Space reconstruction method and space reconstruction system using wi-fi channel status information(CSI) based on environmental adaptation and object form, and a medium having computer readable program for executing the method

Abstract

The present invention relates to a method for spatial reconstruction based on environmental adaptation and object shape utilizing Wi-Fi CSI, a spatial reconstruction system, and a computer-readable program for executing the spatial reconstruction method, wherein Wi-Fi CSI can efficiently perform spatial analysis and recognition, and in particular, apply meta-learning techniques to recognize behavioral patterns to enhance adaptability to the environment, and learn and reconstruct the object shape of the space through CSI data. According to the present invention, a method for spatial reconstruction utilizing Wi-Fi CSI is provided, characterized by comprising: a Wi-Fi CSI data collection processing step for collecting Wi-Fi CSI data; a behavioral pattern recognition step for recognizing behavioral patterns while excluding environmental factors based on the collected Wi-Fi CSI data; and a spatial configuration step for configuring a space including objects based on the collected Wi-Fi CSI data.

Description

Technology Field

[0001] The present invention relates to a spatial reconstruction method and system based on environmental adaptation and object shape utilizing Wi-Fi CSI, and a medium storing a computer-readable program for executing the spatial reconstruction method. More specifically, the invention relates to a spatial reconstruction method and system based on environmental adaptation and object shape utilizing CSI, and a medium storing a computer-readable program for executing the spatial reconstruction method, which can efficiently perform spatial analysis and recognition by utilizing Wi-Fi Channel State Information (CSI), and in particular, can recognize behavioral patterns by applying meta-learning techniques to enhance adaptability to the environment, and learn and reconstruct object shapes in the space through CSI data. Background Technology

[0002] The most common method for predicting human behavior is the use of wearable devices, including sensors. Korean Registered Patent No. 10-2089002 discloses a configuration that provides feedback on monitoring results regarding user behavior based on multiple sensor data measured using various sensors embedded in a wearable device.

[0003] However, conventionally, predicting human behavior in this way required attaching separate devices, such as sensors, to the body. Attaching devices to the body is cumbersome and restricts movement, and there are also problems involving costs associated with using the devices.

[0004] Another method is behavior prediction technology based on optical cameras. However, this method also has inherent problems, such as the necessity of an optical camera of a certain size to provide a light source, and the difficulty of installing the camera in confined spaces.

[0005] Meanwhile, the characteristics of Wi-Fi signals change depending on objects, people, and environmental factors within a space. Therefore, Wi-Fi CSI provides important data for analyzing the environment and human behavior within a space.

[0006] Specifically, behavioral pattern recognition using Wi-Fi CSI is a technology that analyzes the behavior or movements of people or objects by utilizing wireless signal data collected from Wi-Fi networks. Wi-Fi CSI represents the propagation characteristics during the transmission and reception of Wi-Fi signals, and through this data, it is possible to track a person's location, movement, and even posture or actions in an indoor environment.

[0007] However, these environmental factors are a major cause of reduced accuracy in behavioral pattern analysis. Furthermore, directly reconstructing the shape of objects or spaces using Wi-Fi CSI is a very complex problem. Prior art literature

[0009] Republic of Korea Published Patent Application 10-2024-0102615 (Published July 3, 2024) Republic of Korea Published Patent Application 10-2023-0070849 (Published May 23, 2023) Republic of Korea Registered Patent Application 10-2397311 (Announced May 13, 2022) The problem to be solved

[0010] Accordingly, the present invention, which aims to solve the aforementioned conventional problems, is intended to provide a spatial reconstruction method and system based on environmental adaptation and object shapes utilizing CSI, a spatial reconstruction system, and a computer-readable program for executing the spatial reconstruction method, which can efficiently perform spatial analysis and recognition by utilizing Wi-Fi CSI, and in particular recognize behavioral patterns by applying meta-learning techniques to enhance adaptability to the environment, and learn and reconstruct object shapes in the space through CSI data.

[0011] Specifically, the present invention has another objective of providing an environment-adapting and object-shaped-based spatial reconstruction method and spatial reconstruction system utilizing Wi-Fi CSI, which can recognize environment-adapting behavioral patterns in response to environmental variability in various environments by utilizing an environment-adapting artificial intelligence (AI) model based on the meta-learning algorithm of MAML (Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks (2017)) and Wi-Fi CSI data, and can learn the shape of objects, analyze the location and structure of objects, and visualize and reconstruct the entire space in 3D through an object shape learning and spatial reconstruction algorithm, and a medium storing a computer-readable program for executing the spatial reconstruction method.

[0012] The problems solved by the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0014] According to one aspect of the present invention for achieving the objectives and other features of the present invention, a method for spatial reconstruction using Wi-Fi CSI is provided, comprising: a Wi-Fi CSI data collection processing step for collecting Wi-Fi CSI data; a behavior pattern recognition step for recognizing a behavior pattern by excluding environmental factors based on the collected Wi-Fi CSI data; and a space configuration step for configuring a space including objects based on the collected Wi-Fi CSI data; wherein the behavior pattern recognition step comprises a basic learning step for pre-learning to recognize basic behavior patterns by learning Wi-Fi CSI data through a hybrid AI model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM); a meta-learning step for learning the variability of Wi-Fi CSI data and learning initialization parameters for behavior pattern recognition using a Model-Agnostic Meta-Learning (MAML) module after the basic learning step; and a behavior pattern determination step for recognizing a final behavior pattern by excluding environmental factors of the Wi-Fi CSI signal after the meta-learning step.

