Method and device for analyzing users of specific space using wi-fi channel status information

Abstract

A method and device for analyzing users of a specific space using Wi-Fi channel status information are disclosed. The disclosed method for analyzing users of a specific space using Wi-Fi channel status information comprises the steps of: collecting a multi-modal-based dataset having CSI as a first modality and an image as a second modality; training a CSI-utilizing gender / age analysis AI model and a CSI-utilizing location prediction AI model using the dataset; and predicting gender / age / location information related to users located in a specific space using the trained models. The present technology may provide a method for analyzing users of a specific space using Wi-Fi channel status information that analyzes the gender, age, location, and the like of users located in the specific space.

Description

Method and device for analyzing users of a specific space using Wi-Fi channel status information

[0001] The present invention relates to a method and device for analyzing users of a specific space using Wi-Fi channel status information, and more specifically, to a method and device for analyzing users of a specific space using Wi-Fi channel status information for analyzing the gender, age, location, etc. of users located in a specific space.

[0002] Conventional techniques for determining a person's location have primarily relied on camera-based technology. Korean Patent Publication No. 10-2021-0064536 (Title: User Location Calculation Device) describes a technique for deriving a depth-of-field (DoF) area from images acquired from at least two cameras and then determining the user's location based on the relationship between the user and objects within the derived DoF area.

[0003] With the advancement of image processing technology, determining a person's location using images, as described in the aforementioned literature, offers significant advantages in terms of accuracy and computational speed. However, it also faces the issue of personal information, which has recently become a major issue. In particular, image-collecting devices, such as multiple cameras placed in confined spaces, often cause users to feel uneasy and uncomfortable rather than comfortable.

[0004] Recently, various attempts have been made to analyze objects using Wi-Fi's Channel State Information (CSI). Wi-Fi is evolving beyond a communications technology that simply provides connectivity and network services, into a sensing technology that enables security, safety, and family care services in smart homes and Internet of Things (IoT) applications. Wi-Fi sensing technology supports a variety of features and applications, such as motion detection, activity detection and recognition, and vital sign detection.

[0005] CSI can provide detailed information about how Wi-Fi signals are affected by physical features of the environment (such as walls, furniture, and human bodies), offering the potential for analyzing this information for object analysis. While still a nascent field of research, it holds limitless potential for advancement alongside the field of artificial intelligence (AI).

[0006] Meanwhile, multimodality refers to an environment in which computers interact in various forms and meanings. Modality refers to the communication channels used during the interaction process, such as a keyboard for entering text into a computer. Furthermore, multimodality refers to technology that enables user-friendly technology that enables natural communication between humans and computers by integrating various input methods, such as voice, gestures, gaze, facial expressions, and biometric signals (such as pulse, heart rate, iris scans, and fingerprints), in addition to traditional text.

[0007] In particular, recent advances in deep learning and neural network-based technologies have led to significant advancements in technologies capable of integrating diverse data types, such as object recognition, speech recognition, and text processing. Consequently, multimodal research is also integrating with AI technology, resulting in powerful synergies through enhanced performance in data processing and analysis. In AI, multimodality is expected to yield outstanding results in analysis and learning by simultaneously processing diverse data, including text, images, audio, and video.

[0008] An embodiment of the present invention provides a method for analyzing users of a specific space using Wi-Fi channel status information that analyzes the gender, age, location, etc. of users located in a specific space.

[0009] Meanwhile, other unspecified purposes of the present invention will be additionally considered within the scope that can be easily inferred from the detailed description and effects thereof below.

[0010] A method for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment may include the steps of collecting a multi-modal based dataset having CSI as a first modality and video as a second modality; learning a CSI-utilizing gender / age analysis AI model and a CSI-utilizing location prediction AI model using the dataset; and predicting gender / age / location information related to users located in a specific space using the learned models.

[0011] The collecting step may include a step of designing a dataset that inputs CSI information for a predetermined space and outputs a 3D pose prediction result, a gender result, and an age result based on image data for the predetermined space; and a step of optimizing the placement of shared Wi-Fi devices under a number of space-environment conditions and constructing a dataset based on the designed dataset.

[0012] The above multiple space-environment conditions may include a space-environment set by a combination of one or more of a closed space, an open space, less than 100 pyeong, and more than 100 pyeong.

