User authentication method and device based on finger vein and finger joint patterns

Abstract

The present invention relates to a method and apparatus for user authentication based on finger vein and finger joint patterns, comprising: an ultrasonic sensor for acquiring an image or video of a finger vein inside a finger (first data); an optical sensor for acquiring an image or video of the external shape of a finger (second data); a region of interest extraction unit for extracting a region of interest for each of the first and second data; a data processing unit for performing segmentation and skeleton processing on the extracted region of interest; a data matching unit for aligning the first and second data after segmentation and skeleton processing are completed; a feature point extraction unit for extracting a plurality of feature points based on the processed first and second data; a feature point registration unit for storing the extracted plurality of feature points in a server or database and registering them for each user; and a user authentication unit for authenticating a user based on the registered feature points.

Description

Technology Field

[0001] The present invention relates to a method and apparatus for user authentication based on finger vein and finger joint patterns. More specifically, the invention provides a method and apparatus for user authentication based on finger vein and finger joint patterns that can precisely identify and authenticate a user by acquiring finger vein data inside the finger using an ultrasonic sensor and acquiring joint shape data outside the finger using an optical sensor, calculating feature points by segmenting and skeletonizing regions of interest extracted from the two data, and comprehensively analyzing the intersection points of the finger vein pattern and the finger joint pattern as well as the structural features of each pattern. Background Technology

[0003] Existing biometric technologies have primarily utilized methods to identify individuals based on external human characteristics such as fingerprints, irises, faces, and voices; however, these technologies have security limitations, including significantly reduced recognition accuracy due to factors such as shooting angle, lighting, humidity, skin damage, or contamination, and easy misidentification by forged images or models.

[0004] In particular, fingerprint recognition had a problem where pattern recognition rates frequently decreased due to epidermal damage or dryness.

[0005] Accordingly, finger vein recognition technology that utilizes the internal vascular structure of the body rather than the external appearance is gaining attention; however, relying solely on finger vein images has limitations, as image alignment errors are prone to occur due to factors such as the position, angle, and pressure of the finger, making it difficult to ensure the consistency and accuracy of authentication.

[0006] Furthermore, existing authentication systems utilizing a single biometric information find it difficult to completely eliminate the possibility of misrecognition due to various external factors, such as biometric similarities between individuals, environmental variations, and noise during image acquisition; therefore, highly reliable multi-biometric information fusion technology is required.

[0007] Accordingly, the present invention aims to provide a fused authentication technology capable of combining and analyzing internal blood vessel patterns and external structural patterns by acquiring internal finger vein data using an ultrasonic sensor and simultaneously collecting external finger joint shape data through an optical sensor.

[0008] Through this, users can be identified reliably and accurately by precisely comparing and verifying the intersection feature points of finger vein patterns and joint patterns while minimizing the influence of changes in finger posture or lighting environments, and furthermore, security against falsification of biometric information can be significantly enhanced.

[0009] Therefore, a user authentication method and device based on finger vein and finger joint patterns are required. Prior art literature

[0011] (Patent Document 0001) KR 10-2361752 B1 (2022.02.08.)(Patent Document 0002) KR 10-1899493 B1 (2018.09.11.)(Patent Document 0003) KR 10-1626837 B1 (2016.05.27.)(Patent Document 0004) KR 10-2025-0035975 A (2025.03.13.) The problem to be solved

[0012] The purpose of the present invention is to provide a user authentication method and apparatus based on finger veins and finger joint patterns that can improve the accuracy and reliability of user authentication by utilizing finger veins and finger joint patterns together. Existing finger vein recognition technologies have problems such as image registration errors occurring depending on the position, angle, or pressure changes of the finger, or the possibility of misrecognition due to inter-individual similarity or environmental factors as they are based on single biometric information.

[0013] To solve these problems, the present invention proposes a high-precision user authentication technology through combined analysis of multiple biometric information by acquiring finger vein data inside the finger via an ultrasonic sensor and collecting joint shape data outside the finger via an optical sensor, and then deriving intersection feature points of finger vein patterns and finger joint patterns by segmenting and skeletonizing regions of interest extracted from the two data.

[0014] The present invention can reliably identify the unique biometric characteristics of a user while minimizing the influence of changes in finger posture or lighting environments, and can simultaneously improve security and accuracy by overcoming the limitations of single biometric authentication. means of solving the problem

[0016] The objective of the present invention described above can be achieved by: an ultrasonic sensor for acquiring an image or video of a finger vein inside a finger (first data); an optical sensor for acquiring an image or video of the external shape of a finger (second data); a region of interest extraction unit for extracting a region of interest for each of the first and second data; a data processing unit for performing segmentation and skeleton processing on the extracted region of interest; a data matching unit for aligning the first and second data after the segmentation and skeleton processing is completed; a feature point extraction unit for extracting a plurality of feature points based on the processed first and second data; a feature point registration unit for storing the extracted plurality of feature points in a server or database and registering them for each user; and a user authentication unit for authenticating a user based on the registered feature points.

[0017] The region of interest extraction unit is controlled to extract a first ROI from the finger vein image, including the first joint of the finger closest to the fingernail and a portion spaced from it by a predetermined distance (22-24 mm) in the vertical direction, and to extract a second ROI from the finger image, including the first joint of the finger and a portion spaced from it by a predetermined distance (22-24 mm) in the vertical direction, respectively, and the data processing unit is controlled to perform segmentation and skeleton processing on the extracted first ROI and second ROI, respectively.

[0018] The above feature point extraction unit is characterized by being controlled to extract the intersection point of the finger vein pattern and the finger joint pattern as a first feature point, extract the feature of the finger joint pattern itself as a second feature point, and extract the feature of the finger vein itself as a third feature point.

[0019] The second feature point is characterized by including at least one of the following: the number of skeleton lines of the joint pattern, the total length of the joint pattern, the ratio of the number of branch points to the number of endpoints of the joint pattern, the skeleton width of the joint pattern, and the position coordinates of the branch points or endpoints of the joint pattern expressed on a normalized finger coordinate system.

[0020] The above third feature point is characterized by including at least one of the endpoint, branch point, and intersection point of the designated vein skeleton line, the line length of the individual vessel skeleton, the distance between the endpoint and the branch point, the branching angle, the curvature of the designated vein skeleton curve, and the distance between the centerlines of the individual vessel skeleton lines.

[0021] The above third feature point is characterized by including at least one of the number of skeleton lines of the designated vein, the total length of the designated vein skeleton lines, the ratio of the number of branch points to the number of endpoints of the designated vein, the width of the designated vein skeleton, and the position coordinates of the branch points or endpoints of the designated vein expressed on a normalized finger coordinate system.

[0022] The above user authentication unit is controlled to identify and authenticate a user based on a plurality of authentication steps, wherein the plurality of authentication steps include a first authentication step that identifies and authenticates a user based on a first feature point, a second authentication step that identifies and authenticates a user based on a second feature point, and a third authentication step that identifies and authenticates a user based on a third feature point.

[0023] The above user authentication unit is characterized by being controlled to identify and authenticate a user by sequentially performing the above first authentication step, the above second authentication step, and the above third authentication step.

[0024] The user authentication unit is characterized by being controlled to identify and authenticate a user by performing at least two selected steps among the first authentication step, the second authentication step, and the third authentication step simultaneously or in sequence.

[0025] According to one embodiment, a user authentication method based on finger vein and finger joint patterns comprises: a step of acquiring an image or video of a finger vein inside a finger (first data) using an ultrasonic sensor; a step of acquiring an image or video of the external shape of a finger (second data) using an optical sensor; a step of extracting an area corresponding to a first joint to a 1.5 joint from the fingertip as a first ROI from the first data, and extracting an area adjacent to the first joint by a predetermined distance from it as a second ROI from the second data; a step of generating first processed data and second processed data by performing segmentation and skeleton processing on each of the first and second ROIs; a step of aligning and matching the positions of the first and second processed data; a step of extracting a plurality of feature points based on the first and second processed data; and a step of storing the extracted plurality of feature points in a server or database and registering them for each user. and a step of authenticating a user based on the registered feature points; wherein the plurality of feature points may include the intersection point of a finger vein pattern and a finger joint pattern, and relative position information of a finger vein branch point detected centered on the intersection point.

[0026] After the step of authenticating a user based on the registered feature points, the method further includes the step of calculating a first authentication reliability based on the user's behavioral habit data; wherein the step of calculating the first authentication reliability based on the user's behavioral habit data may include: a step of obtaining first habit data by analyzing data obtained using the optical sensor to extract the trajectory of the user's finger moving within the shooting area of ​​the optical sensor; a step of obtaining second habit data by analyzing data obtained from the optical sensor to calculate the entry slope of the finger relative to the reference plane based on the reference plane of the shooting area of ​​the optical sensor; a step of obtaining third habit data by analyzing data obtained from the optical sensor to detect the positional shaking of the fingertip in consecutive frame images before and after the point of entry into the shooting area of ​​the optical sensor according to a preset threshold value; a step of registering or updating the user's reference behavioral habit data by storing the first habit data, the second habit data, and the third habit data in a server or database; and a step of calculating the first authentication reliability according to a preset standard by comparing the user's behavioral habit data obtained during authentication based on the feature points with the user's reference behavioral habit data.

[0027] The method further includes the step of calculating a second authentication reliability based on the user's gesture data; wherein the step of calculating a second authentication reliability based on the user's gesture data comprises: a step of acquiring gesture data by analyzing data acquired from the optical sensor to detect a gesture trajectory performed by a finger within the shooting area of ​​the optical sensor; a step of identifying at least one action among a finger contact motion, a finger tilt transition motion, a circular trajectory motion, or a zigzag trajectory motion from the gesture data; and a step of calculating a second authentication reliability according to a preset standard by comparing the identified gesture data with a preset standard gesture data of the user, wherein the step of calculating a second authentication reliability may be performed only when at least one of a first environment corresponding to the user's authentication information registration procedure, a second environment identified as high risk according to a preset standard, and a third environment in which an error occurred during the process of authenticating the user based on the registered feature points is detected.