[0015] In one aspect of the present invention, the spatial reconstruction step comprises: a product feature recognition step that recognizes product features or product shapes including the shape and size of a product by utilizing an autoencoder or GAN (Generative Adversarial Networks) model that receives Wi-Fi CSI data as input and learns the product shape and features; a product position-orientation estimation step that estimates product position data including the position and direction of a product by analyzing the time and angle at which a signal reaches a product using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques; and a spatial reconstruction step that visualizes a 3D space by generating a mesh after reconstructing the space in a point cloud format by combining the product shape recognized in the product feature recognition step and the product position data learned in the product position-orientation estimation step.

[0016] According to another aspect of the present invention, a spatial reconstruction system utilizing Wi-Fi CSI is provided, comprising: a Wi-Fi CSI data collection processing unit configured to collect Wi-Fi CSI data; a behavior pattern recognition unit configured to recognize behavior patterns by excluding environmental factors based on the collected Wi-Fi CSI data; and a spatial configuration unit configured to configure a space including objects based on the collected Wi-Fi CSI data; wherein the behavior pattern recognition unit comprises a basic learning unit that pre-learns to recognize basic behavior patterns by learning Wi-Fi CSI data through a hybrid AI model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM); a meta-learning learning unit that learns the variability of Wi-Fi CSI data and learns initialization parameters for behavior pattern recognition using a MAML meta-learning module; and a behavior pattern determination unit that recognizes a final behavior pattern by excluding environmental factors of the Wi-Fi CSI signal.

[0017] In another aspect of the present invention, the spatial composition unit may include: an object feature recognition unit that recognizes an object shape including the shape and size of an object by utilizing an autoencoder model that receives Wi-Fi CSI data as input and learns the shape and features of an object; an object location data estimation unit that estimates object location data including the position and direction of an object by analyzing the time and angle at which a signal reaches an object using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques; and a spatial reconstruction step that visualizes a 3D space by combining the object shape recognized by the object feature recognition unit and the object location data estimated by the object location data estimation unit to reconstruct the space in a point cloud format, and then generating a mesh.

[0018] In another aspect of the present invention, the object feature recognition unit comprises, wherein the autoencoder learns the shape of the object during the process of compressing and restoring input data into a low-dimensional latent space, and recognizes the shape of the object using the latent space vector of Equation 1 below and the restored Wi-Fi CSI data of Equation 2.

[0019] Equation 1

[0020] Equation 2

[0021] (Here, Z is the latent space vector, X CSI is Wi-Fi CSI data, and f encoder is the encoder function, is restored Wi-Fi CSI data, f decoder is a decoder function)

[0022] The above object position data estimation unit estimates using the distance (d) to the object obtained through Equation 3 below and the angle of incidence (θ) obtained through Equation 4 below, and

[0023] Equation 3

[0024] (Here, c is the speed of light, and Δt is the signal propagation time)

[0025] Equation 4

[0026] (Here, θ is the angle of incidence of the object, λ is the wavelength of the signal, and ΔΦ is the phase change)

[0027] The above space reconstruction unit can be configured to use point cloud coordinates (P) obtained through Equation 5 below.

[0028] Equation 5

[0029] (Here, h is the height of the object) Effects of the invention

[0031] According to the method and system for spatial reconstruction based on environment adaptation and object shape utilizing Wi-Fi CSI according to the present invention, by utilizing Wi-Fi CSI to recognize behavioral patterns and visualizing the location and shape of the space and objects in the space as a 3D map, it is possible to achieve efficient spatial reconstruction.

[0032] The effects of the present invention are not limited to those mentioned above, and other unmentioned problems will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0034] FIG. 1 is a flowchart schematically illustrating the reconstruction process of an environment adaptation and object shape-based spatial reconstruction method using Wi-Fi CSI according to the present invention. FIG. 2 is a flowchart schematically illustrating the behavioral pattern recognition process included in the method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention. FIG. 3 is a flowchart schematically illustrating the spatial configuration process included in the method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention. FIG. 4 is a diagram illustrating the overall configuration and data flow for performing an environment adaptation and object shape-based spatial reconstruction method using Wi-Fi CSI according to the present invention, and includes a process of receiving CSI data as input and recognizing behavioral patterns after environment adaptation through a hybrid AI model and a MAML meta-learning module. FIG. 5 is a diagram illustrating, in the form of a block diagram, each component for object learning and spatial reconstruction performed in the method for environment adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention, and includes the linkage relationship between the object learning module, the location estimation module, and the 3D map generation module. FIG. 6 is a diagram illustrating the flow of object shape learning and 3D map reconstruction performed in the method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention, and includes the process of estimating the location and shape of an object based on Wi-Fi CSI data, generating a point cloud based thereon, and then visualizing it as a 3D map. FIG. 7 is a block diagram schematically illustrating the configuration of an environment adaptation and object shape-based spatial reconstruction system utilizing Wi-Fi CSI according to the present invention. FIG. 8 is a diagram schematically showing the configuration of a behavior pattern recognition unit included in an environment adaptation and object shape-based spatial reconstruction system utilizing Wi-Fi CSI according to the present invention. FIG. 9 is a diagram schematically showing the configuration of a spatial configuration unit included in an environment adaptation and object shape-based spatial reconstruction system utilizing Wi-Fi CSI according to the present invention. Specific details for implementing the invention

[0035] Further objects, features, and advantages of the present invention can be more clearly understood from the following detailed description and the accompanying drawings.

