where they influence sound propagation and interaction. Outer shells, on the other hand, may define the indoor boundaries of the building complex, such as the walls, floors, or ceilings that enclose indoor spaces. By distinguishing between inner and outer shells, the method can accurately model both the internal and external geometries of the building, providing a comprehensive representation that is well-suited for detailed simulations or analyses.

[0161] The intersection solver may further enhance the structural integrity and accuracy of the 3D model by ensuring that both inner and outer shells are watertight and free of errors. This capability may be particularly advantageous in scenarios where the original input geometry contains gaps, overlaps, or other imperfections that would otherwise compromise the reliability of the model. By resolving these issues, the method may facilitate the generation of high-quality 3D models that are optimized for their intended applications, whether in acoustical analysis, virtual reality environments, or architectural design. By incorporating an intersection solver, the method provides a robust tool for refining and validating the geometry of simplified 3D models. The ability to establish coherent shells ensures that the model is both accurate and computationally efficient, enabling it to support a wide range of use cases while maintaining the integrity of its spatial and structural representation.

[0162] In one embodiment, the plurality of spatial partitioning data structures (SPDSs) is a plurality of volumetric pixels or a plurality of volume unitary elements. These SPDSs can serve as discrete, structured units for representing and organizing the inner volume of the 3D model. By discretizing the space into smaller, manageable components, the SPDSs may facilitate efficient processing, reconstruction, and analysis of complex geometries. The use of SPDSs may provide a flexible and adaptable framework for handling a wide variety of 3D geometries and applications.

[0163] The SPDSs may include, but are not limited to, voxel grids or hierarchical spatial structures such as octrees. A voxel grid can divide the inner volume of the 3D model or any 3D models as mentioned in this disclosure, into uniform cubic units, where each voxel represents a defined volume with consistent dimensions. This structure is particularly advantageous for its simplicity and regularity, enabling efficient computation for applications such as filling volumes, defining boundaries, or performing spatial queries. Octrees, on the other hand, offer a hierarchical approach, dividing the spaceP7391 PC00

[0164] 31

[0165] into progressively smaller subdivisions based on local geometry complexity. This adaptive structure allows for a finer resolution in areas with detailed geometry while maintaining coarser subdivisions in less complex regions, optimizing computational resources.

[0166] In another embodiment, the plurality of SPDSs is a plurality of volumetric elements, such as voxels, wherein the plurality of voxels are cubes, and wherein the cubes have a cube edge of 0.2 m, preferably 0.1 m, more preferably 0.05 m, even more preferably 0.4 m. The cube edge dimensions may vary based on the requirements of the application. For example, the cube edge may be 0.2 meters, 0.1 meters, 0.05 meters, 0.3 meters, or even 0.4 meters, depending on the desired level of detail and the scale of the 3D model. Smaller voxel sizes may be used for applications requiring higher precision, such as acoustical simulations in intricate spaces, while larger voxel sizes may suffice for less detailed representations, improving computational efficiency in large-scale models.

[0167] The broad definition of SPDSs may also encompass other forms of spatial partitioning structures, such as tetrahedral meshes, convex polyhedra, or adaptive grid systems. This versatility may ensure that the method can be tailored to various input geometries and use cases, from highly detailed architectural models to simplified representations for simulation or visualization. Additionally, the SPDSs may store metadata or attributes associated with their corresponding volumes, such as material properties, proximity relationships, or acoustic characteristics. This capability may enhance the method's ability to integrate spatial and contextual data into the reconstruction and analysis process.

[0168] By leveraging a broad and flexible definition of SPDSs, the method can ensure compatibility with diverse geometries and computational requirements. Whether implemented as voxel grids, octrees, or other partitioning structures, SPDSs can provide a structured approach to discretizing and managing the inner volume of a 3D model or any other 3D model such as the at least one 3D model of an element, enabling efficient processing and analysis across a wide range of applications. This adaptability may allow the method to optimize the trade-off between computational efficiency and geometric accuracy, ensuring that the reconstructed models remain both practical and precise for their intended use cases.P7391 PC00

[0169] 32

[0170] In one embodiment, the space proxy is named according to the at least one space annotation labelled in or derived from the 3D representation of the 3D model of a building complex, or wherein the space proxy is named with a generic indexing. The space proxy may be named according to at least one space annotation that is labeled in or derived from the 3D representation of the building complex. These space annotations may provide semantic information about the functional or spatial purpose of specific areas within the 3D model, such as "Living Room," "Office," or "Corridor". By naming the space proxy based on such annotations, the method can maintain consistency with the original 3D representation and enhance the interpretability of the reconstructed model. This naming approach may also facilitate downstream applications, such as acoustical simulations, where the identification of specific spaces plays a critical role in setting up simulation parameters.

[0171] Alternatively, when space annotations are not available or applicable, the space proxy may be named using a generic indexing scheme. In this case, each space proxy may be assigned a unique identifier, such as "Space 1", "Space 2", and so on. Generic indexing can provide a straightforward and flexible method for naming space proxies, ensuring that all reconstructed volumes are accounted for, even in cases where the input data lacks detailed semantic information. This dual naming approach, supporting both annotated labels and generic indexing, allows the method to adapt to various levels of metadata richness in the 3D representation.

[0172] In one embodiment, the method further comprises performing at least one raytracing analysis, wherein at least one ray trace identifies the at least one indoor boundary from the nearest SPDS comprised in the plurality of SPDSs, thereby identifying the at least one indoor boundary of the at least one indoor area comprised in the building complex. The method may further comprise performing at least one raytracing analysis to identify the indoor boundaries of the reconstructed model. Raytracing in this context can analyze the spatial relationships between the indoor boundaries and the SPDSs by detecting a line of sight between a given boundary (referred to as "indoor boundary A") and the nearest SPDS. The raytrace may determine whether the direct line of sight is obstructed by another boundary (referred to as "indoor boundary B"). If such an obstruction is detected, the raytrace may exclude indoor boundary A from being selected, ensuring that only the directly accessible or visible boundaries are identified. This process helps in refining the boundary identification by eliminating indirect or occluded boundaries, thereby improving the accuracy of the reconstruction.P7391 PC00

[0173] 33

[0174] The term "indoor boundary" in this context may specifically refer to the geometric components forming the boundary, such as triangles in a triangulated representation. Each triangle may represent a discrete surface element of the 3D model’s boundary geometry. Raytracing may thus operate at the level of individual triangles, using their spatial and angular relationships with the SPDSs to determine the nearest boundary elements. This approach ensures that the granularity of the analysis aligns with the geometric representation of the reconstructed model, allowing for precise identification and correlation of boundaries.