[0028] The method further includes the step of detecting an abnormal blood vessel pattern and providing a health abnormality warning; wherein the step of detecting an abnormal blood vessel pattern and providing a health abnormality warning may include: the step of analyzing the first data collected using the ultrasonic sensor to detect a blood vessel stenosis pattern according to a first criterion set based on a change in blood vessel cross-sectional area or blood flow velocity; the step of analyzing the first data to detect an abnormal blood vessel branching pattern according to a second criterion set based on a change in blood vessel branching angle or branching interval; the step of generating a user health abnormality signal based on abnormal blood vessel pattern data including the detected blood vessel stenosis pattern and the abnormal blood vessel branching pattern; and the step of transmitting warning data including the generated health abnormality signal to the terminal of the user. Effects of the invention

[0030] According to the present invention, by acquiring finger vein data inside the finger using an ultrasonic sensor and analyzing joint shape data outside the finger using an optical sensor, there is an effect of significantly reducing recognition errors compared to single biometric information-based authentication.

[0031] According to the present invention, by extracting and mutually utilizing the intersection feature points of the finger vein pattern and the finger joint pattern, it is possible to provide stable authentication results even with changes in the position, angle, and pressure of the finger.

[0032] According to the present invention, by generating noise-removed normalized data through region of interest extraction, segmentation, and skeleton processing, it is possible to extract and register user-specific feature points with high precision.

[0033] According to the present invention, since multi-stage verification of finger veins, finger joints, and cross-feature points is possible through a multi-stage feature point-based authentication procedure, it is possible to effectively block attempts at forgery or imitation and improve security.

[0034] According to the present invention, by applying a multi-biometric information fusion method based on finger vein and finger joint patterns, the influence of external environments such as lighting conditions or skin condition is minimized, and the unique biometric characteristics of the user can be stably identified.

[0035] Meanwhile, the effects according to the embodiments are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0037] FIG. 1 is a schematic diagram illustrating the overall configuration of a user authentication device based on finger vein and finger joint patterns according to one embodiment of the present invention. FIG. 2 is a diagram illustrating a region of interest extraction process and a data processing process according to an embodiment of the present invention. FIG. 3 is a diagram illustrating the operation concept of a feature point extraction unit according to one embodiment of the present invention. FIG. 4 is a schematic diagram illustrating the structure of a second feature point according to one embodiment of the present invention. FIG. 5 is a schematic diagram illustrating the topological features of a third feature point according to an embodiment of the present invention. FIG. 6 is a schematic diagram illustrating the structure of a third feature point reflecting a shape-based feature according to an embodiment of the present invention. FIG. 7 is a schematic diagram illustrating the authentication steps of a user authentication device based on finger vein and finger joint patterns according to one embodiment of the present invention. FIG. 8 is a conceptual diagram illustrating the configuration of an authentication procedure in which a plurality of authentication steps are performed sequentially according to an embodiment of the present invention. FIG. 9 is a conceptual diagram illustrating the configuration of an authentication procedure in which at least two of a plurality of authentication steps are performed simultaneously or selectively according to an embodiment of the present invention. FIG. 10 is a flowchart schematically illustrating the overall flow of a user authentication method based on finger vein and finger joint patterns according to one embodiment of the present invention. Specific details for implementing the invention

[0038] Hereinafter, embodiments are described in detail with reference to the attached drawings. However, various modifications may be made to the embodiments, and thus the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, and substitutions to the embodiments are included within the scope of the rights.

[0039] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed forms, and the scope of this specification includes modifications, equivalents, or substitutions that fall within the technical concept.

[0040] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0041] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0042] The terms used in the embodiments are for illustrative purposes only and should not be interpreted as intended to be limiting. 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.

[0043] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the embodiments pertain. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0044] In addition, when describing with reference 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 embodiments, if it is determined that a detailed description of related prior art could unnecessarily obscure the essence of the embodiments, such detailed description is omitted.

[0045] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined only by the scope of the claims.

[0046] In the embodiments of the present invention, unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the embodiments of the present invention.

[0047] The shapes, sizes, ratios, angles, numbers, etc. disclosed in the drawings for explaining embodiments of the present invention are exemplary, and therefore the present invention is not limited to the depicted details. Furthermore, in describing the present invention, if it is determined that a detailed description of related known technology may unnecessarily obscure the essence of the present invention, such detailed description is omitted. Where terms such as "includes," "has," or "is made up" are used in this specification, other parts may be added unless "only" is used. Where a component is expressed in the singular, it includes cases where it includes the plural unless specifically stated otherwise.

[0048] In interpreting the components, they are interpreted to include a margin of error even in the absence of a separate explicit statement.

[0049] In the case of describing a positional relationship, for example, when the positional relationship between two parts is described using expressions such as 'on,' 'upper,' 'lower,' or 'next to,' one or more other parts may be located between the two parts unless 'immediately' or 'directly' is used.

[0050] When elements or layers are referred to as "on" another element or layer, this includes cases where another layer or element is placed directly on top of or in between. Throughout the specification, the same reference numerals refer to the same components.

[0051] The size and thickness of each component shown in the drawings are illustrated for convenience of explanation, and the present invention is not necessarily limited to the size and thickness of the illustrated components.

[0052] The features of each of the various embodiments of the present invention may be combined or combined with one another, either partially or wholly, and as will be fully understood by those skilled in the art, various technical interlocking and operation are possible, and each embodiment may be implemented independently of one another or together in an interlocking relationship.

[0053] Among the attached drawings, FIG. 1 is a schematic diagram illustrating the overall configuration of a finger vein and finger joint pattern-based user authentication device according to an embodiment of the present invention; FIG. 2 is a diagram illustrating a region of interest extraction process and a data processing process according to an embodiment of the present invention; FIG. 3 is a diagram illustrating the operation concept of a feature point extraction unit according to an embodiment of the present invention; FIG. 4 is a diagram illustrating the structure of a second feature point according to an embodiment of the present invention; FIG. 5 is a diagram illustrating the topological features of a third feature point according to an embodiment of the present invention; FIG. 6 is a diagram illustrating the structure of a third feature point reflecting shape-based features according to an embodiment of the present invention; FIG. 7 is a schematic diagram illustrating the authentication steps of a finger vein and finger joint pattern-based user authentication device according to an embodiment of the present invention; FIG. 8 is a conceptual diagram illustrating the configuration of an authentication procedure in which a plurality of authentication steps are performed sequentially according to an embodiment of the present invention; and FIG. 9 is a diagram illustrating the configuration of an authentication procedure in which at least two of a plurality of authentication steps are performed simultaneously or selectively according to an embodiment of the present invention. Figure 10 is a conceptual diagram and is a flowchart schematically illustrating the overall flow of a user authentication method based on finger vein and finger joint patterns according to one embodiment of the present invention.

[0054] A user authentication device (1) based on finger vein and finger joint pattern according to the present invention (hereinafter referred to as the user authentication device (1)) comprises: an ultrasonic sensor (100) for acquiring an image or video of a finger vein inside a finger (first data); a light sensor (110) for acquiring an image or video of the external shape of a finger (second data); a region of interest extraction unit (120) for extracting a region of interest for each of the first data and second data acquired from the ultrasonic sensor (100) and the light sensor (110); a data processing unit (130) for performing segmentation and skeleton processing on the region of interest of the first data and second data extracted by the region of interest extraction unit (120); a data matching unit (140) for aligning the first data and second data processed by the data processing unit (130); a feature point extraction unit (150) for extracting a plurality of feature points based on the processed first data and second data; and a feature point that stores the extracted plurality of feature points in a server or database (200) and registers them for each user. It consists of a user authentication unit (170) that identifies and authenticates a user by performing a multi-stage authentication procedure based on a registration unit (160) and registered feature points.

[0055] At this time, the user authentication unit (170) can authenticate the user by utilizing the first feature point (intersection point of the finger vein pattern and the finger joint pattern), the second feature point (structural feature of the finger joint pattern), and the third feature point (topographical and shape-based feature of the finger vein pattern) in stages or in parallel.

[0056] Additionally, the user authentication device (1) is connected to a server or database (200) via a network to perform feature point registration, comparison, and authentication result verification, and can operate together with a user terminal (300) that transmits and receives authentication requests and authentication results.

[0057] In this way, the user authentication device (1) can provide high accuracy and security compared to existing authentication methods that rely on single biometric information by fusing and analyzing internal blood vessel data through the ultrasonic sensor (100) and external node shape data through the optical sensor (110).

[0058] The ultrasonic sensor (100) is configured to acquire an image or video (first data) of a finger vein present inside the user's finger, and can detect internal biometric information in the user authentication device (1).

[0059] Here, the first data refers to data obtained by imaging the vascular structure within the finger using ultrasound signals, and may include morphological and topographical information such as the distribution, thickness, branching shape, curvature, and directionality of the blood vessels, but is not limited thereto.

[0060] This first data is biometric information that quantifies and visualizes the unique pattern of the internal vein of the user's finger, and is subsequently used as basic data to be analyzed in the region of interest extraction unit (120).

[0061] The ultrasonic sensor (100) includes a transmitter and a receiver, the transmitter transmits ultrasonic waves of a certain frequency band toward the surface of the finger, and the receiver detects ultrasonic signals reflected back from the internal tissues and blood vessel boundaries of the finger. The detected reflected signals are converted into electrical data through an internal signal processing circuit, and then converted into image data of the finger vein, i.e., first data, through an image reconstruction algorithm. In this process, detailed features such as the branching point, thickness, curvature, and blood flow direction of the blood vessel can be clearly expressed.

[0062] The ultrasonic sensor (100) can be configured to simultaneously detect vascular structures in the longitudinal and cross-sectional directions of the finger by arranging a plurality of transducer elements in a linear or matrix form. For example, the linear array method enables fast scanning, while the matrix array method enables the acquisition of cross-sectional images with greater detail. These array forms can be varied depending on the design purpose and the required level of accuracy.

[0063] Additionally, the ultrasonic sensor (100) may include an automatic sensitivity adjustment function to obtain a stable image even when the intensity of the reflected signal changes due to changes in the position of the finger or deviations in contact pressure. This allows for securing an image of a certain quality even if the position or angle at which the same user places the finger changes slightly. This configuration is merely an exemplary form and may be implemented in a different automatic correction method depending on the embodiment.

[0064] The ultrasonic sensor (100) can automatically start operation the moment a finger is placed on the recognition area to continuously acquire multiple frame images for a short period of time, and then perform temporal averaging or adaptive filtering to remove noise and disturbances and generate first data including clear boundary information. The generated first data includes quantified bio-information such as morphological features of internal blood vessels in the finger, distance between branching points, blood vessel thickness, curvature, and directionality, and is subsequently transmitted to the region of interest extraction unit (120) to be used to designate the area to be analyzed.