[0036] Before providing a detailed description of the present invention, it should be understood that the present invention is capable of various modifications and may have various embodiments, and that the examples described below and illustrated in the drawings are not intended to limit the present invention to specific embodiments, but rather include all modifications, equivalents, and substitutions that fall within the spirit and scope of the present invention.

[0037] When it is stated that one component is "connected" or "connected" to another component, it should be understood that while it may be directly connected or connected to that other component, there may also be other components in between. On the other hand, when it is stated that one component is "directly connected" or "directly connected" to another component, it should be understood that there are no other components in between.

[0038] The terms used in this specification are used merely to describe specific embodiments and are not intended to limit the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0039] Additionally, terms such as "...part," "...unit," and "...module" described in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware, software, or a combination of hardware and software.

[0040] Furthermore, in the description referring to the attached drawings, identical components are assigned the same reference numeral regardless of drawing symbols, and redundant descriptions thereof are omitted. In describing the present invention, if it is determined that a detailed description of related prior art could unnecessarily obscure the essence of the present invention, such detailed description is omitted.

[0041] Hereinafter, a space reconstruction method and system based on environment adaptation and object shape utilizing Wi-Fi CSI according to a preferred embodiment of the present invention, a space reconstruction system, and a medium storing a computer-readable program for executing the space reconstruction method will be described in detail with reference to the attached drawings.

[0042] First, a method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention will be described in detail with reference to FIGS. 1 to 6.

[0043] FIG. 1 is a flowchart schematically illustrating the reconstruction process of the method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention, FIG. 2 is a flowchart schematically illustrating the behavioral pattern recognition process included in the method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention, and FIG. 3 is a flowchart schematically illustrating the spatial configuration process included in the method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention. FIG. 4 is a diagram illustrating the overall configuration and data flow for performing an environment adaptation and object shape-based spatial reconstruction method using Wi-Fi CSI according to the present invention, FIG. 5 is a diagram illustrating each component for object learning and spatial reconstruction performed in the environment adaptation and object shape-based spatial reconstruction method using Wi-Fi CSI according to the present invention in the form of a block diagram, and FIG. 6 is a diagram illustrating the flow of object shape learning and 3D map reconstruction performed in the environment adaptation and object shape-based spatial reconstruction method using Wi-Fi CSI according to the present invention.

[0044] Here, FIG. 4 is a diagram including a process of receiving CSI data as input and recognizing behavioral patterns after adapting to the environment through a hybrid AI model and a MAML meta-learning module, FIG. 5 is a diagram including the linkage relationship between an object learning module, a location estimation module, and a 3D map generation module, and FIG. 6 is a diagram including a process of estimating the location and shape of an object based on Wi-Fi CSI data, generating a point cloud based thereon, and then visualizing it as a 3D map.

[0045] The method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention, as shown in FIGS. 1 to 6, largely comprises a Wi-Fi CSI data collection and processing step (S100), a behavior pattern recognition step (S200), and a spatial reconstruction step (S300).

[0046] Specifically, the method for environmental adaptation and object shape-based spatial reconstruction using Wi-Fi CSI according to the present invention comprises, as shown in FIGS. 1 to 6, a Wi-Fi CSI data collection and processing step (S100) for collecting and processing Wi-Fi CSI data; a behavior pattern recognition step (S200) for excluding environmental factors (or environmental elements) and recognizing behavior patterns based on the Wi-Fi CSI data collected and processed in the Wi-Fi CSI data collection and processing step (S100); and a spatial configuration step (S300) for configuring a space including objects based on the Wi-Fi CSI data collected and processed in the Wi-Fi CSI data collection and processing step (S100).

[0047] The above Wi-Fi CSI data collection and processing step (S100) is a process for collecting and processing Wi-Fi CSI data, which is Wi-Fi CSI data.

[0048] Specifically, the Wi-Fi CSI data collection processing step (S100) collects Wi-Fi CSI through a receiving device at each path in the environment from a signal transmitted from a Wi-Fi router or an access point. Wi-Fi CSI includes propagation information (strength, phase, frequency, etc.) for multiple channels of the transmitted and received signal.

[0049] And the Wi-Fi CSI collection processing step (S100) removes noise, distortion, interference, etc. from the collected CSI data based on preset criteria.

[0050] Furthermore, the behavior pattern recognition step (S200) is a process that excludes environmental factors (or environmental elements) and recognizes behavior patterns based on the Wi-Fi CSI data collected and processed in the Wi-Fi CSI data collection and processing step (S100).