[0175] The raytracing process may involve emitting rays from a reference point, such as the centroid of a space proxy or a selected SPDS, and extending the rays outward to detect intersections with triangles representing the indoor boundaries. If a ray intersects multiple triangles, the process may evaluate the spatial order of these intersections to determine the nearest boundary. By discarding triangles that are not directly visible due to occlusion by other boundaries, the method can ensure that the reconstructed model accurately reflects the spatial configuration of the indoor areas. This implementation of raytracing may be particularly advantageous in scenarios involving complex or layered geometries, where direct visibility plays a critical role in defining functional spaces. For example, in acoustical simulations, the accuracy of sound propagation modeling depends on the correct identification of reflective and absorptive boundaries. Raytracing ensures that only the relevant, directly accessible boundaries are included in the reconstructed model, improving the fidelity and reliability of the simulation.

[0176] The method’s ability to leverage raytracing for boundary identification, combined with the precise representation of boundaries as triangles, can provide a robust framework for handling intricate geometries. This approach may ensure that the reconstructed 3D model is both computationally efficient and geometrically accurate, making it suitable for a wide range of applications, from acoustical analysis to architectural visualization. The present disclosure also relates to a computer-implemented method for generating an impulse response for a listening point comprised in at least one indoor volume as reconstructed according to the aforementioned method, wherein the method may comprise receiving a 3D model of the at least one indoor volume, the position of at least one sound source in the at least one indoor volume; and determining using a wave-based solver and / or a geometrical acoustics solver, an impulse response of aP7391 PC00

[0177] 34

[0178] wave-based and / or a ray-based propagation of an impulse emitted at the at least one sound source in the at least one indoor volume and received at the listening point. The method may comprise further receiving acoustic properties of at least one boundary in the at least one indoor volume.

[0179] In one embodiment, the method further comprises rendering a base audio signal by convolving the base audio signal with the impulse response, thereby creating a rendered audio signal. Convolution in this context refers to a mathematical operation where the characteristics of the impulse response are applied to the base audio signal, effectively embedding the acoustic properties of the reconstructed 3D space into the audio signal. The resulting rendered audio signal may accurately simulate how sound behaves and propagates within the specific indoor volume at the listening point. This process allows the acoustic properties of the space, such as reverberation, absorption, and reflection, to be faithfully incorporated into the rendered audio, enhancing its realism and spatial fidelity.

[0180] The convolution process may be implemented using digital signal processing techniques, where the impulse response acts as a filter that modifies the frequency and temporal characteristics of the base audio signal. For example, if the impulse response captures the reverberant qualities of a large hall, the rendered audio signal will reflect these characteristics, producing an immersive and realistic auditory experience. This capability may be particularly useful for applications in architectural acoustics, sound design, or virtual reality environments, where accurate audio rendering is essential to create a realistic sense of space.

[0181] In another embodiment, the rendered audio signal may provide an audio rendering of the base audio signal in the 3D model of the internal space at the listening point. This means that the rendered signal represents the sound as it would be perceived at the listening point within the reconstructed indoor space or reconstructed 3D model, taking into account the acoustic interactions between the sound source, the indoor boundaries, the at least one element and the surrounding environment. The rendered audio signal may include spatial cues, such as directional reflections or delays, that enhance the perception of depth and position within the space.

[0182] In the context of the present disclosure, the term internal space may be used interchangeably with the terms indoor volume or indoor space. The internal spaceP7391 PC00

[0183] 35

[0184] refers generally to any enclosed or semi-enclosed spatial region within a building complex that is delimited by at least one structural boundary. It corresponds to the portion of the 3D model that can be intended to represent a functional indoor area, such as a room, corridor, hall, or similar architectural compartment, and which is reconstructed by the computer-implemented method described herein.

[0185] The internal space may thus comprise the same regions identified and reconstructed as part of the at least one indoor volume, as defined herein. Whether referred to as an indoor volume, indoor space, or internal space, the intent is to designate a space that is at least partially surrounded by building elements and is spatially distinct from outdoor or external regions. The terminology may vary throughout the description depending on the context, but all such expressions are to be understood as referring to the same general concept of an enclosed or enclosed-like region suitable for further spatial analysis or acoustical simulation.

[0186] The rendered audio signal may be used in various applications requiring precise spatial audio rendering. For example, in virtual reality or gaming environments, the rendered signal can create an immersive experience by accurately simulating how sounds originate and propagate within the modeled space. In architectural acoustics, it may assist in evaluating the auditory qualities of a room or hall, enabling designers to optimize the space for specific acoustic purposes. Similarly, the rendered audio signal may be employed in audiovisual production or sound system design, ensuring that audio content aligns seamlessly with the spatial characteristics of the environment. By convolving the base audio signal with the impulse response and providing an audio rendering within the 3D model of the internal space, the method enables a detailed and realistic representation of sound behavior. This capability ensures that the auditory experience is not only computationally accurate but also perceptually immersive, supporting a wide range of technical and creative applications.

[0187] In one embodiment, the computer-implemented method further comprises generating a volumetric mesh of the at least one indoor volume, wherein the volumetric mesh is used for computing the wave-based solver, and wherein each mesh of the volumetric mesh has a predetermined mesh size. The volumetric mesh can be used as a computational framework for executing the wave-based solver. The volumetric mesh discretizes the indoor volume or the 3D model of interest, i.e., that would need an acoustic analysis, into smaller, finite elements, enabling numerical computations thatP7391 PC00

[0188] 36

[0189] simulate the propagation of sound waves within the space. Each element of the volumetric mesh may represent a specific portion of the indoor volume, facilitating precise modeling of acoustic interactions, such as wave reflections, absorptions, and diffractions. The mesh may have a predetermined mesh size, which influences the resolution and accuracy of the simulation.

[0190] The predetermined mesh size may be carefully selected based on the requirements of the simulation. A smaller mesh size may provide higher accuracy, particularly for modeling high-frequency sound waves, as it ensures that the wave behavior is captured with greater detail. Conversely, a larger mesh size may reduce computational complexity, making the simulation more efficient but potentially less precise. In some extreme cases, the simulation could eventually crash and / or not converge. This adaptability allows the method to balance computational demands with the desired level of accuracy, depending on the specific application.

[0191] In another embodiment, the smallest dimension of the at least one 3D model of an element as defined in the aforementioned method, is equal or larger than the predetermined mesh size. To avoid excessive computational overhead, the smallest dimension of any remaining 3D model of an element in the building complex may be equal to or larger than the predetermined mesh size. This ensures that excessively small objects, such as chairs, cables, or other minor elements, do not reduce the mesh size disproportionately. For example, if a small object with dimensions significantly smaller than the overall scale of the indoor volume is present in the 3D model, the mesh would need to adapt to this small size locally, resulting in an overly fine mesh in that area. This, in turn, could cause the simulation to require excessive computational resources and time. By ensuring that small elements are either excluded, simplified, or appropriately scaled, the method can maintain a practical and efficient mesh size while preserving the overall fidelity of the 3D model.