[0065] Unlike optical-based sensors, the ultrasonic sensor (100) is not affected by external factors such as skin surface condition, lighting intensity, and skin color, so it can minimize recognition errors that may occur in appearance-based authentication technologies such as fingerprint recognition or face recognition. In other words, it can maintain consistent image quality even under changes in conditions such as external lighting, skin damage, or dryness, thereby preventing a decrease in recognition performance due to environmental changes.

[0066] In this way, the ultrasonic sensor (100) can non-invasively acquire the finger vein pattern present inside the user's finger, thereby reliably generating first data of a consistent quality without being affected by external factors. Since this internal blood vessel information is a unique biosignal in which physiological differences between individuals are clearly evident, it can secure a much higher level of reliability and security than existing authentication methods that use only external information.

[0067] The optical sensor (110) is configured to acquire an external shape image or video (second data) of a finger, and can acquire external structure information including a finger joint pattern from the user authentication device (1).

[0068] Here, the second data is data that visualizes external features observable on the surface of the finger, and may include nail boundaries, finger contours, joint folds, ridge patterns on the skin surface, curvatures and protrusions between joints, etc., but is merely an example and is not limited thereto.

[0069] This second data may include shape, contour, and texture information for identifying finger joint patterns, and can subsequently be used as basic data for specifying the area to be analyzed in the area of ​​interest extraction unit (120).

[0070] The light sensor (110) may include a lens, an image sensor, and a lighting module to accurately capture the shape of the finger's exterior. The lens may be configured with a focal length and distortion correction parameters to capture the contours of the finger's surface and joint creases without distortion, and the image sensor may acquire stable second data even in situations where there is minute movement of the finger through continuous frame shooting. Additionally, the lighting module may maintain uniform illumination in the shooting area to minimize contour loss caused by diffuse reflection or shadows, and may apply a polarizing filter as needed to suppress surface reflection components and enhance the contrast of low-contrast patterns. This configuration is merely an exemplary form of implementation, and depending on the embodiment, a visible light or near-infrared light source may be used, and a fixed exposure or automatic exposure control method may be adopted.

[0071] The optical sensor (110) can collect multiple frames for a certain period of time from the moment the finger enters the shooting area. This provides sufficient information so that the region of interest extraction unit (120) can stably designate the area to be analyzed based on the first joint of the finger.

[0072] The light sensor (110) can generate second data with reduced shaking and noise by performing time averaging, motion correction, or distortion correction processing on a plurality of captured frames, and can perform perspective correction and magnification normalization to correct for deviations in size or position of the finger.

[0073] The second data obtained in this way is transmitted to the data processing unit (130) and can quantify the structure of the finger joint pattern through segmentation and skeleton processing. Subsequently, it can be aligned based on the same finger coordinate system as the first data in the data matching unit (140) and can be used as an input for extracting structural features (second feature points) of the finger joint pattern in the feature point extraction unit (150).

[0074] Additionally, the second data obtained from the optical sensor (110) can provide quantifiable structural information such as the number of lines, total length, ratio of branch points to endpoints, skeleton width, and location coordinates of branch points or endpoints when representing the skeleton. This data can be advantageous for the data matching unit (140) to perform intersection-based alignment when aligning with the finger vein pattern of the first data, and the user authentication unit (170) can improve the accuracy and reliability of the user authentication process by utilizing the first feature point (intersection point of the finger vein pattern and finger joint pattern) and the second feature point (structural feature of the finger joint pattern) in stages or in parallel.

[0075] The light sensor (110) can automatically adjust exposure, gain, and lighting intensity to respond to changes in ambient light or a decrease in contrast caused by partial occlusion of the finger. In addition, by acquiring multiple images for a certain period of time from the moment the user's finger is placed within the shooting area, constant image quality can be maintained even with changes in shooting conditions.

[0076] As a result, the optical sensor (110) can reliably acquire second data, which is an external shape image or video of a finger, and clearly express the contour and structure of the finger joint pattern, thereby ensuring the input quality of subsequent processing steps, including region of interest designation, segmentation and skeleton processing, data alignment and feature point extraction.

[0077] Through this, the user authentication device (1) can achieve improved accuracy, stability, and security compared to a method relying on a single biometric information by fusion analyzing the first data obtained from the ultrasonic sensor (100) and the second data obtained from the optical sensor (110).

[0078] As shown in FIG. 2, the region of interest extraction unit (120) is configured to extract a region of interest to be analyzed for each of the finger vein image or video (first data) provided from the ultrasonic sensor (100) and the external shape image or video (second data) of the finger provided from the optical sensor (110), and can play a key role in the preprocessing stage of the user authentication device (1).

[0079] Here, the first data is information that visualizes the internal vascular structure of the finger, and is internal biological data including the distribution, thickness, branching shape, curvature, etc. of the blood vessels, and the second data is image data that represents the external shape and joint structure of the finger, and may include external visual patterns such as contours, joint wrinkles, and fingernail boundaries. The region of interest extraction unit (120) can analyze these first and second data to calculate a first ROI (Region of Interest) and a second ROI, respectively, so as to designate a region corresponding to the same location of the same finger.

[0080] Here, the first ROI refers to an area including the first joint of the finger closest to the fingernail in the first data and a section spaced apart from it by a predetermined distance (e.g., 22–24 mm) in the direction of the fingertip and the direction of the palm, respectively, and the second ROI may refer to an area defined by the same criteria in the second data. At this time, the distance or size between ROIs may be adjusted according to sensor resolution, finger size, image magnification, etc., and this is merely an exemplary setting and may be configured differently depending on the embodiment.

[0081] The region of interest extraction unit (120) can consistently determine the center, size, aspect ratio, and rotation angle of the first ROI and the second ROI by setting the length direction of the finger as an internal reference axis and utilizing the position of the fingernail boundary and the first joint as reference points.

[0082] For example, the position of the first node can be detected by analyzing the change in curvature of the node fold obtained from the optical sensor (110), and then the corresponding position of the first data can be calculated by referring to the blood vessel density change pattern obtained from the ultrasonic sensor (100). This method is merely an exemplary method, and depending on the embodiment, the ROI position may be calculated by using various techniques such as boundary detection, pattern matching, and histogram-based contrast analysis in combination.

[0083] The region of interest extraction unit (120) can confirm the node position by analyzing changes in tissue boundaries, blood vessel density, and reflection intensity of the ultrasound image for the first data and verifying whether it matches the reference point detected in the second data, thereby securing an accurate correspondence. The first ROI and the second ROI are finally determined by calculating an area extended by a predetermined distance (e.g., 22~24 mm) in the finger length direction centered on the reference point calculated in this way, and the center coordinates, boundary coordinates, rotation angle, and magnification information of each ROI can be generated together as metadata.

[0084] For example, if a finger is captured in a tilted state, the rotation angle of the ROI boundary can be corrected to save it in an aligned form; however, this is merely one example, and the ROI correction algorithm or metadata configuration method may differ in actual implementation.

[0085] The region of interest extraction unit (120) can perform size normalization, perspective correction, and distortion correction processes to reduce size discrepancies caused by changes in finger posture or deviations in shooting distance. For the same user, a reference frame-based tracking technique or an inter-frame matching algorithm may be applied to ensure consistency of the ROI position. Additionally, if there is slight image shaking due to changes in finger contact pressure or light reflection, time averaging or multi-frame synthesis may be performed to control the ROI boundary so that it is stably defined. These normalization and correction processes are merely exemplary methods, and depending on the embodiment, various algorithms such as filtering, optical flow-based correction, or adaptive matching techniques may be applied.

[0086] The region of interest extraction unit (120) can transmit the first ROI and second ROI defined in this way to the data processing unit (130) and control the data processing unit (130) to perform segmentation and skeleton processing for each ROI. The transmitted control information may include coordinate system parameters of the ROI, an effective region mask, a boundary reliability index, etc., and the data processing unit (130) can refer to this to remove background pixels and stably extract the central structure of the finger joint pattern and finger vein pattern. The detailed configuration of the control information is merely an exemplary item and can be adjusted according to the embodiment to suit the purpose of analysis.

[0087] That is, the region of interest extraction unit (120) can improve the accuracy and reliability of the segmentation, skeleton, alignment, and feature point extraction processes performed by the data processing unit (130), the data matching unit (140), and the feature point extraction unit (150) by precisely matching the same biological part of the same finger in the first data and the second data.

[0088] Through this, the user authentication device (1) can minimize the alignment error between the finger vein pattern and the finger joint pattern, and subsequently improve the stability and recognition rate of the multi-step authentication procedure performed by the user authentication unit (170).

[0089] As shown in FIG. 2, the data processing unit (130) receives image data of each of the first ROI and second ROI provided by the region of interest extraction unit (120), and can refine and process the finger vein pattern and finger joint pattern within the image.

[0090] That is, the data processing unit (130) can perform processes such as image segmentation, centerline extraction, coordinate alignment, and noise removal based on respective regions of interest for the finger vein image (first data) obtained from the ultrasonic sensor (100) and the finger external shape image (second data) obtained from the optical sensor (110).

[0091] The data processing unit (130) can perform segmentation to identify valid areas within the first ROI and the second ROI.

[0092] Segmentation is a processing step for separating structures of interest (e.g., blood vessels, node contours) and the background within an image. The data processing unit (130) may generate an effective region mask based on pixel brightness, texture, and boundary information, but this is merely an example, and the segmentation method may be replaced with various algorithms such as deep learning-based segmentation, boundary detection, and region growing depending on the embodiment.

[0093] The data processing unit (130) can generate a centerline-shaped skeleton based on the segmentation results.

[0094] Skeleton processing is a process of extracting topological axes by simplifying the structure of the vein and finger joint patterns, and the data processing unit (130) can remove unnecessary outlines while preserving the centerline structure of each pattern.

[0095] For example, in the case of finger vein images, the central axis of the blood vessel, and in the case of finger joint images, the centerline of the crease, can be extracted and converted into a skeleton.

[0096] This skeleton processing method is merely an example, and depending on the embodiment, distance transform-based thinning or phase-preserving filters may also be applied.

[0097] The data processing unit (130) can normalize the skeletonized data.

[0098] Normalization is a process of consistently adjusting the size, rotation, and position based on the same finger coordinate system, and the data processing unit (130) can match the coordinate systems of each image by referencing the size and rotation angle information of the ROI as metadata.