[0051] The above behavior pattern recognition step (S200) extracts key features to track human movement or behavior based on CSI data. For example, it recognizes movement patterns by analyzing changes in CSI over time, phase changes, path movement, etc., and classifies specific behaviors based on the extracted features. For example, since the CSI patterns differ when a person walks, runs, or sits, human activity is classified through an algorithm that distinguishes these patterns; this behavior classification can utilize a machine learning model.

[0052] In the present invention, the behavior pattern recognition step (S200) utilizes an AI model (i.e., a MAML meta-learning model) that applies Model-Agnostic Meta-Learning (MAML) for rapid adaptation to environmental changes and learns parameters for adapting to various environments for behavior pattern recognition based on Wi-Fi CSI data, and may utilize a hybrid AI model composed of trained CNN and LSTM (pre-trained CNN and LSTM) to process Wi-Fi CSI data and recognize human behavior patterns. MAML is a meta-learning technique capable of rapid adaptation to various environments, and learns model parameters capable of adapting by reflecting signal changes in new environments. Here, the use of the MAML meta-learning module may be applied optionally.

[0053] Specifically, the behavior pattern recognition step (S200) comprises a basic learning step (S210) for pre-learning to recognize basic behavior patterns by learning Wi-Fi CSI data in various environments using a hybrid AI model, a meta-learning step (S220) for learning the variability of Wi-Fi CSI data and learning optimal initialization parameters for behavior pattern recognition using a MAML meta-learning module to quickly adapt to a new environment after the basic learning step (S210), and a behavior pattern determination step (S230) for determining a final behavior pattern by excluding environmental factors of the Wi-Fi CSI signal after quickly adapting to environmental changes following the meta-learning step (S220).

[0054] The above basic learning step (S210) uses a hybrid AI model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) to process Wi-Fi CSI data. The CNN extracts spatial features of the data, and the LSTM learns temporal patterns.

[0055] CNN's output Z CNN It is expressed as follows.

[0056]

[0057] Here, X CSI represents Wi-Fi CSI input data, and f CNN represents a CNN network.

[0058] And sequence learning through LSTM is expressed as follows.

[0059]

[0060] Here, f LSTM represents the LSTM network, and Z LSTM represents the output after learning the temporal pattern of the sequence.

[0061] Ultimately, the model predicts class probabilities for behavioral patterns. Prints.

[0062]

[0063] Furthermore, in the meta-learning step (S220) described above, the meta-learning algorithm (MAML) is an algorithm designed to enable rapid adaptation to various environments. This MAML algorithm enables rapid adaptation to new environments by learning the initial parameters of the model through meta-learning.

[0064] Meta-learning algorithms (MAML) include an internal learning phase and a meta-update phase.

[0065] The internal learning stage is a given environment E from model f θ is data D E loss function using The parameters are updated according to [this]. In this case, the internal learning is expressed as follows.

[0066]

[0067] Here, α is the learning rate, and θ represents the model parameter.

[0068] And the meta update steps are multiple environments E1, E2, ..., E n Based on the results learned from, the model's initial parameter θ is updated through meta-learning. The meta-update is expressed as follows.

[0069]

[0070] Here, β represents the meta-learning rate.

[0071] Furthermore, the above behavior pattern determination step (S230) is a process of recognizing a behavior pattern by excluding environmental factors (environmental factors) of the Wi-Fi CSI signal after rapidly adapting to environmental changes, and the environmental factors in the behavior pattern determination step (S230) are as follows.

[0072] - Walls, furniture, and other obstacles: Wi-Fi signals are reflected or refracted by obstacles, which can alter the characteristics of the received CSI signal.

[0073] - Human movement: Signal strength and phase may change when a person enters or passes through the Wi-Fi signal path.

[0074] - Multipath Propagation: Signals are received through multiple paths, and the arrival times vary by path, which can cause distortion in CSI.

[0075] - Other radio interference: Other nearby electronic devices (e.g., microwave ovens, Bluetooth devices, etc.) may affect the Wi-Fi signal.

[0076] In the behavior pattern determination step (S230) above, the exclusion of the aforementioned environmental factors may involve applying various filtering techniques to remove environmental noise. For example, methods such as Kalman filters and Wiener filters may be used to remove unnecessary fluctuations from the signal.

[0077] In addition, in the above behavior pattern determination step (S230), the exclusion of the aforementioned environmental factors can be stabilized by, for example, adjusting the signal strength or phase value to a certain range.

[0078] Next, the space configuration step (S300) is a process for configuring a space including objects based on Wi-Fi CSI data collected and processed in the Wi-Fi CSI data collection and processing step (S100), wherein the shape of objects within the environment is learned using Wi-Fi CSI data, and a 3D map of the space is reconstructed based thereon, wherein the location, size, and shape of the objects are recognized and reconstructed in the form of a point cloud.