[0192] In a further embodiment, the predetermined mesh size may be proportional to the overall volume of the 3D model of the building complex. By scaling the mesh size according to the building’s dimensions, the method can adapt the resolution of the volumetric mesh to the size of the indoor volume. For larger volumes, a proportionally larger mesh size may be applied to maintain computational efficiency, while smaller volumes may use a finer mesh to capture detailed acoustic behaviors. This proportionalP7391 PC00

[0193] 37

[0194] approach ensures that the method remains versatile and scalable, accommodating building complexes of varying sizes and complexities.

[0195] The generation of a volumetric mesh with an appropriate mesh size offers several technical advantages. By discretizing the indoor volume of the 3D model, the method provides a structured framework for solving complex wave equations, enabling accurate simulations of sound propagation within the space. The alignment of mesh size with the dimensions of the 3D model of the elements ensures that critical features are retained, enhancing the accuracy and reliability of the results. Additionally, by scaling the mesh size to the building volume, the method optimizes computational resources, making it suitable for a wide range of applications, from small-scale room acoustics to large-scale architectural simulations. This capability ensures that the method is both adaptable and efficient, supporting the precise analysis of acoustical behavior in diverse environments.

[0196] In one embodiment, the acoustic properties of the at least one boundary may be extracted or derived based on boundary names or other metadata comprised in the 3D representation of the building complex. The metadata may include, for example, material identifiers, object types, class names, semantic tags, or layer names associated with boundary surfaces in the model. These attributes can be mapped to known acoustic properties, such as frequency-dependent absorption coefficients, diffusion coefficients, or impedance values, by referencing one or more predefined lookup tables or material property databases.

[0197] Such mapping may rely on internally stored data or standardized acoustic material libraries, such as ISO 12354-1 tables, EN 12354 reference values, or proprietary datasets used in simulation tools. For instance, a boundary labeled “concrete wall” may be mapped to a corresponding entry in the database specifying its absorption coefficient at standard octave bands, while a label such as “glass partition” may be associated with frequency-specific transmission loss values. The system may implement this mapping through direct name matching, tag parsing, or rule-based inference, depending on the format and structure of the metadata.

[0198] In cases where the metadata is incomplete, missing, or ambiguous, the method may apply fallback strategies. For example, the system may use default acoustic properties for unknown boundaries, infer material type from geometric context (e.g., thin surface panels near openings may be presumed to be glass), or prompt the user to assignP7391 PC00

[0199] 38

[0200] values manually. This can ensure that the acoustic simulation remains complete and operational even in the absence of perfect metadata coverage, while still allowing for refinement and calibration when more detailed information becomes available.

[0201] The impulse response derived from the 3D model of the building complex, and as generated according to the method described herein, preferably with the reconstructed 3D model, can be used in numerous ways across a variety of domains. For example, it can be used to render sound by convolving the impulse response with a base audio signal, producing an auditory representation of how the sound propagates and interacts within the reconstructed space. This rendered sound can be used for immersive applications such as virtual reality, augmented reality, or gaming, where accurate spatial audio enhances the realism and user experience. Additionally, the impulse response can be used by acoustic engineers to evaluate the acoustical properties of a space, such as its reverberation time, clarity, or speech intelligibility. These analyses can help engineers optimize the design of a space for specific acoustical goals, such as improving the acoustics of a concert hall or ensuring adequate sound insulation in residential or commercial buildings.

[0202] The impulse response may also be applied in predictive modeling, where engineers use it to simulate the impact of potential design changes, such as the addition of sound-absorbing panels or the modification of room dimensions. It can be used to validate the compliance of a space with regulatory standards for noise control or sound performance. In some cases, the impulse response may be used for training or fine-tuning machine learning models that aim to predict acoustical behavior or classify spaces based on their acoustic properties. Beyond engineering applications, the impulse response may be useful in creative domains, such as audio production or sound design, where it can be employed to create effects that mimic real-world spaces or to enhance spatial audio in films or music.

[0203] By exploring a wide range of applications for the impulse response, the method provides a flexible framework that can adapt to various use cases. Whether the focus is on engineering analysis, creative applications, or immersive technologies, the method offers versatile tools for understanding and leveraging the acoustical properties of indoor spaces. This adaptability ensures its applicability across diverse industries and scenarios.P7391 PC00

[0204] 39

[0205] In one aspect of the present disclosure, the computer-implemented method for generating an impulse response for a listening point comprised in at least one indoor volume as described herein may involve integrating a secondary computer-implemented method for reconstructing a 3D model of a building complex. The term "secondary computer-implemented method" can refer to the steps involved in obtaining and reconstructing the indoor volumes of a building complex using spatial partitioning data structures (SPDSs) and space proxies, as defined in the present disclosure. This method may be referred to as secondary in the context of acoustical simulations, as it provides the reconstructed 3D geometry required for the primary task of generating the impulse response. However, it should be understood that this designation may not imply dependency or subordination in any specific implementation.

[0206] The secondary computer-implemented method can operate independently or be integrated directly into the acoustical simulation workflow. The steps of the secondary method may include obtaining a 3D representation of the building complex, filling the inner volume with SPDSs, applying space proxies, and reconstructing the indoor volumes based on spatial correlation criteria such as proximity or geometric adjacency. By providing accurate 3D models of indoor spaces or indoor volumes, the secondary method ensures that the subsequent acoustical simulation receives input data that faithfully represents the spatial environment of interest for the user, thereby improving the accuracy of the simulated impulse responses.

[0207] This flexible structure allows the methods to be implemented either sequentially or concurrently, depending on the computational requirements or application-specific needs. The designation of the geometry reconstruction as secondary can be intended to provide clarity within this disclosure but does not limit the implementation to any specific order or structure. For example, in some embodiments, the geometry reconstruction may be integrated directly into the acoustical simulation as part of a unified computational workflow.

[0208] Detailed description of the drawings

[0209] Fig. 1 shows a 3D model of a building complex 100, illustrating the spatial organization of indoor and outdoor volumes. The indoor volume 110 comprises enclosed spaces, including a first room 111 and a second room 112. These rooms represent individual indoor areas that may be relevant for acoustical analysis, spatial partitioning, or other computational processes related to the reconstruction of the 3D model.P7391 PC00

[0210] 40

[0211] The outdoor volume 113 is shown surrounding the indoor volume 110. The outdoor volume may correspond to external airspace, open areas adjacent to the building, or any other volume that is external to the enclosed indoor spaces. The boundaries between the indoor volume 110 and the outdoor volume 113 may be defined by walls, floors, ceilings, windows, doors, or other structural elements of the building complex. This figure provides a general representation of the spatial structure of the building complex, which serves as a foundation for further computational processing, such as spatial partitioning using SPDSs, and boundary identification.