[0099] For example, if a finger is photographed at an angle, the rotation angle of the skeleton can be corrected to align it with the reference axis (the direction of finger length).

[0100] This normalization process is merely an example, and geometric alignment, coordinate transformation, or statistical standardization methods may be applied depending on the embodiment.

[0101] The data processing unit (130) can perform noise removal and quality improvement processing based on the normalized first ROI and second ROI.

[0102] For example, Gaussian filters, median filters, or wavelet-based noise suppression techniques can be applied to remove point noise caused by illumination non-uniformity or ultrasonic reflection within an image.

[0103] Additionally, the data processing unit (130) can improve image quality so that the boundaries between blood vessels and wrinkles are clearly distinguished through brightness contrast enhancement, but this is merely an example, and depending on the embodiment, the filter type, correction range, or interpolation method may be set differently.

[0104] The data processing unit (130) can output the first processing data and the second processing data generated through the above process.

[0105] Here, the first processed data is data containing the skeleton and structural metadata of the vein image, and can be used to extract the third feature point (topographic and shape-based features of the vein pattern).

[0106] The second processed data is data containing skeleton and boundary information of a finger joint image, and can be used to extract a second feature point (structural feature of the finger joint pattern).

[0107] The data processing unit (130) transmits these two data to the data matching unit (140) and can provide a formatted output structure so that they can be matched based on the same coordinate system in a subsequent step.

[0108] Through this, the data processing unit (130) processes the finger vein and finger joint image data into a structural form based on the ROI defined in the region of interest extraction unit (120), thereby providing a foundation for the first, second, and third feature points to be stably calculated in the feature point extraction unit (150).

[0109] As a result, the data processing unit (130) can improve the recognition reliability and processing efficiency of the entire user authentication process by removing unnecessary elements of the image and ensuring structural consistency.

[0110] The data matching unit (140) can align and match the first processing data and the second processing data generated by the data processing unit (130) so that they can be compared based on the same finger coordinate system.

[0111] Here, the first processed data refers to the result of a finger vein pattern obtained by performing segmentation and skeleton processing on the first ROI, and the second processed data refers to the result of a finger joint pattern obtained by performing segmentation and skeleton processing on the second ROI.

[0112] The data matching unit (140) can sequentially or repeatedly perform an alignment step to match spatial relationships so that two processed data represent the same position of the same finger, and a matching step to minimize alignment errors by determining the corresponding relationship of the skeleton structure in the matched coordinate system.

[0113] The data matching unit (140) can perform initial alignment by referring to the reference point and metadata provided by the region of interest extraction unit (120). In the initial alignment, a finger coordinate system is established with the finger length direction as the reference axis, and the center, rotation angle, and scale of the two processed data can be primarily matched by using the fingernail boundary and the first joint position as reference points.

[0114] For example, the corresponding position of the first data can be mapped based on the first node position detected in the second data, and the principal axes of the two skeletons can be aligned through similar transformations (translation, rotation, isotropic scaling). This is just one example, and depending on the embodiment, linear transformations including anisotropic scaling or shear transformation may also be applied.

[0115] The data matching unit (140) can further reduce the residual position error of the two skeletons by performing fine alignment after the initial alignment. In the fine alignment, the endpoint candidates and branching point candidates of the skeletons are selected as a set of reference points, and the transformation parameters can be updated in a way that minimizes the sum of the distances between the reference points.

[0116] For example, iterative nearest point (ICP) matching, Procrustean analysis, and Euclidean distance minimization-based point matching techniques may be applied, and RANSAC, trimmed mean, or weighted least squares may be used in combination to reduce the influence of disturbance points. These procedures are merely exemplary implementations and may be replaced, depending on the embodiment, with graph matching, Hausdorff distance-based matching, matching using Shape Context, etc.

[0117] The data matching unit (140) may include a nonlinear correction step to correct non-rigid deformation caused by micro-deformation of the finger or changes in contact pressure. The nonlinear correction can use deformation models such as thin plate splines or freeform deformation grids (B-spline FFD) to mitigate local mismatches and finely shift local coordinates within a range that preserves the phase of the skeleton. This is merely one example and may be implemented as multi-resolution matching or normalized flow-based correction depending on the embodiment.

[0118] The data matching unit (140) can calculate a matching error and a quality indicator to quantify the reliability of alignment and matching. The quality indicator may include the mean squared error between corresponding points, Hausdorff distance, branch point matching rate, endpoint matching rate, and the degree of overlap of the intersecting area, and may generate a final transformation matrix or transformation field as metadata and provide it to the feature point extraction unit (150). If necessary, auxiliary information such as an intersecting mask (potential intersection area between the vein skeleton and the finger joint skeleton), a list of candidate corresponding branch point pairs, and an alignment index of the skeleton width profile may be output together. The items and calculation method of this auxiliary information are exemplary configurations and may be adjusted to meet algorithm requirements according to the embodiment.

[0119] The data matching unit (140) can perform multi-stage verification and fallback procedures to ensure processing stability. It can check whether the initial alignment converges and attempt alternative initialization (such as selecting a different set of reference points) if it falls below a threshold value. If the disturbance point ratio exceeds a threshold value during the fine alignment process, it can reapply the disturbance point removal rule and perform re-alignment. If the final alignment quality falls below the threshold, it can perform fallbacks such as adjusting the number of retries, selecting an alternative algorithm path, or applying a conservative variation model. These procedures are exemplary flows and may be modified according to the embodiment to accommodate system delay, processing resources, and target accuracy.

[0120] The data matching unit (140) can output the first processed data and the second processed data, which have completed segmentation and skeleton processing, in an aligned state within the same finger coordinate system. This output can reduce the search space of the skeleton intersection candidate location so that the feature point extraction unit (150) can stably calculate the first feature point (the intersection point of the finger vein pattern and the finger joint pattern), and can support the direct comparison and calculation of the second feature point (structural feature of the finger joint pattern) and the third feature point (topographical and shape-based feature of the finger vein pattern) within the same coordinate system. Additionally, in an operation scenario where the user authentication unit (170) performs the first authentication step, the second authentication step, and the third authentication step sequentially, or performs two or more selected steps simultaneously or sequentially, it can provide a basis for consistent step-by-step score calculation and threshold determination by referring to the alignment and matching results and quality indicators.

[0121] In conclusion, the data matching unit (140) can perform precise coordinate alignment based on coordinate, size, and rotation information defined by the first ROI and the second ROI, with a function centered on aligning the first data and the second data, which have completed segmentation and skeleton processing, in the same finger coordinate system.

[0122] This allows for the creation of a matching environment where the first, second, and third feature points can be reliably calculated, and enables stable support for the operation of multiple authentication procedures in a stepwise or parallel manner.

[0123] As a result, the data matching unit (140) can reduce image matching errors caused by changes in finger posture, shooting angle, contact pressure, etc., and can subsequently improve the accuracy and consistency of feature point extraction and user authentication steps.

[0124] As shown in FIGS. 3 to 6, the feature point extraction unit (150) receives the first processed data and the second processed data generated by the data processing unit (130) as input and can produce a plurality of feature points for user authentication.

[0125] In this case, the first feature point refers to the intersection point of the finger vein pattern and the finger joint pattern, the second feature point refers to the structural features of the finger joint pattern itself, and the third feature point refers to the features of the finger vein itself.

[0126] The feature point extraction unit (150) can calculate feature points under standardized conditions so that two skeletons can be compared on the same standard by referring to the alignment result and finger coordinate system information provided by the data matching unit (140).

[0127] The feature point extraction unit (150) can detect the intersection point between the vein skeleton of the first processed data and the finger joint skeleton of the second processed data to calculate the first feature point.

[0128] Specifically, the feature point extraction unit (150) can calculate the normalized coordinates of the intersection point, the tangent direction of the two skeletons, and the intersection angle to calculate the first feature point, and can calculate contextual information around the intersection point (e.g., distance to an adjacent branch point, distance to an adjacent endpoint) as an additional indicator. This configuration of additional indicators is merely an example and, depending on the embodiment, may be replaced or supplemented with the curvature of the intersection point, local skeleton width, topological connectivity, etc.

[0129] Additionally, the feature point extraction unit (150) can quantitatively analyze the skeleton of the finger joint pattern to calculate the second feature point.

[0130] At this time, the second feature point refers to a structural feature of the finger joint pattern itself and may include at least one of the number of skeleton lines of the joint pattern, the total length of the joint pattern skeleton lines, the ratio of the number of branch points to the number of endpoints of the joint pattern, the width of the joint pattern skeleton (relative thickness information based on the width profile), and the position coordinates of the branch points or endpoints of the joint pattern expressed on a normalized finger coordinate system.

[0131] Specifically, the feature point extraction unit (150) can first aggregate the number of skeleton lines according to connectivity criteria to calculate the second feature point, thereby reflecting the complexity of the nodal folds, and can apply a minimum length threshold to exclude meaningless short lines. Then, by calculating the total length of the skeleton lines as the sum of the arc lengths of all line segments and converting it into physical units using the scale information of the second ROI, it is possible to support distinguishing differences in extension even when having the same number of lines.

[0132] In addition, the ratio of the number of branch points to the number of end points can be calculated based on the topological connectivity of the skeleton graph to provide a summary indicator of structural complexity, and in special cases where the number of end points is 0, a correction rule of adding a small constant (ε) to the denominator or an alternative of listing pairs (number of branch points, number of end points) instead of the ratio can be applied.

[0133] At this time, the width of the node pattern skeleton is defined as the sum of the orthogonal distances from each arc length position of the centerline to both boundary lines, and the relative thickness change can be quantified by calculating summary statistics such as the mean, standard deviation, maximum / minimum, or interval histogram of the width profile arranged therein.

[0134] The normalized positional coordinates of a branching point or endpoint can be expressed by normalizing them to be invariant to translation, rotation, and scaling changes using a reference point (e.g., the position of the first joint) and a reference length in a finger coordinate system with the finger length direction as the principal axis, and can be extended to positional-contextual features that combine contextual information such as local tangent direction, curvature, and adjacent line segment lengths as needed.

[0135] The feature point extraction unit (150) can package each indicator into a single second feature point vector and record reliability for each indicator such as skeleton connectivity, boundary clarity, and detection stability, and the length, width, and coordinate values ​​can be re-expressed as a ratio to the reference length to correct for individual differences or differences in shooting conditions.