[0079] Specifically, the spatial construction step (S300) includes an object feature recognition step (S310) that recognizes object features (or object shape) including the shape and size of an object by utilizing an autoencoder or GAN (Generative Adversarial Networks) model that receives Wi-Fi CSI data as input and learns the shape and features of an object; an object position-orientation estimation step (S320) that estimates the position and direction of an object (object position data) by analyzing the time and angle at which a signal reaches an object using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques; and a spatial reconstruction step (S330) that visualizes the 3D space by generating a mesh after reconstructing the space in the form of a point cloud by combining the object shape and position data obtained (i.e. learned) through the object feature recognition step (S310) and the object position-orientation estimation step (S320).

[0080] In the above object feature recognition step (S310), the object shape learning based on an autoencoder is performed as follows.

[0081] Autoencoders learn the shapes of objects during the process of compressing input data into a low-dimensional latent space and restoring it.

[0082] An encoder that compresses Wi-Fi CSI data into a latent space vector can be expressed as follows.

[0083]

[0084] Here, Z is a latent space vector, and X CSI is Wi-Fi CSI data, and f encoder is an encoder function.

[0085] The shape of the object is reconstructed by restoring the above latent space vector back to the original space, which can be expressed as follows.

[0086]

[0087] Here, is the restored Wi-Fi CSI data, and f decoder is a decoder function.

[0088] And to minimize the difference between the reconstructed data and the original data, the loss function is defined as follows.

[0089]

[0090] Next, the object position-orientation estimation step (S320) is configured to estimate the position and orientation of an object by analyzing the time and angle at which a signal reaches the object using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques.

[0091] In the above object position-orientation estimation step (S320), position estimation is performed through the time-of-flight (ToF) and angle-of-arrival (AoA).

[0092] Regarding the arrival time, the distance to the object can be estimated based on the arrival time of the signal and is expressed as follows.

[0093]

[0094] Here, d is the distance to the object, c is the speed of light, and Δt is the propagation time of the signal.

[0095] In addition, the location of an object can be estimated based on the angle of the Wi-Fi signal, and is expressed as follows.

[0096]

[0097] Here, θ is the angle of incidence of the object, λ is the wavelength of the signal, and ΔΦ is the phase change.

[0098] Next, the space reconstruction step (S330) is configured to reconstruct the space in the form of a point cloud by combining the shape and location data of the object obtained (i.e., learned) through the object feature recognition step (S310) and the object location-orientation estimation step (S320), and then generate a mesh to visualize the 3D space.

[0099] The above spatial reconstruction step (S330) consists of generating a 3D point cloud by combining the shape and location data of the object (location and orientation of the object).

[0100] Specifically, the space reconstruction step (S330) generates point cloud data by representing each object in 3D coordinates, and can be expressed as follows.

[0101]

[0102] Here, P is the point cloud coordinate, and h is the height of the object.

[0103] Then, the above point cloud data is converted into a 3D mesh to reconstruct the 3D space.

[0104] The above-described spatial configuration step (S300) can be performed based on Wi-Fi CSI data excluding the Wi-Fi CSI data utilized in the above-described behavioral pattern recognition step (S200).

[0105] Meanwhile, an environment adaptation and object shape-based spatial reconstruction system utilizing Wi-Fi CSI according to the present invention will be described in detail with reference to the attached drawings.

[0106] FIG. 7 is a block diagram schematically showing the configuration of an environment adaptation and object shape-based spatial reconstruction system utilizing Wi-Fi CSI according to the present invention, FIG. 8 is a diagram schematically showing the configuration of a behavior pattern recognition unit included in an environment adaptation and object shape-based spatial reconstruction system utilizing Wi-Fi CSI according to the present invention, and FIG. 9 is a diagram schematically showing the configuration of a spatial configuration unit included in an environment adaptation and object shape-based spatial reconstruction system utilizing Wi-Fi CSI according to the present invention.

[0107] Specifically, the space reconstruction system based on environmental adaptation and object shape utilizing Wi-Fi CSI according to the present invention comprises, as shown in FIGS. 7 to 9, a Wi-Fi CSI data collection and processing unit (100) configured to collect and process Wi-Fi CSI data; a behavior pattern recognition unit (200) configured to exclude environmental factors (or environmental factors) and recognize behavior patterns based on the Wi-Fi CSI data collected and processed by the Wi-Fi CSI data collection and processing unit (100); and a space configuration unit (300) configured to configure a space including objects based on the Wi-Fi CSI data collected and processed by the Wi-Fi CSI data collection and processing unit (100).

[0108] The above Wi-Fi CSI data collection and processing unit (100) is a component configured to collect and process Wi-Fi CSI data.

[0109] Specifically, the Wi-Fi CSI data collection processing unit (100) may be composed of a receiving device or receiving unit that collects Wi-Fi CSI from each path of the environment, which is a signal transmitted from a Wi-Fi router or an access point. Here, the Wi-Fi CSI includes propagation information (strength, phase, frequency, etc.) for multiple channels of the transmitted and received signal.

[0110] And the Wi-Fi CSI collection processing unit (100) is configured to remove noise, distortion, interference, etc. from the collected CSI data based on preset criteria.

[0111] Furthermore, the behavior pattern recognition unit (200) is configured to recognize a behavior pattern by excluding environmental factors (or environmental factors) based on Wi-Fi CSI data collected and processed by the Wi-Fi CSI data collection and processing unit (100).