[0212] Fig. 2 illustrates a 3D model of a building complex 200, illustrating how the indoor volume is processed using spatial partitioning data structures (SPDSs) 202. The SPDSs are used to discretize the space, preferably a fourth room, within the building complex, by forming structured computational representations that facilitate the reconstruction of the 3D model. The SPDSs may be implemented as volumetric elements such as voxels, octrees, or other partitioning structures. In Fig. 2, cubes are utilized as SPDSs.

[0213] A coherent patch of SPDSs 201 is shown within the indoor volume. The coherent patch represents a contiguous group of SPDSs that are in contact with each other, forming a structured representation of an enclosed space, which is the fourth room. The coherent patch may be used to define the spatial extent of an indoor volume and to facilitate boundary reconstruction based on proximity or geometric adjacency.

[0214] Additionally, the figure illustrates a third room 114 within the indoor volume of the building complex. The third room represents another enclosed indoor space, further demonstrating the method’s capability of handling multiple indoor volumes within a building complex. As it is shown in Fig. 2, only the fourth room has been filled with SPDSs, showing that the third room remains empty. The structured partitioning of space using SPDSs allows for an accurate and computationally efficient representation of indoor geometries, such as the inner volume of the fourth room as shown in Fig. 2, which may be used for acoustical simulations, spatial analysis, or other computational applications. Fig. 2 provides a visualization of how the method for reconstructing a 3D model of a building complex as described herein utilizes SPDSs and coherent patches to process and reconstruct 3D models of indoor volumes, ensuring accurate spatial representation and efficient computational handling of complex building geometries.P7391 PC00

[0215] 41

[0216] Fig. 3A illustrates a close-up view of a coherent patch of SPDSs 201 , illustrating how the spatial partitioning data structures conform to the geometry of the building complex. The coherent patch of SPDSs 201 forms a structured representation of an enclosed indoor volume, ensuring that the discretized representation aligns with the existing architectural boundaries.

[0217] Fig. 3B further illustrates the spatial relationship between the coherent patch of SPDSs 201 and the boundaries, defining the indoor and outdoor volumes. In Fig. 3B, the indoor boundary 302 represents a surface separating two adjacent indoor volumes, such as a wall or partition between different rooms or functional areas within the building complex. As can be seen in Fig. 3B, the coherent patch of SPDSs is present in one of the two indoor volumes, showing the actual correct identification of an indoor volume. The outdoor boundary 301 defines the separation between the inner volume and the outdoor volume, encompassing surfaces such as exterior walls, facades, or other structural elements that enclose the building. The relationship between these boundaries and the SPDSs ensures that the reconstructed 3D model accurately captures the segmentation of indoor and outdoor spaces, as well as different indoor volumes.

[0218] Figs. 3A-B provide insight into how the method processes and reconstructs indoor and outdoor geometries using SPDSs, ensuring accurate spatial representation and facilitating downstream applications such as acoustical analysis or computational simulations.

[0219] Fig. 4 illustrates the functioning of the closest point query scheme, which may be an integral part of identifying spatial boundaries during the reconstruction of indoor volumes in 3D models of building complexes. The scheme can operate by analyzing a set of SPDSs 401 , which may represent a coherent patch of spatial partitioning data structures (e.g., voxels or other unitary elements) used to discretize the geometry of the indoor space. The goal of the closest point query scheme is to accurately determine the indoor boundary spatially correlated to the set of SPDSs, ensuring that the reconstructed indoor volume is geometrically consistent with the original model. The correct closest point query 404 is shown selecting the first boundary 402, which is the nearest geometric boundary to the set of SPDSs. In contrast, incorrect closest point queries 405 are shown mistakenly selecting the second boundary 403, which is farther from the SPDSs and would result in erroneous boundary identification if chosen.P7391 PC00

[0220] 42

[0221] The scheme relies on proximity-based criteria to determine the closest boundary. For example, it may compute the Euclidean distance between the SPDSs and the candidate boundaries and select the boundary with the shortest distance. However, simple distance measurements may be insufficient in cases where other boundaries (such as the second boundary 403) obstruct the line of sight or are misinterpreted due to complex geometry. To address this, the scheme can incorporate additional checks, such as raytracing or geometric filtering, to ensure that only unobstructed, spatially relevant boundaries are selected.

[0222] One advantage of this scheme is its ability to minimize errors in boundary selection, which is crucial for maintaining the accuracy of the reconstructed indoor volumes. Incorrect boundary selection could lead to overlapping or missing boundaries, which would negatively impact applications such as acoustical simulations, where precise boundary placement is essential for accurately modeling sound wave propagation and reflections. By employing correct closest point queries, the method ensures that boundaries are correctly assembled, providing a watertight and geometrically consistent 3D model. Furthermore, the scheme is scalable and can be applied to complex 3D models with multiple indoor and outdoor boundaries. The method can dynamically adapt to varying spatial configurations by adjusting the query parameters or incorporating hierarchical data structures (e.g., octrees or bounding volume hierarchies) to efficiently process large datasets. This adaptability enables the method to handle diverse building environments, from small interior rooms to large architectural complexes, while maintaining computational efficiency and accuracy.

[0223] Fig. 5 demonstrates how various 3D models of elements 501 , such as chairs, stools, and other pieces of furniture, can be simplified using a plane projection module and / or a polygon replacement module to generate their corresponding plane projections 502. The original 3D models 501 represent detailed representations of the furniture, including legs, seats, backs, and structural components, which may be computationally expensive to process in large-scale simulations. The plane projection module and / or the polygon replacement module identifies suitable 3D models for simplification based on their geometric characteristics, such as being nearly two-dimensional or having a dominant planar surface. For example, objects like flat stools or simple chairs may exhibit aspect ratios or thicknesses that meet predefined thresholds for projection. Once identified, the module projects the 3D geometry of these objects onto a 2D plane, effectively collapsing the third dimension while preserving essential spatialP7391 PC00

[0224] 43

[0225] characteristics such as the object’s outline, orientation, and size. The resulting plane projections 502 provide a simplified representation of the furniture, which retains its reflective or acoustic properties without the need to process the detailed 3D structure. This simplification is particularly advantageous in computational simulations, such as acoustical analysis, where large environments with numerous objects require efficient handling of complex geometries. By reducing the dimensional complexity, the plane projection module and / or the polygon replacement module ensures that computational resources are allocated efficiently while maintaining sufficient accuracy for sound reflections, occlusions, or other interactions. The simplified plane projections 502 maintain the furniture’s spatial footprint and alignment within the 3D model of the building complex, enabling accurate simulations with minimal computational overhead. Additionally, the method ensures that the selection criteria for applying the plane projection module can be adjusted based on the specific needs of the simulation. For example, thinner objects may undergo more aggressive projection, while bulkier objects may retain additional geometric details if they play a significant role in the simulation. This flexibility allows the method to optimize the balance between accuracy and computational efficiency in large-scale environments.