[0136] For example, the number of skeleton lines and the total length can be stored as normalized values ​​by dividing them by the reference length, and the width profile can be represented for comparison using a length normalization axis. These definitions, calculations, normalization, and quality control procedures are merely examples, and depending on the embodiment, threshold values, correction constants, summary statistics, and coordinate normalization methods may be configured differently to suit system requirements and operational policies.

[0137] Through this, the feature point extraction unit (150) can consistently calculate the second feature point according to a clear rule and provide it along with reliability information so that the user authentication unit (170) can reliably utilize it in stepwise or parallel authentication.

[0138] In addition, the feature point extraction unit (150) can quantitatively analyze the vein skeleton represented by the first processing data in terms of topological and shape, respectively, to calculate the third feature point.

[0139] At this time, the third feature point refers to a set of indicators including topographic feature points and shape-based feature points of the vein pattern, and can be calculated in a normalized coordinate system by referring to the alignment result and finger coordinate system information provided by the data matching unit (140).

[0140] Topographic feature points may include at least one of the endpoints, branches, and intersections of the designated vein skeleton lines, the line length of individual vessel skeletons, the distance between the endpoints and branches, the branching angle, the curvature of the designated vein skeleton curve, and the distance between the centerlines of individual vessel skeleton lines.

[0141] The shape-based feature points may include at least one of the number of skeleton lines of the designated vein, the total length of the designated vein skeleton lines, the ratio of the number of branch points to the number of endpoints of the designated vein, the width of the designated vein skeleton, and the position coordinates of the branch points or endpoints of the designated vein expressed on a normalized finger coordinate system.

[0142] By utilizing the node and edge structure of a skeleton graph in terms of topography and scale and shape indicators such as the number of lines, total length, and width profile in terms of geometry, it is possible to simultaneously ensure identical user reproducibility and differentiation between users. This analysis procedure is merely an example, and the algorithm configuration and parameters may be set differently depending on the embodiment.

[0143] Among the topographic feature points, endpoints can be defined as nodes with a degree of 1 in a skeleton graph, branching points can be defined as nodes with a degree of 3 or higher, and intersection points can be defined as complex connected nodes where multiple lines meet. The feature point extraction unit (150) extracts normalized coordinates of these nodes and can apply quality rules such as minimum connection length and minimum angle variation to remove fine disconnections caused by noise.

[0144] The number and location distribution of endpoints, branching points, and intersection points are key indicators that summarize the topological complexity of the vein pattern, and can subsequently be used as input for graph matching or structural similarity calculation in the comparison step of the user authentication unit (170).

[0145] The line length of an individual vascular skeleton can be calculated by summing the arc lengths of segments separated by endpoints and branching points. The feature point extraction unit (150) calculates the segment length using a polyline length summing method and can convert it into physical units using magnification information provided by the region of interest extraction unit (120) or the data processing unit (130). If necessary, invariance with respect to differences in imaging magnification or finger size can be ensured by normalizing by dividing by a reference length (e.g., the reference value of the length axis of the first ROI or the second ROI). Summary statistics such as the average, standard deviation, upper and lower quantiles, and length histogram of the segment length can also be included in the third feature point.

[0146] The distance between an endpoint and a branch point can be defined as the shortest path length in segment units or the Euclidean distance in a normalized coordinate system. The feature point extraction unit (150) tracks the segments connected from each branch point to calculate a set of distances to adjacent endpoints and can record the mean, variance, quantiles, and minimum / maximum values ​​of the distance distribution. Whether to adopt the path-based distance or the straight-line distance indicator may be selected according to the operation policy, and the two indicators may be stored together to increase the robustness of the comparison step. These selection rules are merely examples and may be configured differently depending on the embodiment.

[0147] The branching angle can be defined as the angle between the tangent vectors of two segments meeting at the branching point. The feature point extraction unit (150) can estimate the segment direction around the branching point (e.g., tangent vectors using two points separated by a certain arc length from the branching point) and calculate the minimum value or representative angle of the interior angle. Since the mean, variance, and histogram of the branching angle reflect the regularity of the blood vessel branching shape, they can be used as effective indicators for differentiation among users. The tangent estimation length and the interior angle calculation rule are merely examples and may be set differently depending on the embodiment.

[0148] The curvature of the venous blood vessel skeleton curve can be defined as the rate of change of direction according to the arc length. The feature point extraction unit (150) can estimate the curvature through polynomial approximation, local rotation angle difference, or curve interpolation, and can calculate the average curvature, maximum curvature, curvature variance, and the ratio of high curvature intervals. The curvature value can reduce noise sensitivity by combining skeleton smoothing or multi-resolution analysis. The curvature estimation method and summary statistical items are merely examples and may be changed according to the embodiment.

[0149] The distance between the centerlines of individual blood vessel skeleton lines can be defined as a statistical value of the nearest distance from a sample point on one skeleton line to another skeleton line. The feature point extraction unit (150) can calculate the distance distribution by nearest point search or co-segment cell-based approximation and record the average interval, minimum / maximum interval, quantile, etc. Since this indicator quantifies the interval pattern between blood vessels running parallel or adjacently, it is possible to simultaneously ensure reproducibility within the same user and differentiation between users.

[0150] As shape-based feature points, the number of skeleton lines can be defined as the number of connection components or paths, and the total length of the designated vein skeleton lines can be calculated as the sum of all segment lengths.

[0151] The ratio of the number of branch points to the number of end points can be used as a summary indicator of topological complexity, and if the number of end points is 0, a correction of adding a small constant ε to the denominator or an alternative of (number of branch points, number of end points) pairs can be applied.

[0152] The width of the skeleton can be expressed as a width profile by measuring the local width at the centerline using distance transformation-based boundary estimation, and may include statistics such as mean, standard deviation, maximum / minimum, and interval histograms.

[0153] The position coordinates of a branching point or endpoint can be stored by normalizing them using a reference point and reference length in a finger coordinate system to be invariant to changes in translation, rotation, or scaling, and contextual information such as local tangent direction, local curvature, and adjacent segment length can be combined if necessary. The selection of these indicators and the normalization method are merely examples and may be configured differently depending on the embodiment.

[0154] The feature point extraction unit (150) can construct a finger vein structure signature by combining topographic feature points and shape-based feature points, and can convert continuous values ​​such as coordinates, length, angle, curvature, and width into a normalized finger coordinate system and then encode them into an expression that is invariant to rotation, translation, and scaling changes. For each indicator, quality indicators such as skeleton connectivity, boundary clarity, and estimation stability can be included to quantify reliability, and if an outlier is detected, a reprocessing request or a candidate reweighting procedure can be applied. This packaging, normalization, and quality control process is merely an example, and depending on the embodiment, the database schema, transmission format, and threshold settings may be adopted differently.

[0155] Consequently, the third feature point can be composed of a normalized feature vector that encompasses the phase and shape of the vein pattern, and can support the user authentication unit (170) in achieving high accuracy and reliability in a stepwise or parallel authentication scenario.

[0156] The feature point extraction unit (150) can convert the calculated feature values ​​(coordinates, length, angle, curvature, width, etc.) into a normalized finger coordinate system and encode them into an expression that is invariant to rotation, movement, and scaling changes.

[0157] For example, the locations of intersections, branches, and endpoints can be used as a set of reference points and expressed using relative coordinates and distance ratios, or the skeleton width can be standardized with respect to the length normalization axis to reduce the influence of deviations between objects. This encoding method is merely one example and can be modified into a hash-based representation, a topological signature vector, or a local patch descriptor depending on the embodiment.

[0158] The feature point extraction unit (150) may calculate quality indicators in parallel to quantify the reliability of the feature points. The quality indicators may include the consistency of intersection detection, the number of corresponding candidates, the stability of branch point detection, the stability of curvature calculation, skeleton connectivity, and a composite score with the matching quality indicator provided by the data matching unit (140). The threshold and weights may be adjusted according to the policy or operating environment, which corresponds to an exemplary configuration.

[0159] The feature point extraction unit (150) can package the calculated first feature point, second feature point, and third feature point into a feature vector set and provide it to the feature point registration unit (160). The packaging items may include a feature point type identifier, normalized coordinates, numerical values ​​such as length, angle, and curvature, a skeleton width profile, finger coordinate system metadata, and quality indicators.

[0160] The feature point extraction unit (150) may selectively provide a minimum configuration centered on the first feature point or an extended configuration including the second and third feature points, in accordance with the operation scenario (sequential or parallel authentication) of the user authentication unit (170). This method of provision is merely an exemplary form and may be adjusted to fit the database schema or transmission format according to the embodiment.

[0161] The feature point extraction unit (150) can perform diagnostic and fallback procedures to prevent error propagation. For example, if the number of branch points of the node pattern skeleton is below a threshold or the total length of the vein skeleton falls below a quality standard, a reprocessing request flag can be set to request reprocessing from the data processing unit (130) or the data matching unit (140). If an excessive or insufficient number of intersection candidates are detected, outliers can be suppressed by reweighting candidate scores or applying a minimum distance criterion between candidates. These diagnostic and fallback procedures correspond to an exemplary flow and can be adjusted according to the embodiment to suit system resources and target accuracy.

[0162] In conclusion, the feature point extraction unit (150) can consistently calculate the first feature point (intersection point of the finger vein pattern and finger joint pattern), the second feature point (structural feature of the finger joint pattern), and the third feature point (feature of the finger vein itself: topographical feature point and shape-based feature point) based on the first processing data and the second processing data in a normalized finger coordinate system.

[0163] The feature point extraction unit (150) structures the output result into a feature vector and provides it to the feature point registration unit (160) and the user authentication unit (170), thereby reliably supporting an operational form in which the first authentication step, the second authentication step, and the third authentication step are performed sequentially in subsequent steps, or two or more selected steps are performed simultaneously or sequentially.

[0164] Through this, the user authentication device (1) can secure high recognition accuracy and reliability even under various shooting conditions and environmental changes.

[0165] The feature point registration unit (160) can structure the first feature point, second feature point, and third feature point calculated by the feature point extraction unit (150) into user-specific registration data and store them in a server or database (200).

[0166] The feature point registration unit (160) can manage numerical values ​​such as coordinates, length, angle, curvature, and width, along with metadata (acquisition time, sensor identifier, first ROI and second ROI identifier, alignment quality indicator, transformation parameter or deformation field reference) by referring to the alignment and matching results provided by the data matching unit (140) and finger coordinate system information.