[0112] The above behavior pattern recognition unit (200) is configured to extract key features to track a person's movement or behavior based on CSI data. For example, it recognizes movement patterns by analyzing changes in CSI over time, phase changes, path movements, etc., and classifies specific behaviors based on the extracted features. For example, since the CSI patterns when a person walks, runs, or sits are different, the person's activities are classified through an algorithm that distinguishes them. This behavior classification can be configured to utilize a machine learning model.

[0113] In the present invention, the behavior pattern recognition unit (200) utilizes an AI model (i.e., a MAML meta-learning module) that applies Model-Agnostic Meta-Learning (MAML) for rapid adaptation to environmental changes and learns parameters for adapting to various environments for behavior pattern recognition based on Wi-Fi CSI data, and can be configured to utilize a hybrid AI model that recognizes human behavior patterns by processing Wi-Fi CSI data, composed of learned CNN and LSTM (pre-trained CNN and LSTM). MAML is a meta-learning technique that can adapt quickly to various environments and learns model parameters that can adapt by reflecting signal changes in new environments. Here, the use of the MAML meta-learning module may be applied optionally.

[0114] Specifically, the behavior pattern recognition unit (200) includes a basic learning unit (210) that pre-learns to recognize basic behavior patterns by learning Wi-Fi CSI data in various environments using a hybrid AI model, a meta-learning learning unit (220) that learns the variability of Wi-Fi CSI data and learns optimal initialization parameters for behavior pattern recognition using a MAML meta-learning module to quickly adapt to a new environment, and a behavior pattern determination unit (230) that recognizes a final behavior pattern by excluding environmental factors (environmental factors) of the Wi-Fi CSI signal from the learning result of the meta-learning learning unit (220).

[0115] The above basic learning unit (210) may use a hybrid AI model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM) to process Wi-Fi CSI data. The CNN extracts spatial features of the data, and the LSTM learns temporal patterns.

[0116] CNN's output ZCNN It is expressed as follows.

[0117]

[0118] Here, X CSI represents Wi-Fi CSI input data, and f CNN represents a CNN network. Wi-Fi CSI data X CSI CNN model f CNN When input into, the CNN processes the data and important features Z CNN It extracts [the data]. This is the process of transforming input data into a high-dimensional feature space.

[0119] And sequence learning through LSTM is expressed as follows.

[0120]

[0121] Here, f LSTM represents the LSTM network, and Z LSTM represents the output after learning the temporal pattern of the sequence.

[0122] Ultimately, the model predicts class probabilities for behavioral patterns. Prints.

[0123]

[0124] Furthermore, the meta-learning algorithm (MAML) in the above-mentioned meta-learning learning unit (220) is an algorithm designed to enable rapid adaptation to various environments. This MAML algorithm enables rapid adaptation to new environments by learning the initial parameters of the model through meta-learning.

[0125] Meta-learning algorithms (MAML) are achieved through internal learning and meta-update processes.

[0126] The internal learning process is a given environment E from model f θ is data D E loss function using The parameters are updated according to [this]. In this case, the internal learning is expressed as follows.

[0127]

[0128] Here, α is the learning rate, and θ represents the model parameters. θ L is the slope of the parameter θ for the loss function. L is the loss function, the model's prediction f θ (D E ) and actual value y E Measures the difference between livers.

[0129] And the meta update process involves various environments E1, E2, ..., E n Based on the results learned from, the model's initial parameter θ is updated through meta-learning. The meta-update is expressed as follows.

[0130]

[0131] Here, β represents the meta-learning rate.

[0132] Furthermore, the behavior pattern determining unit (230) is a process of excluding environmental factors (environmental factors) of the Wi-Fi CSI signal and recognizing a behavior pattern, and the environmental factors excluded by the behavior pattern determining unit (230) are as follows.

[0133] - Walls, furniture, and other obstacles: Wi-Fi signals are reflected or refracted by obstacles, which can alter the characteristics of the received CSI signal.

[0134] - Human movement: Signal strength and phase may change when a person enters or passes through the Wi-Fi signal path.

[0135] - Multipath Propagation: Signals are received through multiple paths, and the arrival times vary by path, which can cause distortion in CSI.

[0136] - Other radio interference: Other nearby electronic devices (e.g., microwave ovens, Bluetooth devices, etc.) may affect the Wi-Fi signal.

[0137] In the behavior pattern determining unit (230) above, the exclusion of the aforementioned environmental factors may be achieved by applying various filtering techniques to remove environmental noise. For example, methods such as Kalman filters and Wiener filters may be used to remove unnecessary fluctuations from the signal.

[0138] In addition, the exclusion of the aforementioned environmental factors in the behavior pattern determining unit (230) can stabilize the change by, for example, adjusting the signal strength or phase value to a certain range.

[0139] Next, the space configuration unit (300) is configured to configure a space including objects based on Wi-Fi CSI data collected and processed by the Wi-Fi CSI data collection and processing unit (100), and is configured to learn the shape of objects in the environment using Wi-Fi CSI data and reconstruct a 3D map of the space based on this, and is configured to recognize the location, size, and shape of the objects and reconstruct them in the form of a point cloud.