[0226] Fig. 6 illustrates a geometry simplification process involving voxelization, plane projection and / or polygon replacement for various 3D models of furniture. The models include a chair with armrests 600, a table 610, and a piece of furniture 620. The simplification process begins by voxelizing the original 3D models, which discretizes the geometry into a grid of volumetric spatial partitioning data structures (SPDSs), such as voxels. The voxelization captures the external structure and spatial properties of the models, enabling subsequent simplification steps to be performed efficiently. The voxelization process evaluates the geometric properties of the 3D models to determine whether plane projection is a suitable simplification technique. For example, if the voxel grid reveals that a model is nearly two-dimensional (such as thin, planar objects or objects with dominant flat surfaces) the plane projection module may be triggered. The chair 600, after voxelization, retains much of its 3D structure, but some of its details may meet the projection threshold, and the chair 600 is then simplified to a simple polygon representation. The table 610, however, is identified as a suitable candidate for plane projection based on its voxelization results, as the tabletop forms a dominant planar surface. The plane projection module projects the geometry of the table onto a 2D plane, reducing the dimensional complexity while preserving the spatial footprintP7391 PC00

[0227] 44

[0228] and key reflective properties of the table’s surface. Similarly, the piece of furniture 620, which includes flat panels or shelving, undergoes plane projection after voxelization since some of its dimensional characteristics meet the projection threshold.

[0229] This combined voxelization and plane projection approach optimizes the geometry simplification by dynamically selecting the appropriate method based on the object’s structure. The voxelization step ensures that only models suitable for planar simplification are processed using the plane projection module, preventing unnecessary or inaccurate simplification of complex objects. This process reduces computational overhead by discarding non-critical details while retaining the essential characteristics of the objects needed for acoustical simulations or spatial analysis. The advantages of this approach are twofold. The voxelization process enables systematic evaluation of the geometry, ensuring that simplification methods are applied selectively, and the plane projection reduces the processing burden for large-scale environments containing numerous objects, without sacrificing critical spatial properties. This adaptability allows the method to handle a variety of object types within a 3D model of a building complex, making it suitable for diverse applications such as architectural modeling, acoustical simulations, and real-time rendering of audio sound files.

[0230] Figs. 7A-B illustrate how a complex 3D model of an element, such as a vase, can be simplified using a bounding volume replacement module. In Fig. 7A, the original 3D model of the vase has a detailed geometry, with curved surfaces and intricate contours that are computationally expensive to process. The bounding volume replacement module generates a simplified representation, shown in Fig. 7B, where the vase is replaced by a convex bounding volume. The bounding volume approximates the external shape of the vase while reducing the overall complexity by eliminating unnecessary details. This simplification offers computational advantages, particularly for large-scale simulations, by reducing the number of geometric components that need to be processed. For example, in acoustical simulations, where the outer shape of an element may be more significant than its internal details, the bounding volume representation ensures that the element's acoustic impact is captured efficiently while minimizing computational overhead. The bounding volume shown in Fig. 7B retains the general external dimensions of the vase, allowing for accurate reflections and interactions with sound waves without the need for processing fine details.P7391 PC00

[0231] 45

[0232] Figs. 8A-B illustrates the concept of bounding volume replacement with a more structurally complex object, a cabinet. Fig. 8A depicts the original 3D model of the cabinet, which includes specific features such as a geometrically complex front door. The detailed internal and external geometries of the cabinet can increase computational demand when the 3D model is used directly in simulations or visualizations. The bounding volume replacement module simplifies this geometry, as shown in Fig. 8B, by generating a bounding volume that approximates the overall outer shape of the cabinet. In this example, the bounding volume is configured to maintain the key dimensions and spatial footprint of the cabinet while discarding smaller, computationally expensive details such as the surface features of the front door or shelves. This replacement ensures that the essential acoustic or spatial characteristics of the cabinet are preserved, making the simplified representation suitable for simulations where only the overall shape affects the outcome. For instance, in sound wave simulations, the bounding volume effectively models how the cabinet reflects or absorbs sound while avoiding the computational costs of modeling its internal structure. The bounding volume replacement shown in these figures demonstrates how the method balances simplification and accuracy, ensuring that computational efficiency is achieved without significantly compromising the fidelity of the 3D model. By using bounding volumes, the method can handle large and complex environments with numerous objects while keeping simulations or visualizations manageable and scalable. This simplification strategy can be applied dynamically, allowing for adaptive refinement of the model depending on the simulation's precision requirements or available computational resources.

[0233] Figs. 9A-B illustrate the simplification of a window geometry using a plane projection module. Fig. 9A shows the window with its detailed 3D structure, including the frame and pane surfaces. The plane projection module simplifies this representation by projecting the 3D geometry onto a 2D plane, as shown in Fig. 9B, thereby creating a 2D model of the window. This projection retains key features of the window, such as its size, frame, and overall shape, while discarding the third dimension, which is unnecessary for many computational processes. The plane projection module is typically applied to objects that are nearly two-dimensional, such as windows, panels, or other thin structures. The module identifies suitable candidates for projection by evaluating dimensional properties such as aspect ratios or geometric thresholds. For example, a window with a dominant planar surface and minimal depth may meet theseP7391 PC00

[0234] 46

[0235] criteria and be projected into a 2D plane while preserving its reflective properties and spatial position within the 3D model. This simplification reduces computational complexity, making simulations involving large-scale environments more manageable without compromising accuracy in applications like acoustical simulations, where reflections off flat surfaces are key.

[0236] Figs. 10A-B illustrate the use of a polygon replacement module following outer shell voxelization to simplify the geometry of a chair, more specifically a stool. Fig. 10A shows the original 3D model of the chair, which includes specific geometric details such as the seat, legs, and connecting bars. The outer shell voxelization module first identifies the outer shell of the chair by discretizing its geometry into a set of volumetric spatial partitioning data structures (SPDSs), such as voxels. The outer shell voxelization focuses on capturing the object’s external surfaces while ignoring internal details that do not significantly affect computational outcomes. After voxelization, the polygon replacement module generates a simplified version of the chair, shown in Fig. 10B, by replacing the outer shell with a polygonal approximation. The module evaluates the dimensionality of the outer shell and extracts key feature edges before substituting the original geometry with a set of polygons that approximate the chair’s external shape. The resulting simplified representation retains the essential spatial footprint and structural characteristics of the chair while discarding minor details that would otherwise increase computational overhead.