[0167] The registration data may be designed to include a type identifier (distinguishing between first, second, and third feature points), a quality indicator, and a policy key for weights, priorities, and combination rules (feature combination or score combination) to be utilized in a phased or parallel authentication scenario of the user authentication unit (170).

[0168] The feature point register (160) can ensure consistency and reproducibility by verifying input quality. For example, it can check minimum requirements such as alignment quality indicators, skeleton connectivity, branching point detection consistency, and intersection candidate reliability, and if the criteria are not met, it can set a reprocessing request flag to request reprocessing from the data processing unit (130) or the data matching unit (140). At this time, the threshold configuration and judgment rules can be adjusted to suit the service purpose and processing resources.

[0169] The feature point register (160) can generate and manage multi-sample-based templates for the same user. Generating multi-sample-based templates refers to a procedure for deriving a representative template by statistically combining first, second, and third feature points collected in multiple acquisition scenarios. The first feature point may calculate the cluster center as the representative location after coordinate alignment, and the second feature point may express the number of skeleton lines, total length, ratio of branch points to endpoints, skeleton width, and normalized coordinates as a weighted average or distribution range. The third feature point may package the frequency and relative position of endpoints, branch points, and intersection points, line length, branch angle, curvature, and distance between center lines into summary statistics and a location histogram. Operations such as outlier removal, quality-weighted combination, and multi-template storage (latest / optimal / renewed type) are merely examples and may be configured differently depending on the embodiment.

[0170] The feature point register (160) can configure user-specific indexes for quick lookup and comparison. The index may include a user identifier, registration date and time, finger coordinate system parameters, summary vectors of the first, second, and third feature points, quality indicators, and transformation parameter references, and may adopt a hash-based index, a graph signature index, or a multidimensional tree structure as needed. Frequently referenced templates and threshold / weight tables may be kept in a cache to reduce authentication delay.

[0171] The feature point register (160) may apply storage and transmission protection to ensure security and integrity. Feature vectors and metadata may be stored in an encrypted manner, integrity verification tags may be added to the transmission section, and the original image and the registration template may be stored separately. To reduce the risk of re-identification, a method of converting some features into irreversible mappings (e.g., hash, quantization, sketch) may be adopted, and access control, key management, audit logs, and retention / destruction policies may be operated. The specific mechanism is merely an example and may vary depending on regulations and policies.

[0172] The feature point register (160) can perform registration renewal and lifecycle management. When new acquired data is added, an appropriate strategy among cumulative reflection, replacement, supplementation, and rollback can be selected, and if the variation in the distribution of the second feature point exceeds a threshold within a certain period, the weight of the latest sample can be increased. If an abnormal structural deformation is detected at the third feature point, previous template recovery can be applied. The validity period, maximum storage count, and quality-capacity trade-off policy can be set to suit service requirements.

[0173] The feature point registration unit (160) may provide a subset of registered data in accordance with the request of the user authentication unit (170). For example, for high-speed authentication, it may support a stepwise provision method in which a minimum configuration focused on the first feature point is provided first, and the second and third feature points are subsequently provided as needed. In addition, type-specific weights, thresholds, and combination rules may be provided together as metadata to match sequential or parallel authentication scenarios. For user verification and error handling, diagnostic information such as the reason for failure, recommended measures, and registered version may be generated and notified to the user's terminal (300). The interface configuration is exemplary and may be adjusted to suit the actual service form.

[0174] In conclusion, the feature point register (160) can structure the first, second, and third feature points by user and safely store and manage them in a server or database (200), and can support the user authentication unit (170) to perform high-reliability comparison and authentication through quality verification, multiple sample combination, indexing, security protection, and update and lifecycle management.

[0175] Through this, the user authentication device (1) can improve overall authentication performance and security by ensuring consistency and reproducibility of registered data even under various environments and shooting conditions.

[0176] As shown in FIGS. 7 to 9, the user authentication unit (170) can identify and authenticate the user by comparing the registration template stored for each user in the feature point registration unit (160) with the first feature point, second feature point, and third feature point provided by the feature point extraction unit (150).

[0177] When an authentication request is received, the user authentication unit (170) may initiate a comparison procedure by referring to the alignment and matching results of the data matching unit (140), finger coordinate system information, the first, second, and third feature point vectors of the feature point extraction unit (150), and the user-specific registration template stored in the feature point registration unit (160). The comparison procedure may consist of a process for calculating step-by-step matching scores and a process for performing a final decision, and each step may have an independent threshold or an integrated threshold according to a setting policy.

[0178] The operation mode can be flexibly configured to perform the first authentication step, the second authentication step, and the third authentication step sequentially, and can also support a combination mode that performs at least two selected steps simultaneously or sequentially.

[0179] This flexibility in the execution sequence of steps can provide multi-layered benefits in real-world environments, such as reduced latency, decreased computational resources, increased robustness in high-risk situations, and minimized privacy exposure, and can precisely maintain a balance between false positive and false negative rates when combined with context-aware threshold and weight updates.

[0180] In this case, the first authentication step refers to a step of identifying and authenticating a user based on a first feature point (the intersection of the finger vein pattern and the finger joint pattern), the second authentication step refers to a step of identifying and authenticating a user based on a second feature point (structural features of the finger joint pattern), and the third authentication step refers to a step of identifying and authenticating a user based on a third feature point (topographical and shape-based features of the finger vein pattern).

[0181] The user authentication unit (170) can perform an intersection-based comparison in the first authentication step. The first feature point comparison can be evaluated based on distance or structure, including contextual information around the intersection points such as normalized coordinates of the intersection points, intersection angles, and distances to adjacent branches. For example, registered intersection configurations and input intersection configurations can be aligned and matched by applying Euclidean distance matching, iterative nearest point (ICP) matching, and graph matching. This is merely one example and, depending on the embodiment, can be replaced with a morphological or topological comparison utilizing Hausdorff distance, Shape Context, or topological signatures (Reeb graph, etc.). The user authentication unit (170) can record the score and quality indicators (number of corresponding intersection points, average error, omission rate, etc.) calculated in this step.

[0182] The user authentication unit (170) can compare structural indicators of the finger joint pattern in the second authentication step. The second feature points may include at least one of the number of joint pattern skeleton lines, the total length of the skeleton lines, the ratio of the number of branch points to the number of endpoints, the skeleton width (relative thickness based on the width profile), and the normalized position coordinates of the branch points or endpoints. The user authentication unit (170) can compare these indicators in vector form in a normalized finger coordinate system and calculate a score using the distance between vectors, weighted sum error, or statistical similarity (e.g., Mahalanobis distance). If necessary, feature subset selection, dimensionality reduction, or feature weight rebalancing may be performed, which corresponds to exemplary procedures.

[0183] The user authentication unit (170) can compare topographic and shape-based indicators of the vein pattern in the third authentication step. Topographic feature points may include at least one of the endpoints, branches, intersections, individual vessel skeleton line lengths, distances between endpoints and branches, branch angles, curvatures of the vein skeleton curves, and distances between the centerlines of individual vessel skeleton lines, and shape-based feature points may include at least one of the number of skeleton lines, total length of skeleton lines, ratio of branches to endpoints, skeleton width, and normalized coordinates of branches or endpoints. The user authentication unit (170) can generate a vector combining features of these two categories, compare it with a registration template, and calculate a structural similarity score. For example, graph-based comparison, phase invariant comparison, or histogram cross-measures of length, angle, and curvature distributions may be used, which are merely examples.

[0184] The user authentication unit (170) can apply a fusion strategy in addition to step-by-step comparison. Feature-level fusion can compare the first, second, and third feature point vectors by combining them early into a single vector, score-level fusion can calculate a final score by weighted summing and normalizing the step-by-step scores, and decision-level fusion can integrate the step-by-step authentication results based on voting or rules. Weights, thresholds, and combination rules can be adjusted according to policies, security levels, and environmental conditions. Additionally, learning-based recognition techniques (CNN / Siam network, ResNet-based embedding comparison, GNN, etc.) can be used in parallel as needed, and can be operated in combination with non-learning matching.

[0185] The user authentication unit (170) can calculate step-by-step and integrated quality indicators to quantify the reliability of the authentication. The quality indicators may include matching error, corresponding point coverage, intersection point matching rate, branch structure consistency, curvature stability, skeleton connectivity, and a composite score with the matching quality indicator provided by the data matching unit (140). Based on these indicators and scores, the user authentication unit (170) generates a final decision (authentication / rejection) and reliability value, and can trigger a fallback to a retry, additional step, or other step combination if necessary.

[0186] The user authentication unit (170) can flexibly configure the order and combination of step execution according to the operation policy. For example, in the basic policy, sequential execution in the order of the first authentication step → second authentication step → third authentication step may be adopted, and in the high-speed authentication policy, the first authentication step and the second authentication step may be performed simultaneously, and the third authentication step may be omitted when the result satisfies the threshold. In the security enhancement policy, all three steps may be performed in parallel, and authentication may be approved only when the detailed criteria are satisfied. This policy configuration is merely one example and may be set differently according to environmental conditions, processing resources, and security level depending on the embodiment.

[0187] In addition, the user authentication unit (170) can automatically correct thresholds to satisfy the false positive rate (FAR) and false rejection rate (FRR) targets, and manage a policy-specific threshold table near the equivalent error rate (EER) to link flexible operation with the achievement of quantitative targets.

[0188] The user authentication unit (170) can query a registration template stored in a server or database (200) and record the comparison results, and can perform receiving authentication requests, notifying progress status, and delivering final results in conjunction with the user's terminal (300). In addition, it can diagnose abnormal situations such as substandard input quality or unstable matching and perform exception processing such as providing guidance on re-acquisition, recommending lighting adjustment, or requesting finger position relocation.

[0189] In conclusion, the user authentication unit (170) can flexibly operate a stepwise or parallel authentication scenario based on the first feature point, the second feature point, and the third feature point to produce an authentication result of high accuracy and reliability through comparison with a registration template. Through this, the user authentication device (1) can secure consistent performance even under various shooting conditions and environmental changes, and achieve a security level that meets policy requirements.

[0190] The server or database (200) can safely store registration data provided from the feature point register (160) and perform lookup, operation, and transmission to support comparison and decision-making of the user authentication unit (170).