[0140] Specifically, the spatial configuration unit (300) includes an object shape recognition unit (310) that recognizes object features (or object shapes) including the shape and size of an object by utilizing an autoencoder or GAN (Generative Adversarial Networks) model that receives Wi-Fi CSI data as input and learns the shape and features of an object, an object location data estimation unit (320) that estimates the position and direction (object location data) of an object by analyzing the time and angle at which a signal reaches an object using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques, and a spatial reconstruction unit (330) that visualizes a 3D space by generating a mesh after reconstructing the space in a point cloud format by combining the object shape and location data obtained (i.e. learned) through the object shape recognition unit (310) and the object location data estimation unit (320).

[0141] In the above object shape recognition unit (310), object shape learning based on an autoencoder is performed as follows.

[0142] Autoencoders learn the shapes of objects during the process of compressing input data into a low-dimensional latent space and restoring it.

[0143] An encoder that compresses Wi-Fi CSI data into a latent space vector can be expressed as follows.

[0144]

[0145] Here, Z is a latent space vector, and X CSI is Wi-Fi CSI data, and f encoder is an encoder function.

[0146] The shape of the object is reconstructed by restoring the above latent space vector back to the original space, which can be expressed as follows.

[0147]

[0148] Here, is the restored Wi-Fi CSI data, and f decoder is a decoder function.

[0149] And to minimize the difference between the reconstructed data and the original data, the loss function is defined as follows.

[0150]

[0151] This measures how well the model reconstructs the input data, and a smaller value indicates higher reconstruction performance.

[0152] Next, the object position-orientation estimation step (S320) is configured to estimate the position and orientation of an object by analyzing the time and angle at which a signal reaches the object using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques.

[0153] In the above object position data estimation unit (320), position estimation is performed through the time-of-flight (ToF) and angle-of-arrival (AoA).

[0154] Regarding the arrival time, the distance to the object can be estimated based on the arrival time of the signal and is expressed as follows.

[0155]

[0156] Here, d is the distance to the object, c is the speed of light, and Δt is the propagation time of the signal.

[0157] In addition, the location of an object can be estimated based on the angle of the Wi-Fi signal, and is expressed as follows.

[0158]

[0159] Here, θ is the angle of incidence of the object, λ is the wavelength of the signal, and ΔΦ is the phase change.

[0160] It is used to estimate the angle based on the phase difference of the signal.

[0161] Next, the space reconstruction unit (330) is configured to reconstruct the space in the form of a point cloud by combining the shape and location data of the object obtained (i.e., learned) through the object shape recognition unit (310) and the object location data estimation unit (320), and then generate a mesh to visualize the 3D space.

[0162] The above-mentioned space reconstruction unit (330) is composed of generating a 3D point cloud by combining the shape and location data (location and direction of the object).

[0163] Specifically, the space reconstruction unit (330) generates point cloud data by representing each object in 3D coordinates, and can be expressed as follows.

[0164]

[0165] Here, P is the point cloud coordinate and h is the height of the object, and the 3D position is defined by converting from polar coordinates to Cartesian coordinates.

[0166] Then, the above point cloud data is converted into a 3D mesh to reconstruct the 3D space.

[0167] The above-described space configuration unit (300) can use Wi-Fi CSI data excluding Wi-Fi CSI data utilized in the above-described behavior pattern recognition unit (200).

[0168] Meanwhile, although the above description explains a method for environmental adaptation and object-shaped spatial reconstruction using Wi-Fi CSI, it goes without saying that a computer-readable recording medium storing a program for implementing the method for environmental adaptation and object-shaped spatial reconstruction using Wi-Fi CSI, and a program stored on the computer-readable recording medium for implementing the method for environmental adaptation and object-shaped spatial reconstruction using Wi-Fi CSI, can also be implemented.

[0169] In other words, those skilled in the art will readily understand that the above-described method for environmental adaptation and object-shaped spatial reconstruction utilizing Wi-Fi CSI may be provided by being tangibly implemented as a program of instructions for implementing it, and thus included in a computer-readable recording medium. That is to say, it may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination. The program instructions recorded on the computer-readable recording medium may be those specifically designed and configured for the present invention, or they may be those known and available to those skilled in computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical media such as CD-ROMs and DVDs; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, and USB memory. The above computer-readable recording medium may be a transmission medium, such as an optical or metallic wire or a waveguide, that includes a carrier wave transmitting a signal specifying program instructions, data structures, etc. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The above hardware device may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

[0170] According to the method and system for spatial reconstruction based on environment adaptation and object shape utilizing Wi-Fi CSI and the spatial reconstruction method according to the present invention as described above, and a medium storing a computer-readable program for executing the spatial reconstruction method, there is an advantage in that efficient spatial reconstruction can be achieved by utilizing Wi-Fi CSI to recognize behavioral patterns and visualizing the location and shape of the space and spatial objects as a 3D map.