[0237] This simplification process is particularly advantageous for applications like acoustical simulations, where the outer contours of objects significantly influence sound reflections and absorptions, but fine surface details have a negligible impact on the results. By replacing the original 3D model with a polygonal substitute, the method ensures that computational resources are allocated efficiently, enabling large and complex 3D environments to be processed without excessive computational burden. Figs. 11 A-B show the use of a decimation module to simplify a 3D model of an element by reducing its geometric complexity. In Fig. 11 A, the original 3D model contains a large number of surface elements, including detailed faces, edges, and vertices that define the object’s shape. The decimation module applies a decimation algorithm, which systematically reduces the number of these surface elements while preserving the overall structure and key geometric features of the model. The decimation process involves analyzing the geometry of the model and identifying redundant or non-criticalP7391 PC00

[0238] 47

[0239] surface elements, such as small or closely spaced triangles, that can be removed or merged without significantly affecting the object’s visual or functional representation. As shown in Fig. 11 B, the decimated model retains its essential shape and dimensions but has fewer surface elements, resulting in a simplified representation that is computationally more efficient to process.

[0240] The decimation module may use different techniques to achieve this reduction, such as collapsing edges, merging neighboring polygons, or selectively removing vertices. The level of decimation can be controlled based on predefined thresholds or tolerances, allowing for dynamic adaptation depending on the specific application. For example, a higher degree of decimation may be applied when computational efficiency is prioritized, such as in large-scale acoustical simulations or real-time rendering, while a lower degree may be used when higher accuracy is required. This decimation-based simplification offers significant computational advantages. By reducing the number of surface elements, the method decreases the memory and processing requirements needed for simulations involving complex 3D models. In acoustical simulations, where the primary concern is often the interaction of sound waves with the object’s outer surfaces, decimating minor details that do not contribute meaningfully to reflections or absorptions helps optimize performance without sacrificing the accuracy of the simulation.

[0241] The flexibility of the decimation module ensures that the method can be applied to various types of objects within the 3D model of a building complex, from furniture and structural components to architectural elements. By selectively retaining key features while discarding non-critical details, the decimation module provides a balance between computational efficiency and geometric fidelity, supporting its application in diverse computational environments.

[0242] Embodiment List

[0243] Disclosed herein are the following embodiments:

[0244] 1. A computer-implemented method for reconstructing a 3D model of a building complex, wherein the 3D model of a building complex has a 3D model of a building complex volume, and wherein the 3D model of a building complex comprises:P7391 PC00

[0245] 48

[0246] at least one indoor volume and at least one outdoor volume, wherein the at least one outdoor volume partially or entirely surrounds the at least one indoor volume; or

[0247] at least two indoor volumes,

[0248] wherein the at least one indoor volume or at least one of the at least two indoor volumes comprises an inner volume, wherein the method comprises the steps of:

[0249] obtaining a 3D representation of the 3D model of a building complex; filling the inner volume with a plurality of spatial partitioning data structures (SPDSs), wherein the inner volume is delimited by at least one indoor boundary of the at least one indoor volume;

[0250] applying at least one space proxy to the plurality of SPDSs, wherein each of the at least one space proxy represents a computational proxy of the at least one indoor volume and wherein the computational proxy forms a coherent patch of SPDSs, wherein the coherent patch of SPDSs comprises a plurality of SPDSs that are in contact with each other; and reconstructing the at least one indoor volume comprised in the 3D model of a building complex by assembling together the at least one indoor boundary spatially correlated to the at least one space proxy, wherein the spatial correlation is based on proximity or geometric adjacency.

[0251] 2. The computer-implemented method according to item 1 , wherein the proximity is defined as a shortest measurable distance between an SPDS and the at least one indoor boundary, satisfying a predetermined distance threshold.

[0252] 3. The computer-implemented method according to any one of the preceding items, wherein the geometric adjacency is defined as a spatial relationship where an SPDS and the at least one indoor boundary are in direct contact or share at least one geometric feature, such as an edge, vertex, or face, within a predefined tolerance.

[0253] 4. The computer-implemented method according to any one of the preceding items, wherein the method for reconstructing a 3D model of a building complex is suitable for acoustic simulation.P7391 PC00

[0254] 49

[0255] 5. The computer-implemented method according to any one of the preceding items, wherein the 3D representation of the 3D model of a building complex is based on point cloud data, at least one mesh model, at least one polygonal model, at least one wireframe model, at least one voxel model, at least one Non-Uniform Rational B-splines (NURBS) model and / or at least one 3D scene.

[0256] 6. The computer-implemented method according to any one of the preceding items, wherein the 3D representation of the 3D model of a building complex is extracted from a BIM model, such as an Industry Foundation Classes (I FC) file, or is extracted from a Drawing Exchange Format (DXF) file, an Universal Scene Description (USD) file and / or a GL Transmission Format (GLTF) file.

[0257] 7. The computer-implemented method according to any one of the preceding items, wherein the at least one space proxy is a SPDS space representation or a SPDS-based proxy.

[0258] 8. The computer-implemented method according to any one of the preceding items, wherein the spatial correlation between the plurality of SPDSs and the at least one indoor boundary is identified by generating at least one closest point query, wherein the at least one closest point query identifies the at least one indoor boundary from a nearest SPDS comprised in the plurality of SPDSs, thereby identifying the at least one indoor boundary of the at least one indoor volume comprised in the 3D model of a building complex.

[0259] 9. The computer-implemented method according to any one of the preceding items, wherein the at least one indoor volume is at least one room.

[0260] 10. The computer-implemented method according to any one of the preceding items, wherein the at least one indoor volume is at least two indoor volumes, such as at least two rooms.

[0261] 11. The computer-implemented method according to any one of the preceding items, wherein the 3D model of a building complex comprises at least one 3D model of an element such as at least one 3D model of an object, wherein the at least one 3D model of an element is arranged in the at least one indoor area.P7391 PC00

[0262] 50

[0263] 12. The computer-implemented method according to any one of the preceding items, wherein the method further comprises applying to the at least one 3D model of an element a geometry simplification, thereby obtaining a simplified 3D model of a building complex.

[0264] 13. The computer-implemented method according to any one of the preceding items, wherein the geometry simplification extracts geometry metadata from the 3D representation, such as at least one material property and / or at least one spatial annotation, such that the at least one 3D model of an element is simplified based on a category associated with the at least one 3D model of an element, and wherein the category is defined in the geometry metadata.