[0013] The learning step may include: a step of converting the CSI information into a dimension compatible with the image data so that the CSI information can be applied to a 3D pose prediction neural network model; a step of extracting keypoint information by applying the converted CSI information to a 3D pose prediction neural network model; and a step of designing and learning an optimal layer model using the keypoint information as input.

[0014] The above-mentioned converting step may include a process of encoding amplitude and phase data of the CSI information and aligning them to the same dimension as the image data through a dimension reduction and merging process.

[0015] The above 3D pose prediction neural network model may be a dense-pose RCNN model.

[0016] The collecting step may include: a step of designing a dataset that inputs CSI information for a predetermined space and outputs bounding box coordinate results of object detection based on image data for the predetermined space; and a step of optimizing the placement of Wi-Fi device routers under a number of space-environment conditions and constructing a dataset based on the designed dataset.

[0017] The above learning step may include a step of encoding amplitude and phase data of the CSI information, performing a dimension reduction and merging process, and performing a final bounding box and loss optimization.

[0018] The method may further include a step of confirming the coverage area per router for each of the plurality of spatial-environmental conditions, merging CSI data when installing a plurality of routers, and enhancing the CSI utilization location prediction AI model by utilizing the merged dataset.

[0019] Additionally, a computer-readable recording medium storing one or more computer programs according to one embodiment may include instructions for performing any one of the above methods.

[0020] In addition, a device for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment may include: a data acquisition unit for acquiring CSI data and image data for designing a multi-modal based dataset using CSI as a first modality and an image as a second modality; a processor for collecting the dataset, learning a CSI-utilizing gender / age analysis AI model and a CSI-utilizing location prediction AI model using the dataset, and predicting gender / age / location information related to users located in a specific space using the learned models; and a display for displaying the predicted results.

[0021] This technology can provide a method for analyzing users of a specific space using Wi-Fi channel status information that analyzes the gender, age, location, etc. of users located in a specific space.

[0022] Figure 1 illustrates the overall flow of a method for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment.

[0023] FIG. 2 is a diagram illustrating a detailed flow of a step of collecting user analysis information of a specific space using Wi-Fi channel status information according to one embodiment.

[0024] FIG. 3 is a diagram illustrating a detailed flow of a learning step of a method for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment.

[0025] FIG. 4 is a diagram illustrating a detailed flow of a step of collecting user analysis information of a specific space using Wi-Fi channel status information according to one embodiment.

[0026] FIG. 5 is a diagram illustrating a detailed flow of a learning step of a method for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment.

[0027] FIG. 6 is a diagram illustrating an example of a hardware implementation of a user analysis device according to one embodiment.

[0028] It is to be understood that the attached drawings are provided for reference only to help understand the technical concept of the present invention, and the scope of the present invention is not limited thereby.

[0029] Below, the most preferred embodiments of the present invention are described. In the drawings, thicknesses and spacing are expressed for convenience of explanation and may be exaggerated compared to the actual physical thickness. In describing the present invention, well-known components irrelevant to the gist of the present invention may be omitted. When assigning reference numbers to components in each drawing, it should be noted that, whenever possible, identical components are given the same number, even if they appear in different drawings.

[0030] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Therefore, the actual implementation is not limited to the specific embodiments disclosed, and the scope of this specification includes modifications, equivalents, or alternatives within the technical concepts described in the embodiments.

[0031] Although terms such as "first" or "second" may be used to describe various components, these terms should be interpreted solely to distinguish one component from another. For example, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component.

[0032] When it is said that a component is "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but there may also be other components in between.

[0033] Singular expressions include plural expressions unless the context clearly dictates otherwise. In this specification, the terms "comprises" or "has" should be understood to indicate the presence of a described feature, number, step, operation, component, part, or combination thereof, but not to exclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0034] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art. Terms defined in commonly used dictionaries should be interpreted to have a meaning consistent with their meaning in the context of the relevant technology, and will not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0035] Hereinafter, embodiments will be described in detail with reference to the attached drawings. In the description with reference to the attached drawings, identical components are assigned the same reference numerals regardless of the drawing numbers, and redundant descriptions thereof will be omitted.

[0036]

[0037] Figure 1 illustrates the overall flow of a method for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment.