[0191] Here, the registered data refers to a set of data structured for each user, comprising the first feature point (intersection of the vein pattern and the finger joint pattern), the second feature point (structural feature of the finger joint pattern), and the third feature point (topographical and shape-based feature of the vein pattern) calculated by the feature point extraction unit (150), and the registered template refers to a representative feature vector combined and normalized based on multiple samples for the same user and its metadata. The server or database (200) can store these data in conjunction with a user identifier and provide indexing, search, version control, security protection, and auditing functions.

[0192] A server or database (200) can systematically store metadata such as finger coordinate system parameters, first, second, and third feature point vectors, quality indicators, alignment quality indicators, reference keys for alignment and alignment results (e.g., version, hash), transformation parameters, or deformation fields through a data schema. Additionally, normalization units of numerical values ​​such as coordinates, lengths, angles, curvatures, and widths, along with scale information, can be stored together to ensure reproducibility.

[0193] For example, the first feature point can be designed to include the normalized coordinates of the intersection, the intersection angle, and the distances to adjacent branch points and endpoints; the second feature point can be designed to include the number of skeleton lines, the total length, the ratio of branch points to endpoints, the skeleton width (width profile), and the normalized coordinates of branch points or endpoints; and the third feature point can package and store statistical and distribution information regarding the frequency and location distribution of endpoints, branch points, and intersections, segment length, endpoint-branch point distance, branch angle, curvature, and distance between centerlines. This schema configuration is merely an example, and the items and precision may vary depending on the implementation goals and system resources.

[0194] The server or database (200) may configure an index structure for fast authentication processing. The index may include a user-specific summary vector, feature point hash, graph signature, multidimensional tree, etc., and may support the user authentication unit (170) in quickly searching for close candidates between the input feature point and the registration template. If necessary, a cache layer may be provided to keep recently retrieved templates, thresholds, weights, and policy information resident in memory to reduce latency. This indexing and caching method is merely an exemplary operational form and may be replaced with other data structures depending on the scale of service and security level.

[0195] A server or database (200) can perform security protection during the storage and transmission phases. It can encrypt and store feature vectors and metadata, and apply integrity verification tags and secure transmission protocols to the communication segments. Additionally, it can store original images and registration templates separately and convert some features into irreversible mappings (hash, quantization, sketch, etc.) to reduce the risk of re-identification. Furthermore, it can satisfy regulatory and organizational policies by operating key management, periodic key rotation, HSM-based key storage, immutable preservation (append-only) of audit logs, access control (role- and scope-based), and data retention and destruction policies.

[0196] A server or database (200) can manage the lifecycle of registered data. When registering a new data, it can check the quality standards and trigger reacquisition, reprocessing, and re-alignment procedures for data that does not meet the standards. When updating, it can regenerate the registration template by applying the quality weight of the latest sample, or maintain a rollback or conservative template for indicators with high variability. It can also record versions and change history to support the user authentication unit (170) in specifying and comparing templates at different points in time.

[0197] A server or database (200) may optionally provide a subset of first feature points, a second and third feature point extension set, or an integrated feature vector via a query API to support phased or parallel authentication scenarios. In this case, the API may be configured to include authentication / authorization (role / scope), request rate limiting, masking of sensitive fields in the response, and detailed audit logging. Additionally, it may manage a table of phased weights and thresholds required for score combination or decision combination, and may update and deploy values ​​according to policy changes or environmental conditions (e.g., high-risk mode). The API may be implemented in REST or gRPC, but this is merely an example and may vary depending on the system structure.

[0198] The server or database (200) may operate replication, sharding, and snapshot backups to ensure availability and consistency. Data loss can be minimized through automatic switching and log-based recovery in the event of a failure, and proximity provisioning using read-only replicas or edge nodes can be performed for latency-sensitive authentication traffic. Additionally, recovery performance can be managed by defining and monitoring RPO and RTO goals as operational metrics. This distributed recovery configuration is an exemplary architecture and can be replaced depending on throughput, latency, and cost constraints.

[0199] The server or database (200) can monitor the operational status and accumulate diagnostic information. It can collect and analyze storage capacity, query delay, failure rate, quality indicator distribution, and trends in false positive and false rejection rates to automatically correct thresholds and policies or notify the administrator of warnings. If necessary, it can transmit the registration / authentication progress status or warning data to the user's terminal (300), and can support the transmission of related warning data when additional services, such as health abnormality signals, are in operation. The monitoring and warning procedures are merely exemplary flows and may vary depending on the actual service policy.

[0200] The server or database (200) may not store unnecessary original data to follow the principle of minimizing personal information, and may apply pseudonymization, aggregation, and retention period limits. In a multi-tenant environment, data isolation between users may be ensured by applying role-based access control and tenant boundaries, and if data transfer or export is required, it may be controlled through policy review and audit trails. Additionally, procedures for data residency, user consent, purpose restrictions, and processing of data subject rights (viewing, correction, deletion) may be reflected in the operational policy.

[0201] In conclusion, the server or database (200) can safely and efficiently store, manage, and provide user-specific registration templates with the first feature point, second feature point, and third feature point as cores, and can support high-reliability operation of the feature point register (160) and user authentication unit (170) through indexing, security, availability, auditing, and policy management.

[0202] Through this, the user authentication device (1) can maintain consistent authentication performance and security levels even in various environments and load conditions.

[0203] Referring to FIG. 10, in this embodiment, the user authentication device (1) can authenticate a user based on finger vein and finger joint patterns.

[0204] First, the user authentication device (1) can acquire an image or video (first data) of the finger vein inside the finger using an ultrasonic sensor.

[0205] That is, the user authentication device (1) can obtain first data that visualizes the vascular structure inside the finger by driving the ultrasonic sensor (100) and through the irradiation of the transmitting unit, the collection of reflected signals from the receiving unit, and the reconstruction procedure of the signal processing unit.

[0206] For example, the user authentication device (1) can set the array type and investigation method, gain and sensitivity, number of frames and sampling interval according to the situation, and can increase the contrast and boundary clarity of blood vessels by applying temporal multi-frame averaging and adaptive noise suppression in addition to beamforming, envelope detection, log compression, and automatic gain control. This processing pipeline is merely an example to aid understanding, and depending on the embodiment, the frequency band, focusing method, filter design, smoothing technique, etc., may be replaced or modified with other configurations that produce equivalent effects.

[0207] Through this, the user authentication device (1) can obtain first data of stable quality that is not affected by external factors such as lighting conditions, skin color, and skin condition, and can reliably use it as basic data for subsequent interest region extraction, segmentation, skeleton processing, and alignment steps.

[0208] Next, the user authentication device (1) can acquire an external shape image or video (second data) of the finger using a light sensor.

[0209] That is, the user authentication device (1) can drive the light sensor (110) to control the focus, exposure, and gain according to the scene conditions, sequentially apply lighting equalization, polarization reflection suppression, and optical distortion correction, and then acquire a series of frames and perform time averaging or motion correction to obtain second data that clearly expresses the finger contour and knuckle wrinkles.

[0210] For example, the user authentication device (1) can select a visible light or near-infrared wavelength, adjust the exposure time to preserve both highlight and shadow areas simultaneously, correct for distortions caused by differences in perspective and magnification, and apply polarization filtering to reduce surface reflections.

[0211] In addition, contour loss can be minimized by collecting frames over a specific time window and averaging only the stable segments, or by compositing after motion correction in cases of minute movement. These procedures and parameter configurations are merely examples to aid understanding, and depending on the embodiment, the spectrum band, exposure control method, distortion / perspective correction algorithm, frame compositing technique, etc., may be replaced or modified with other configurations that produce equivalent effects.

[0212] Through this, the user authentication device (1) can obtain second data of stable quality in which the effects of ambient lighting changes, partial reflections, and fine shaking are suppressed, and the second data can be used as input data to increase the accuracy of region of interest extraction and segmentation / skeleton processing in subsequent stages.

[0213] Next, the user authentication device (1) can extract an area corresponding to the first to 1.5 joints from the fingertip from the first data as the first ROI, and extract a first joint and an area adjacent thereto by a predetermined distance from it from the second data as the second ROI.

[0214] At this time, the first segment refers to the center position of the finger joint (DIP) closest to the fingernail, and the 1.5 segment may refer to the midpoint on the finger length axis between the first segment and the next joint (PIP). If PIP detection is difficult, the 1.5 segment may be defined as a position spaced apart from the first segment in the palm direction by a predetermined ratio (e.g., 50% of the DIP-PIP distance) or a predetermined distance according to the design. These ratio and distance settings are merely examples and may be determined differently depending on the embodiment.

[0215] That is, the user authentication device (1) sets the finger length direction as an internal reference axis, detects the fingernail boundary, finger contour, and the curvature inflection of the joint wrinkle in the second data to calculate the first joint position, and then accurately maps the calculated reference point to the first data coordinate system to define and extract the center, range, and rotation angle of the first ROI and the second ROI according to a consistent rule.

[0216] For example, the user authentication device (1) sets the length range of the first ROI along the finger length axis from the fingertip to the first joint to half of the first joint, calculates the width of the first ROI based on a ratio through finger width measurement, and can rotate and align the ROI boundary using an estimated tilt value when the finger is tilted with respect to the camera plane.

[0217] Additionally, in the second data, a second ROI is configured to include a band area defined by a predetermined distance above and below the center of the first node as a reference point, and the coordinates, size, rotation angle, and effective area mask of the two ROIs can be generated as metadata in a standard coordinate system that corrects for perspective and magnification differences. This procedure and parameter selection are merely examples to aid understanding, and depending on the embodiment, they may be replaced or modified with other configurations that produce equivalent effects, such as a learning-based node detector, edge-corner fusion detection, or calculation of an auxiliary reference point based on changes in ultrasonic density.

[0218] Through this, the user authentication device (1) can secure a standardized analysis area so that the first data and the second data acquired from different sensors refer to the same anatomical region, and subsequently, by maintaining positional consistency in the segmentation, skeleton processing, alignment, and feature point extraction stages, it can reduce alignment errors and increase the reliability of feature points.

[0219] Afterwards, the user authentication device (1) can generate first processed data and second processed data by performing segmentation and skeleton processing on each of the first ROI and the second ROI.

[0220] That is, the user authentication device (1) can generate first processed data and second processed data including metadata based on a normalized finger coordinate system by segmenting the designated vein blood vessel area in the first ROI and deriving a centerline skeleton and width profile through phase-preserving thinning, and segmenting the finger joint pattern in the second ROI and extracting a joint skeleton, endpoint / branching point candidates, and connectivity information through thinning of the same principle.