[0265] 14. The computer-implemented method according to item 13, wherein the category is an element name, preferably an object name, such as chair, table or cable.

[0266] 15. The computer-implemented method according to any one of the preceding items, wherein the IFC file comprises at least one space annotation, and wherein the at least one space annotation is extracted from an IFC class named IFCSpace, wherein the IFCSpace is defined in the IFC file.

[0267] 16. The computer-implemented method according to any one of the preceding items, wherein the geometry simplification comprises a bounding volume replacement module, wherein the bounding volume replacement module replaces the at least one 3D model of an element with a simplified 3D model of an element, wherein the simplified 3D model of an element approximates outer contours of the at least one 3D model of an element.

[0268] 17. The computer-implemented method according to item 16, wherein the simplified 3D model of an element is a bounding box, a convex hull volume, a cylindrical volume, a spherical volume, an ellipsoidal volume, an oriented bounding box, a capsule volume, or any combinations thereof.

[0269] 18. The computer-implemented method according to any one of the preceding items, wherein the geometry simplification comprises a plane projectionP7391 PC00

[0270] 51

[0271] module, wherein the plane projection module projects the at least one 3D model of an element to a 2D projection plane, thereby replacing the at least one 3D model of an element by at least one 2D model of an element.

[0272] 19. The computer-implemented method according to item 18, wherein the at least one 3D model of an element is determined to be nearly two-dimensional when an aspect ratio threshold is satisfied, such that a ratio between the largest and smallest dimensions of the at least one 3D model of an element is greater than or equal to a plane projection threshold.

[0273] 20. The computer-implemented method according to any one of the preceding items, wherein the geometry simplification comprises a decimation module, wherein the decimation module executes a decimation algorithm, wherein the decimation algorithm removes at least one surface element of the at least one 3D model of an element, thereby reducing the number of the at least one surface element of the at least one 3D model of an element.

[0274] 21. The computer-implemented method according to any one of the preceding items, wherein the geometry simplification comprises an outer shell voxelization module, wherein the outer shell voxelization module comprises identifying at least one outer shell of the at least one 3D model of an element, creating a set of SPDSs matching the at least one outer shell, and wherein the set of SPDSs defines at least one 3D model of an element outer boundary, and returning a triangulated representation of the at least one 3D model of an element outer boundary, thereby returning a representation of outer faces of the set of SPDSs.

[0275] 22. The computer-implemented method according to any one of the preceding items, wherein the geometry simplification comprises an exclusion module, wherein the exclusion module removes the at least one 3D model of an element if the at least one 3D model of an element is a 1 D object such as at least one cable, at least one wire, and / or at least one railing.

[0276] 23. The computer-implemented method according to item 22, wherein the 1 D object has a 1 D object length equal to or below a length threshold.P7391 PC00

[0277] 52

[0278] 24. The computer-implemented method according to any one of the preceding items, wherein the geometry simplification comprises a polygon replacement module, wherein the polygon replacement module comprises filling the at least one 3D model of an element with a secondary set of SPDSs, comparing a first and a second subsidiary set of SPDSs of the secondary set of SPDSs, wherein the first and the second subsidiary sets are neighbouring sets of SPDSs, thereby defining a dimensionality of the at least one 3D model of an element, extracting feature edges of the at least one 3D model of an element based on the dimensionality, and replacing each of the feature edges by a polygon representation.

[0279] 25. The computer-implemented method according to item 23, wherein the feature edges are defined by an element boundary between two faces of the at least one 3D model of an element.

[0280] 26. The computer-implemented method according to any one of the preceding items, wherein the method further comprises applying an intersection solver to the simplified 3D model of a building complex, wherein the intersection solver determines or establishes a plurality of coherent shells.

[0281] 27. The computer-implemented method according to item 26, wherein the intersection solver analyzes a plurality of triangles comprised in the 3D model of a building complex, preferably comprised in the at least one indoor boundary and / or in the at least one 3D model of an element, and wherein the intersection solver reconstructs edges and vertices based on an intersection solver predetermined threshold, such that the plurality of triangles forms a plurality of coherent shells.

[0282] 28. The computer-implemented method according to any one of items 26-27, wherein the plurality of coherent shells are inner shells and / or outer shells, wherein the inner shells are at least one primary boundary of the at least one 3D model of an element, and the outer shells are the at least one indoor boundary.P7391 PC00

[0283] 53

[0284] 29. The computer-implemented method according to any one of the preceding items, wherein the plurality of SPDSs is a plurality of volumetric pixels or a plurality of volume unitary elements.

[0285] 30. The computer-implemented method according to item 29, wherein the plurality of SPDSs is a plurality of volumetric elements, such as voxels, wherein the plurality of voxels are cubes, and wherein the cubes have a cube edge of 0.2 m, preferably 0.1 m, more preferably 0.05 m, even more preferably 0.4 m.

[0286] 31. The computer-implemented method according to any one of the preceding items, wherein the space proxy is named according to the at least one space annotation labelled in or derived from the 3D representation of the 3D model of a building complex, or wherein the space proxy is named with a generic indexing.

[0287] 32. The computer-implemented method according to any one of the preceding items, further comprising performing at least one raytracing analysis, wherein at least one ray trace identifies the at least one indoor boundary from the nearest SPDS comprised in the plurality of SPDSs, thereby identifying the at least one indoor boundary of the at least one indoor area comprised in the building complex.

[0288] 33. A computer-implemented method for generating an impulse response for a listening point comprised in at least one indoor volume as reconstructed according to any one of items 1-32, wherein the method comprises receiving a 3D model of the at least one indoor volume, the position of at least one sound source in the at least one indoor volume, and acoustic properties of at least one boundary in the at least one indoor volume;

[0289] determining using a wave-based solver and / or a geometrical acoustics solver, an impulse response of a wave-based and / or a ray-based propagation of an impulse emitted at the at least one sound source in the at least one indoor volume and received at the listening point.P7391 PC00

[0290] 54

[0291] 34. The computer-implemented method according to item 33, wherein the method further comprises rendering a base audio signal by convolving the base audio signal with the impulse response, thereby creating a rendered audio signal.

[0292] 35. The computer-implemented method according to any one of items 33-34, wherein the rendered audio signal provides an audio rendering of the base audio signal in the 3D model of the internal space at the listening point.

[0293] 36. The computer-implemented method according to any one of items 33-35, wherein the computer-implemented method further comprises generating a volumetric mesh of the at least one indoor volume, wherein the volumetric mesh is used for computing the wave-based solver, and wherein each mesh of the volumetric mesh has a predetermined mesh size.

[0294] 37. The computer-implemented method according to any one of items 33-36, wherein the smallest dimension of the at least one 3D model of an element according to any one of items 1-30 is equal or larger than the predetermined mesh size.