[0038] FIG. 2 is a diagram illustrating a detailed flow of a step of collecting user analysis information of a specific space using Wi-Fi channel status information according to one embodiment.

[0039] And, FIG. 3 is a diagram showing a detailed flow of a learning step of a method for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment.

[0040] First, referring to FIG. 1, a method for analyzing users of a specific space using Wi-Fi channel state information (CSI; Channel State Information) (hereinafter, simply referred to as “method for analyzing users of a specific space using CSI” or more simply “method for analyzing users”) includes a step of collecting a multi-modal dataset (S10), a step of learning an AI model using the multi-modal dataset (S20), and a step of predicting gender / age / location information using the learned model (S30).

[0041] According to one embodiment, an AI learning model based on a multimodal dataset may include a CSI-based gender / age analysis AI model and a CSI-based location prediction AI model. The former CSI-based gender / age analysis AI model can predict gender / age information related to users located in a specific space. The latter CSI-based location prediction AI model can predict location information related to users located in a specific space. Each model may be created through different dataset collection processes and AI model training processes. However, they share the commonality of being a multimodal dataset utilizing CSI data and video data.

[0042] In one embodiment, a CSI-based gender / age analysis AI model and a CSI-based location prediction AI model can be used to accurately predict gender / age / location information related to users located in a specific space. In the present invention, the "specific space" may refer to a space requiring prediction (e.g., a convenience store, supermarket, etc.). The "user" may refer to a person using the space.

[0043] Below, we will explain the process of training a CSI-utilizing gender / age analysis AI model using a multi-modal dataset and predicting gender / age information related to users located in a specific space.

[0044] First, in step (S10) according to one embodiment, the user analysis device collects a multi-modal based dataset with CSI as the first modality and video as the second modality.

[0045] Channel Signal Intensity (CSI) provides detailed information about the state of a wireless channel and quantitatively represents the various changes (e.g., attenuation, phase shift, multipath effects) that a Wi-Fi signal experiences as it passes through its environment. CSI can be a sampled value of the Channel Frequency Response (CFR) or a frequency-domain value of the Channel Impulse Response (CIR). In other words, CSI can be expressed in the form of a CFR, which can be obtained through frequency-domain analysis of the CFR. CIR can be viewed as a time-domain transformation of the CFR. CSI can be obtained by collecting CSI data of Wi-Fi signals in various spaces (e.g., enclosed spaces, open spaces, spaces of various sizes, etc.).

[0046] A video is visual data that captures people or other objects as a digital image or video frame. It is composed of pixels containing color and brightness information, and each pixel can typically have a 2D form with coordinates along the x-axis and y-axis. Video can be acquired by collecting image data captured through CCTV or cameras.

[0047] More specifically, step (S10) may include a step (S12) of designing a dataset and a step (S14) of building a dataset, as illustrated in FIG. 2.

[0048] In step (S12), the user analysis device designs a dataset that inputs CSI information for a predetermined space (x, input data) and outputs 3D pose prediction results, gender results, and age results based on image data for the predetermined space (y, output data).

[0049] 3D pose prediction results based on image data may include, for example, a process called DensePose Estimation. DensePose Estimation maps the human body in a 2D image onto a 3D surface model. DensePose, a system developed by Meta Platforms, Inc., uses deep learning to detect people in 2D images and map them onto the 3D structure of the human body to obtain various poses. DensePose can identify more than 24 major parts of the human body, including joints, arms, head, and torso, and this function enables AI to describe human poses.

[0050] For example, gender-based results based on image data can involve analyzing a person's gender in an image and predicting their gender. Trained using an artificial intelligence neural network, gender can be determined based on factors such as body shape and facial features.

[0051] Age-based results based on image data can, for example, include analyzing a person's age in an image and predicting their age. This process is trained using an artificial intelligence neural network, and, similar to the gender analysis process described above, age can be estimated based on body shape, facial features, and other factors. The age-prediction process can be combined with facial recognition technology to improve accuracy.

[0052] Meanwhile, the present invention is not limited to the above-described embodiments, and various algorithms other than Dense-Pose may be applied for 3D pose prediction based on image data, and various algorithms using features other than the above-described features may be applied for gender prediction and age prediction. In addition, according to another embodiment, the gender result based on image data and the age prediction result based on image data may be a method in which a person directly annotates the 3D pose prediction result based on image data.