[0171] The embodiments described in this specification and the accompanying drawings are merely illustrative of a part of the technical concept included in the present invention. Accordingly, since the embodiments disclosed in this specification are intended to explain, not limit, the technical concept of the present invention, it is obvious that the scope of the technical concept of the present invention is not limited by these embodiments. All variations and specific embodiments that can be easily deduced by a person skilled in the art within the scope of the technical concept included in the specification and drawings of the present invention should be interpreted as being included within the scope of the rights of the present invention. Explanation of the symbols

[0173] S100: Wi-Fi CSI data collection and processing step S200: Behavior pattern recognition stage S210: Basic Learning Stage S220: Meta-learning stage S230: Behavioral pattern determination step S300: Space configuration stage S310: Object feature recognition stage S320: Object position-orientation estimation step S330: Space Reorganization Phase 100: Wi-Fi CSI Data Collection and Processing Unit 200: Behavior Pattern Recognition Unit 210: Basic Learning Department 220: Meta-learning learning department 230: Behavior Pattern Determination Unit 300: Spatial composition 310: Object shape recognition unit 320: Object location data estimation section 330: Spatial Reconstruction Section

Claims

Claim 1 A method for spatial reconstruction using Wi-Fi CSI, comprising: a Wi-Fi CSI data collection processing step for collecting Wi-Fi CSI data; a behavior pattern recognition step for recognizing behavior patterns by excluding environmental factors based on the collected Wi-Fi CSI data; and a space configuration step for configuring a space including objects based on the collected Wi-Fi CSI data; wherein the behavior pattern recognition step comprises a basic learning step for pre-training to recognize basic behavior patterns by learning Wi-Fi CSI data through a hybrid AI model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM); a meta-learning step for learning the variability of Wi-Fi CSI data and learning initialization parameters for behavior pattern recognition using a Model-Agnostic Meta-Learning (MAML) module after the basic learning step; and a behavior pattern determination step for recognizing a final behavior pattern by excluding environmental factors of the Wi-Fi CSI signal after the meta-learning step. Claim 2 In claim 1, the spatial reconstruction step comprises: a product feature recognition step that recognizes product features or product shapes including the shape and size of a product by utilizing an autoencoder or GAN (Generative Adversarial Networks) model that receives Wi-Fi CSI data as input and learns the product shapes and features of the product; a product position-orientation estimation step that estimates product position data including the position and direction of a product by analyzing the time and angle at which a signal reaches the product using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques; and a spatial reconstruction step that visualizes a 3D space by generating a mesh after reconstructing the space in a point cloud format by combining the product shape recognized in the product feature recognition step and the product position data learned in the product position-orientation estimation step; wherein the spatial reconstruction step is characterized by generating point cloud data by expressing each product in 3D coordinates and reconstructing it. Claim 3 A spatial reconstruction system utilizing Wi-Fi CSI, comprising: a Wi-Fi CSI data collection processing unit configured to collect Wi-Fi CSI data; a behavior pattern recognition unit configured to recognize behavior patterns by excluding environmental factors based on the collected Wi-Fi CSI data; and a space configuration unit configured to construct a space including objects based on the collected Wi-Fi CSI data; wherein the behavior pattern recognition unit comprises a basic learning unit that pre-learns to recognize basic behavior patterns by learning Wi-Fi CSI data through a hybrid AI model combining a Convolutional Neural Network (CNN) and a Long Short-Term Memory (LSTM); a meta-learning learning unit that learns the variability of Wi-Fi CSI data and learns initialization parameters for behavior pattern recognition by utilizing a Model-Agnostic Meta-Learning (MAML) module; and a behavior pattern determination unit that recognizes a final behavior pattern by excluding environmental factors of the Wi-Fi CSI signal. Claim 4 In claim 3, the spatial reconstruction system utilizing Wi-Fi CSI is characterized by comprising: a spatial reconstruction unit that recognizes the shape of an object including its shape and size by utilizing an autoencoder model that receives Wi-Fi CSI data as input and learns the shape and features of the object; an object location data estimation unit that estimates object location data including the position and direction of the object by analyzing the time and angle at which a signal reaches the object using Time-of-Flight (ToF) and Angle-of-Arrival (AoA) techniques; and a spatial reconstruction step that visualizes the 3D space by combining the object shape recognized by the object feature recognition unit and the object location data estimated by the object location data estimation unit to reconstruct the space in a point cloud format, and then generating a mesh. Claim 5 In paragraph 4, the object feature recognition unit comprises, wherein the autoencoder learns the shape of the object during the process of compressing and restoring input data into a low-dimensional latent space, and recognizes the shape of the object using the latent space vector of Equation 1 below and the restored Wi-Fi CSI data of Equation 2. Equation 1 Equation 2 (where Z is the latent space vector, X CSI is Wi-Fi CSI data, and f encoder is the encoder function, is restored Wi-Fi CSI data, f decoder (is a decoder function) The above object position data estimation unit estimates using the distance to the object (d) obtained through Equation 3 below and the angle of incidence (θ) obtained through Equation 4 below, and Equation 3 (where c is the speed of light and Δt is the signal propagation time) Equation 4 (where θ is the angle of incidence of the object, λ is the wavelength of the signal, and ΔΦ is the phase change) The space reconstruction unit is characterized by being configured to use the point cloud coordinates (P) obtained through Equation 5 below. Equation 5 (where h is the height of the object) Space reconstruction system using Wi-Fi CSI.

Images