[0295] 38. The computer-implemented method according to any one of items 33-37, wherein the predetermined mesh size is proportional to the 3D model of a building complex volume.

[0296] 39. The computer-implemented method according to any one of items 33-38, wherein the acoustic properties of the at least one boundary are extracted or derived based on boundary names or metadata comprised in the 3D representation of the 3D model of a building complex according to any one of items 1-32.

Claims

P7391 PC0055Claims1. A computer-implemented method for generating an impulse response for a listening point comprised in at least one indoor volume as reconstructed according to a secondary computer-implemented method for reconstructing a 3D model of a building complex, wherein the 3D model of a building complex has a 3D model of a building complex volume, and wherein the 3D model of a building complex comprises:- the at least one indoor volume and at least one outdoor volume, wherein the at least one outdoor volume partially or entirely surrounds the at least one indoor volume; orat least two indoor volumes,wherein the at least one indoor volume or at least one of the at least two indoor volumes comprises an inner volume,wherein the secondary computer-implemented method comprises the steps of: obtaining a 3D representation of the 3D model of a building complex; filling the inner volume with a plurality of spatial partitioning data structures (SPDSs), wherein the inner volume is delimited by at least one indoor boundary of the at least one indoor volume;applying at least one space proxy to the plurality of SPDSs, wherein each of the at least one space proxy represents a computational proxy of the at least one indoor volume and wherein the computational proxy forms a coherent patch of SPDSs, wherein the coherent patch of SPDSs comprises a plurality of SPDSs that are in contact with each other; reconstructing the at least one indoor volume comprised in the 3D model of a building complex by assembling together the at least one indoor boundary spatially correlated to the at least one space proxy, wherein the spatial correlation is based on proximity or geometric adjacency;wherein the computer-implemented method comprises:receiving a 3D model of the at least one indoor volume, the position of at least one sound source in the at least one indoor volume, and acoustic properties of at least one boundary in the at least one indoor volume; and determining using a wave-based solver and / or a geometrical acoustics solver, an impulse response of a wave-based and / or a ray-based propagation of an impulse emitted at the at least one sound source in the at least one indoor volume and received at the listening point.P7391 PC00562. The computer-implemented method according to claim 1 , wherein the proximity is defined as a shortest measurable distance between an SPDS and the at least one indoor boundary, satisfying a predetermined distance threshold and / or wherein the geometric adjacency is defined as a spatial relationship wherein an SPDS and the at least one indoor boundary are in direct contact or share at least one geometric feature, such as an edge, vertex, or face, within a predefined tolerance.

3. The computer-implemented method according to any one of the preceding claims, wherein the 3D representation of the 3D model of a building complex is based on point cloud data, at least one mesh model, at least one polygonal model, at least one wireframe model, at least one voxel model, at least one Non-Uniform Rational B-splines (NURBS) model and / or at least one 3D scene.

4. The computer-implemented method according to any one of the preceding claims, wherein the 3D representation of the 3D model of a building complex is extracted from a BIM model, such as an Industry Foundation Classes (I FC) file, or is extracted from a Drawing Exchange Format (DXF) file, an OBJ file, a 3DM file, a SKP file, an Universal Scene Description (USD) file and / or a GL Transmission Format (GLTF) file.

5. The computer-implemented method according to any one of the preceding claims, wherein the at least one space proxy is a SPDS space representation or a SPDS-based proxy.

6. The computer-implemented method according to any one of the preceding claims, wherein the spatial correlation between the plurality of SPDSs and the at least one indoor boundary is identified by generating at least one closest point query, wherein the at least one closest point query identifies the at least one indoor boundary from a nearest SPDS comprised in the plurality of SPDSs, thereby identifying the at least one indoor boundary of the at least one indoor volume comprised in the 3D model of a building complex.P7391 PC00577. The computer-implemented method according to any one of the preceding claims, wherein the at least one indoor volume is at least one room and / or wherein the at least one indoor volume is at least two indoor volumes, such as at least two rooms.

8. The computer-implemented method according to any one of the preceding claims, wherein the 3D model of a building complex comprises at least one 3D model of an element such as at least one 3D model of an object, wherein the at least one 3D model of an element is arranged in the at least one indoor area and wherein the method further comprises applying to the at least one 3D model of an element a geometry simplification, thereby obtaining a simplified 3D model of a building complex.

9. The computer-implemented method according to any one of the preceding claims, wherein the geometry simplification comprises a polygon replacement module, wherein the polygon replacement module comprises filling the at least one 3D model of an element with a secondary set of SPDSs, comparing a first and a second subsidiary set of SPDSs of the secondary set of SPDSs, wherein the first and the second subsidiary sets are neighbouring sets of SPDSs, thereby defining a dimensionality of the at least one 3D model of an element, extracting feature edges of the at least one 3D model of an element based on the dimensionality, and replacing each of the feature edges by a polygon representation.

10. The computer-implemented method according to any one of the preceding claims, wherein the method further comprises applying an intersection solver to the simplified 3D model of a building complex, wherein the intersection solver determines / establishes a plurality of coherent shells, and wherein the intersection solver analyses a plurality of triangles comprised in the 3D model of a building complex, preferably comprised in the at least one indoor boundary and / or in the at least one 3D model of an element, and wherein the intersection solver reconstructs edges and vertices based on an intersection solver predetermined threshold, such that the plurality of triangles forms a plurality of coherent shells.P7391 PC005811. The computer-implemented method according to any one of the preceding claims, wherein the space proxy is named according to the at least one space annotation labelled in or derived from the 3D representation of the 3D model of a building complex, or wherein the space proxy is named with a generic indexing.

12. The computer-implemented method according to any one of the preceding claims, further comprising performing at least one raytracing analysis, wherein at least one ray trace identifies the at least one indoor boundary from the nearest SPDS comprised in the plurality of SPDSs, thereby identifying the at least one indoor boundary of the the at least one indoor volume comprised in the building complex.

13. The computer-implemented method according to any one of the preceding claims, wherein the method further comprises rendering a base audio signal by convolving the base audio signal with the impulse response, thereby creating a rendered audio signal, and wherein the rendered audio signal provides an audio rendering of the base audio signal in the 3D model of the internal space at the listening point.

14. The computer-implemented method according to any one of the preceding claims, wherein the acoustic properties of the at least one boundary are extracted or derived based on boundary names or metadata comprised in the 3D representation of the 3D model of a building complex.

15. The computer-implemented method according to any one of the preceding claims, wherein the 3D representation of the 3D model of a building complex and / or the at least one 3D model of an element is acquired through photogrammetry, laser scanning or LiDAR, structured light scanning, one or more depth-sensing cameras, or any combinations thereof.

Images