[0053] This dataset design creates a multimodal dataset that combines CSI data as spatial information and dense-pose prediction, gender, and age data as image information.

[0054] Next, in step (S14), the user analysis device optimizes the router placement under multiple spatial-environmental conditions and builds a dataset based on the dataset designed in step (S12).

[0055] The router can be a wireless router that transmits and receives Wi-Fi signals.

[0056] Optimizing router placement involves placing routers in optimal locations to maximize wireless signal quality and range. The location and orientation of a router significantly impact signal strength and stability, so placing it in the optimal location minimizes signal loss and improves data accuracy.

[0057] According to one embodiment, multiple spatial-environmental conditions may include spatial environments defined by one or more combinations of enclosed spaces, open spaces, areas less than 100 pyeong (approximately 320 sq ft), and areas greater than 100 pyeong (approximately 320 sq ft). Because each spatial-environmental condition affects the propagation and performance of Wi-Fi signals differently, it is important to build a dataset based on these spatial-environmental conditions.

[0058] For example, a confined space is a space isolated from the outside world by walls or other structures, such as a home, small office, temporary building, convenience store, or basement. Physical obstacles such as walls or ceilings can weaken Wi-Fi signals, and signals can reflect within the space, resulting in multipath interference. Therefore, step (S14) may consider placing the router in the center of the confined space or using multiple access points to expand coverage. Coverage expansion will be discussed later.

[0059] Additionally, for example, an open space is a more expansive space with fewer physical obstacles than the aforementioned enclosed space, such as a large office, a large supermarket, a warehouse, an open area of ​​a university campus, or an entrance to a national park. Considering the need for long-distance signal transmission and the need for users to be able to access Wi-Fi from various locations within the space, step (S14) may consider placing the router at a high location in the open space or distributing access points.

[0060] For example, areas under 100 pyeong (approximately 330 sq ft) can be sufficiently covered by one or two routers, such as small and medium-sized offices, classrooms, small restaurants, and convenience stores. Step (S140) may consider placing routers in central or strategic locations to maximize coverage.

[0061] Additionally, for example, a space larger than 100 pyeong (approximately 330 sq ft) is a large space requiring multiple routers or access points due to its large surface area, such as a large office, conference center, university building, or large supermarket. Step (S14) may prioritize the placement of multiple routers or access points to maintain signal strength and avoid dead zones.

[0062] A multi-modal dataset can be collected through the steps described above (S12 and S14).

[0063]

[0064] Next, step (S20) according to one embodiment is a process of learning the collected multi-modal dataset, which may include a step of converting to a compatible dimension (S22), a step of extracting keypoint information (S24), and a step of designing and learning an optimal layer model (S26), as illustrated in FIG. 3.

[0065] First, in step (S22), the user analysis device can convert the CSI information into a dimension compatible with image data so that it can be applied to a 3D pose prediction neural network model. The 3D pose prediction neural network model may be, for example, a DensePose RCNN model.

[0066] Since CSI data and image data have different formats and dimensions, in step (S22), the two data formats are converted to be compatible, so that the AI ​​model can effectively combine and process the two data sources.

[0067] More specifically, step (S22) may include encoding amplitude and phase data of CSI information and aligning them to the same dimension as the image data through a dimension reduction and merging process. The dimension reduction and merging process converts and merges high-dimensional CSI data into lower-dimensional data that is easier to handle. Since the amplitude and phase data of a CSI signal may be complex and high-dimensional, feature extraction and dimensionality reduction techniques (such as principal component analysis (PCA), autoencoders, etc.) may be used to process such data.

[0068] Next, in step (S24), the user analysis device applies the CSI information converted in step (S22) to a 3D pose prediction neural network model to extract keypoint information. The 3D pose prediction neural network model may be, for example, a DensePose RCNN (Region-based Convolutional Neural Network (R-CNN)) model, and CSI information aligned to the same dimension as the image data can extract keypoint information by utilizing the DensePose RCNN model.

[0069] Keypoint information represents the location of each part of the human body. Dense-Pose RCNN maps keypoints in a 2D image onto the 3D body surface, enabling the precise 3D position and orientation of each body part.

[0070] In step (S26), the user analysis device designs and trains an optimal layer model using the keypoint information extracted in step (S24) as input. This corresponds to the process of determining the structure of a neural network suitable for a specific task and finding a method for effectively training it.

[0071] More specifically, designing an optimal layer model may include determining layer composition, configuring network architecture, and setting hyperparameters. For example, layer composition decisions include determining the type of layer (e.g., convolutional layer, fully connected layer, LSTM layer), the number of layers, the size of each layer, and the number of filters. Furthermore, for example, network architecture can be configured by combining various layers to form an overall network architecture. This process can take into account the characteristics of the input data and the nature of the output to be predicted. Furthermore, for example, hyperparameter settings include setting hyperparameters such as learning rate, batch size, and activation function selection.

[0072] Optimal layer model training can involve data preprocessing, learning algorithm application, validation, and tuning. For example, data preprocessing involves preprocessing input data (CSI signals and image data) into a format that the model can process. This often involves normalization, dimensionality reduction, and data encoding. Another example is learning algorithm application, where an appropriate learning algorithm can be selected to effectively train the model. Loss functions and optimization algorithms can determine how the model learns from given data. Validation and tuning, for example, involve evaluating the trained model using a validation dataset and fine-tuning the model to optimize performance.

[0073] Through the above-described processes, CSI data and video data can be integrated to provide an AI model optimized for accurately predicting gender and age.

[0074] Below, we describe the process of training a CSI-based location prediction AI model using a multi-modal dataset and then predicting locations associated with users in a specific space. This process is largely the same as that described in Figures 1 through 3 above, with a focus on the differences.

[0075]

[0076] FIG. 4 is a diagram illustrating a detailed flow of a step of collecting user analysis information of a specific space using Wi-Fi channel status information according to one embodiment.

[0077] And, FIG. 5 is a diagram showing a detailed flow of a learning step of a method for analyzing users of a specific space using Wi-Fi channel status information according to one embodiment.

[0078] In one embodiment, the user analysis device collects a multi-modal dataset using CSI as the first modality and video as the second modality. This corresponds to step (S10) described above, and in contrast, this will be referred to as step S10'.

[0079] In step (S10'), the user analysis device collects a multi-modal based dataset with CSI as the first modality and video as the second modality.

[0080] More specifically, step (S10') may include a step (S12') of designing a dataset and a step (S14') of building a dataset, as illustrated in FIG. 4.

[0081] In step (S12'), the user analysis device designs a dataset that inputs CSI information for a predetermined space (x, input data) and outputs bounding box coordinate results of object detection based on image data for the predetermined space (y, output data).

[0082] For example, the bounding box coordinates of an object detected based on image data can be the coordinates of a rectangle surrounding a person in an image. The rectangle represents the location and size of the object and can typically be expressed as coordinates (such as the coordinates of the top left and bottom right in the image). Bounding boxes are used to identify the precise location of an object in an image and estimate its size. An object detection model can predict the bounding box for each object in an image.

[0083] This dataset design creates a multimodal dataset that combines CSI data as spatial information and object-detected bounding box coordinates as image information.

[0084] Next, in step (S14'), the user analysis device optimizes the router placement under multiple spatial-environmental conditions and builds a dataset based on the dataset designed in step (S12').

[0085] A multi-modal dataset can be collected through the steps described above (S12' and S14').

[0086]

[0087] Next, step (S20') according to one embodiment is a process of learning the collected multi-modal dataset, and may include a step (S22') of performing final bounding box and loss optimization, as illustrated in FIG. 5.

[0088] In step (S22'), the user analysis device can encode amplitude and phase data of CSI information and perform final bounding box and loss optimization through dimension reduction and merging processes.

[0089] In more detail, the dimensionality reduction and merging process is to transform and merge high-dimensional CSI data into lower-dimensional data that is easier to handle. Since the amplitude and phase data of CSI signals can be complex and high-dimensional, feature extraction and dimensionality reduction techniques (such as principal component analysis (PCA), autoencoders, etc.) can be used to process such data.

[0090] Final bounding box and loss optimization are the processes of measuring and improving how well the model's predictions match the actual values ​​during the AI ​​model training process. The loss function represents the prediction error, and the goal is to minimize this value. By optimizing the loss, the model can more accurately predict object locations and draw bounding boxes. In object detection, metrics such as Intersection over Union (IoU) can be used as a loss function to improve bounding box accuracy. This can induce the model to generate bounding boxes that are as close as possible to the actual object locations.

[0091] That is, step (S22') combines CSI data and image-based object detection results to develop an AI model that predicts human location. This process utilizes bounding boxes to identify object locations and improves the model's prediction accuracy through loss optimization.

[0092] Through the above-described processes, CSI data and video data can be integrated to provide an AI model optimized for accurately predicting a person's location.

[0093] Meanwhile, the above-described process can further optimize the installation environment and define the maximum coverage area. Specifically, the coverage area per router can be determined for each environment, and when installing multiple routers, CSI data can be merged and the merged dataset can be utilized to enhance the location prediction AI model.

[0094] Specifically, the process of verifying router coverage for each environment can optimize router coverage for each environment, thereby increasing the accuracy and reliability of CSI data collection. By analyzing each router's coverage area, the overall network performance can be assessed and adjustments made based on signal strength, stability, and coverage.

[0095] When installing multiple routers, the process of merging CSI data and utilizing the dataset assumes a multi-router environment. In large spaces or complex environments, multiple routers are used to expand network coverage. By integrating CSI data collected from multiple routers, a more comprehensive dataset is built. This data can be used to predict human locations in a variety of environments and conditions. Furthermore, the merged dataset can be used to train the aforementioned location prediction AI model more accurately and efficiently. The model can learn signal changes and patterns that can occur in various environments, thereby improving the accuracy of location prediction.

[0096]

[0097] Referring back to FIG. 1, in step (S30), the user analysis device can predict gender / age / location information related to a user located in a specific space by using the models learned through steps (S10 to S20, and S10' to S20'), i.e., the CSI-utilizing gender / age analysis AI model and the CSI-utilizing location prediction AI model.

[0098] Through this, it is expected that the gender of a user located in a specific space can be determined with an accuracy of over 90%, the age can be determined with an accuracy of over 85%, and the location of a user located in a specific space can be determined with an accuracy of over 95% based on Recall@50.

[0099]

[0100] FIG. 6 is a diagram illustrating an example of a hardware implementation of a user analysis device according to one embodiment. The user analysis device (1000) according to one embodiment may include a data acquisition unit (1010), a processor (1020), a memory (1030), and a display (1040).

[0101] The data acquisition unit (1010) can acquire CSI data and image data for designing a multi-modal based dataset with CSI as the first modality and image as the second modality. The data acquisition unit (1010) can include at least one of a communication unit for acquiring CSI data and a camera for generating image data.

[0102] The processor (1020) collects a dataset, uses the dataset to train a CSI-utilized gender / age analysis AI model and a CSI-utilized location prediction AI model, and can predict gender / age / location information related to a user located in a specific space using the trained models. The processor (1020) can collect a dataset and train the AI ​​model using a neural network model stored in the memory (1030). In addition, the processor (1020) can perform image processing so that the predicted result is displayed on the display (1040). However, the operation of the processor (1020) is not limited thereto, and the operations described in FIGS. 1 to 5 may also be performed.

[0103] The memory (1030) can store a neural network model. The memory (1030) can temporarily or permanently store data required to perform a user analysis method according to one embodiment. For example, the memory (1030) can store CSI data, image data, datasets, a CSI-based gender / age analysis AI model, a CSI-based location prediction AI model, and prediction results for users located in a specific space.

[0104] The display (1040) can display gender / age / location information about a user located in a predicted specific space in the form of image or video content. The display (1040) can also visualize images or videos for each of gender, age, and location.

[0105] The term "unit" as used herein may mean, for example, a unit comprising one or a combination of two or more of hardware, software, or firmware. The term "unit" may be used interchangeably with terms such as module, unit, logic, logical block, component, or circuit. The "unit" may be the smallest unit of an integrally formed component or a part thereof. The "unit" may also be the smallest unit or a part thereof that performs one or more functions. The "unit" may be implemented mechanically or electronically. For example, the "unit" may include at least one of an application-specific integrated circuit (ASIC) chip, field-programmable gate array (FPGA), or programmable-logic device that performs certain operations, whether known or to be developed in the future.

[0106] The embodiments described above may be implemented using hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using a general-purpose computer or a special-purpose computer, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and software applications running on the operating system. Furthermore, the processing device may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible.

[0107] Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure a processing device to perform a desired operation or, independently or collectively, command the processing device. The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave, for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on a computer-readable recording medium.

[0108] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may store program commands, data files, data structures, etc., alone or in combination, and the program commands recorded on the medium may be those specially designed and configured for the embodiment or may be known and available to those skilled in the art of 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, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memories. Examples of program commands include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc.

[0109] The hardware device described above may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

[0110] Although the embodiments described above have been described with limited drawings, those skilled in the art will appreciate that various technical modifications and variations can be applied based on the described embodiments. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

[0111] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims described below.

[0112] While the technical concept of the present invention has been specifically described in accordance with the above preferred embodiments, it should be noted that the above-described embodiments are intended for illustrative purposes only and are not intended to be limiting. Furthermore, those skilled in the art will readily appreciate that various embodiments are possible within the scope of the present invention.

[0113] [Explanation of symbols]

[0114] 1000: User Analysis Device

[0115] 1010: Data Acquisition Department

[0116] 1020: Processor

[0117] 1030: Memory

[0118] 1040: Display

Claims

1. A step of collecting a multi-modal based dataset with CSI as the first modality and video as the second modality; A step of learning a CSI-utilized gender / age analysis AI model and a CSI-utilized location prediction AI model using the above dataset; and A method for analyzing users of a specific space using Wi-Fi channel status information, comprising: a step of predicting gender / age / location information related to users located in a specific space using the above-mentioned learned models; 2. In paragraph 1, The above collecting steps are: A step of designing a dataset that inputs CSI information for a given space and outputs 3D pose prediction results, gender results, and age results based on image data for the given space; and A method for analyzing users of a specific space using Wi-Fi channel status information, comprising: a step of optimizing router placement in multiple space-environment conditions and constructing a dataset based on the designed dataset; 3. In paragraph 2, A method for analyzing users of a specific space using Wi-Fi channel status information, wherein the above multiple space-environment conditions include a space-environment set by a combination of one or more of a closed space, an open space, less than 100 pyeong, and more than 100 pyeong.

4. In paragraph 2, The above learning steps are: A step of converting the above CSI information into a dimension compatible with the image data so that it can be applied to a 3D pose prediction neural network model; A step of extracting keypoint information by applying the above-mentioned converted CSI information to a 3D pose prediction neural network model; and A method for analyzing users of a specific space using Wi-Fi channel status information, comprising: a step of designing and learning an optimal layer model using the above key point information as input; 5. In paragraph 4, A method for analyzing users of a specific space using Wi-Fi channel status information, wherein the above-mentioned converting step includes a process of encoding amplitude and phase data of the CSI information and aligning them to the same dimension as the image data through a dimension reduction and merging process.

6. In paragraph 4, The above 3D pose prediction neural network model is a dense-pose RCNN model, a method for analyzing users of a specific space using Wi-Fi channel status information.

7. In paragraph 1, The above collecting steps are: A step of designing a dataset that inputs CSI information for a given space and outputs bounding box coordinate results of object detection based on image data for the given space; and A method for analyzing users of a specific space using Wi-Fi channel status information, comprising: a step of optimizing router placement in multiple space-environment conditions and constructing a dataset based on the designed dataset; 8. In paragraph 7, The above learning steps are: A method for analyzing users of a specific space using Wi-Fi channel state information, comprising: encoding amplitude and phase data of the above CSI information, performing a dimension reduction and merging process, and performing a final bounding box and loss optimization; 9. In paragraph 7, A method for analyzing users of a specific space using Wi-Fi channel status information, further comprising: a step of confirming the coverage area per router for each of the plurality of space-environment conditions, merging CSI data when installing a plurality of routers, and enhancing the CSI utilization location prediction AI model by utilizing the merged dataset; 10. A computer-readable recording medium storing one or more computer programs including commands for performing the method of any one of claims 1 to 9.

11. As a user analysis device for a specific space using Wi-Fi channel status information, A data acquisition unit for acquiring CSI data and image data for designing a multi-modal based dataset with CSI as the first modality and image as the second modality; A processor that collects the above dataset, uses the dataset to learn a CSI-utilized gender / age analysis AI model and a CSI-utilized location prediction AI model, and uses the learned models to predict gender / age / location information related to users located in a specific space; and A user analysis device comprising a display for displaying the predicted results.

Images