[0221] For example, it can be implemented with a configuration that applies threshold-based segmentation or learning-based segmentation, performs refinement procedures such as small hole filling, disturbance component removal, and boundary smoothing, and then generates a skeleton using a representative thinning algorithm or distance transformation-based median extraction. This selection of algorithms and parameter configuration is merely an example to aid understanding, and depending on the embodiment, it may be replaced or modified with techniques that produce equivalent effects, such as graph cut, active contour, U-Net-type segmentation, skeleton correction filter, and multi-resolution thinning.

[0222] Through this, the user authentication device (1) can obtain a skeleton-based representation in which noise and background components are removed and the central structure is preserved, and subsequently provide an input that can reliably compare indicators such as coordinates, length, angle, and width in the alignment and matching steps.

[0223] Furthermore, the user authentication device (1) can align and match the positions of the first processed data and the second processed data.

[0224] That is, the user authentication device (1) can match the spatial relationship of the structure represented by the first processing data and the second processing data by performing an initial alignment using a similar transformation including translation, rotation, and scaling using ROI metadata (center, size, rotation angle) to place the two skeletons on the same standard based on a finger coordinate system defined by the finger length axis and the first joint position, and then performing a fine alignment that minimizes residuals by applying iterative nearest point (ICP) or Procrustean analysis.

[0225] At this time, the user authentication device (1) can mitigate local inconsistencies while maintaining the phase of the skeleton by combining disturbance suppression rules such as weighted least squares, trimming, and RANSAC-based candidate rejection to reduce the influence of disturbance points, and by selectively applying thin plate spline (TPS) or B-spline freeform deformation under normalization constraints to correct local non-rigid deformation caused by finger contact pressure or micro-posture changes.

[0226] For example, the user authentication device (1) selects a set of branch points and endpoints of the second processed data as reference points to form a nearest correspondence with the corresponding candidate of the first processed data, updates transformation parameters until a convergence threshold is reached using the sum of squares of the distances between corresponding points as the objective function, and after convergence, can calculate the Hausdorff distance, mean squared error, branch point matching rate, and endpoint matching rate as quality indicators. If the matching quality falls short of the standard, a fallback procedure is performed to replace the initial alignment (e.g., initial estimation based on Shape Context instead of principal axis alignment) or to retry with an alternative path using graph matching, and finally, a transformation matrix or transformation field, a matching quality indicator, and an intersecting area mask are generated and provided to subsequent steps. These procedures and parameter selections are merely examples to aid understanding and may be replaced or modified with other alignment, matching, and correction algorithms that provide equivalent effects depending on the embodiment.

[0227] Through this, the user authentication device (1) can significantly reduce the alignment error caused by differences in shooting angle, finger position, and contact pressure, and establish a basis for stably calculating and comparing the first feature point (intersection point), the second feature point (node ​​structure index), and the third feature point (finger vein structure index) in a common coordinate system.

[0228] Furthermore, the user authentication device (1) can extract multiple feature points based on the first processed data and the second processed data.

[0229] At this time, multiple feature points may include the intersection of the finger vein pattern and the finger joint pattern, and the relative position information of the finger vein branching point detected around the intersection.

[0230] That is, the user authentication device (1) detects the intersection point of the finger vein skeleton and the finger joint skeleton as a first feature point, calculates the normalized coordinates, the intersection angle, and the tangent direction of both skeletons, and can calculate and include contextual information such as the distance and direction to an adjacent branch point or an adjacent end point centered on the intersection point.

[0231] Additionally, the user authentication device (1) can calculate at least one of the number of skeleton lines of the node pattern, the total length of the skeleton lines, the ratio of the number of branch points to the number of endpoints, the skeleton width, and the normalized coordinates of the branch points or endpoints as a second feature point, and can calculate at least one of the endpoints, branch points, intersection points, segment length, distance between endpoints and branch points, branch angle, curvature, distance between center lines, the number of skeleton lines, the total length of the skeleton lines, and the skeleton width as a third feature point.

[0232] For example, the user authentication device (1) can define a local coordinate system around the intersection point to express the relative position to the branch point as a distance-angle vector, and can normalize continuous values ​​such as length, angle, curvature, and width according to the finger coordinate system reference length and reference angle to encode them so as to be invariant to rotation, translation, and scaling changes, and for indicators with a skewed distribution, can increase robustness by performing statistical summaries such as quantiles, histogram crossovers, and weighted averages in parallel. This method of selecting indicators and calculating and normalizing them is merely an example and can be replaced or modified with other measurement, summarization, and encoding procedures that provide equivalent effects depending on the embodiment.

[0233] Through this, the user authentication device (1) can secure cross-based, node-based, and vein-based complementary information in a balanced manner, and subsequently provide a feature vector with high discriminability and reproducibility in the registration and authentication stages.

[0234] Furthermore, the user authentication device (1) can store the extracted multiple feature points in a server or database and register them for each user.

[0235] That is, the user authentication device (1) can package a plurality of feature point vectors composed of a first feature point, a second feature point, and a third feature point together with finger coordinate system parameters, transformation parameters used for alignment, alignment quality indicators, skeleton connectivity indicators, acquisition time, and sensor identifiers, and structure them into registration data mapped to a user identifier. In this process, the consistency and reproducibility of the registration data can be ensured by performing a verification procedure that checks minimum quality standards and requests reprocessing or re-alignment if the standards are not met.

[0236] For example, the user authentication device (1) can represent first feature points, such as intersection coordinates and intersection angles, as cluster centers and variances after coordinate alignment, and second feature points, consisting of the number of skeleton lines of the node pattern, total length, ratio of the number of branch points and the number of endpoints, skeleton width, and normalized coordinates of the branch points or endpoints, by summarizing and storing them as weighted averages and range values.

[0237] Additionally, the user authentication device (1) can express variability by storing a histogram, quantiles, mean, and standard deviation together with a third feature point composed of endpoints, branching points, and intersection points of the vein pattern, segment length, distance between endpoints and branching points, branching angle, curvature, distance between centerlines, number of skeleton lines, total length of skeleton lines, and skeleton width.

[0238] The user authentication device (1) can form a representative registration template by statistically combining data obtained multiple times for the same user with quality weights, construct an index of user-specific summary vectors and key fields for quick lookup, and ensure security by applying encryption and integrity verification during the storage and transmission stages. The specific algorithms and parameters of these representative, indexing, and protection procedures can be replaced or adjusted in a way that produces an equivalent effect according to the implementation environment and policy.

[0239] Through this, the user authentication device (1) can maintain a user-specific registration database that can be referenced quickly and consistently in a subsequent comparison step, and can improve the stability, accuracy, and security of the entire authentication process by providing a representative registration template that averages the noise and differences in shooting conditions between samples.

[0240] Finally, the user authentication device (1) can authenticate the user based on the registered feature points.

[0241] That is, the user authentication device (1) calculates a step-by-step matching score by comparing the user-specific registration template stored in the feature point registration unit (160) with the first feature point (intersection of the finger vein pattern and finger joint pattern), the second feature point (structural feature of the finger joint pattern), and the third feature point (topographical and shape-based feature of the finger vein pattern) provided by the feature point extraction unit (150), and can proceed with decision-making through sequential execution or parallel execution according to the operation policy.

[0242] At this time, the user authentication device (1) can apply an independent threshold or an integrated threshold to each of the intersection-based comparison, node-based comparison, and vein-based comparison, and evaluate quality indicators (matching error, corresponding point coverage, structural consistency, etc.) together to calculate the reliability of the result.

[0243] Additionally, the user authentication device (1) can calculate a final score by fusing the step scores with one or more of feature combinations (feature combinations), score combinations, and decision combinations, and can trigger a fallback to a retry, additional step, or alternative algorithm path if necessary.

[0244] For example, the user authentication device (1) can calculate a score S1 by performing distance-based matching and graph matching in parallel using normalized coordinates, intersection angles, and distance information to adjacent branch points for the first feature point, and for the second feature point, can calculate a score S2 by comparing a vector composed of the number of skeleton lines, total length, ratio of the number of branch points to the number of endpoints, skeleton width, and normalized coordinates of the branch points or endpoints to calculate a vector distance or statistical similarity. For the third feature point, a score S3 can be obtained by evaluating endpoints, branch points, intersection points, segment length, distance between endpoints and branch points, branch angle, curvature, distance between center lines, number of skeleton lines, total length of skeleton lines, skeleton width, etc., as histogram intersection or graph structure similarity.

[0245] The user authentication device (1) can calculate a final score by combining S1, S2, and S3 with weights that are pre-set or adaptively updated, and can determine authentication or rejection based on a threshold that satisfies the target false positive rate and false rejection rate.

[0246] If the input quality is low or the reliability of a specific stage is insufficient, it can be supplemented through procedures such as reflecting additional frames, performing combinations of different stages in parallel, or applying conservative thresholds. This comparison, fusion, and judgment flow is merely an example and may be replaced or modified according to the embodiments with other similarity measurement, weight calculation, and threshold operation methods that provide equivalent effects.

[0247] Through this, the user authentication device (1) can provide authentication results of high accuracy and reliability even under various shooting conditions and environmental changes by utilizing cross-based, node-based, and vein-based complementary information, and can stably achieve a security level that meets policy requirements.

[0248] In another embodiment, the user authentication device (1) can calculate a first authentication reliability based on the user's behavioral habit data.

[0249] Specifically, the user authentication device (1) can calculate the first authentication reliability by quantifying behavioral indicators derived from the optical sensor (110) and evaluating the similarity with the user's usual behavioral pattern.

[0250] Here, the first authentication reliability refers to a reliability value (score or probability) calculated based on behavioral habit data, and the behavioral habit data refers to a time-series indicator including the finger movement trajectory in the shooting area, the entry slope of the finger relative to the reference plane, and the positional shake of the fingertip before and after the point of entry into the shooting area.

[0251] First, the user authentication device (1) can obtain first habit data by analyzing data obtained from the light sensor (110) to extract the trajectory of the user's finger moving within the shooting area.

[0252] That is, the user authentication device (1) can obtain first habit data by tracking the coordinates of the fingertip or a set reference point in consecutive frames with subpixel precision and calculating the movement vector, cumulative path length, average and maximum speed, rate of change of acceleration, and local curvature distribution based on the time-ordered coordinate sequence.

[0253] For example, the user authentication device (1) can stabilize the path by setting a region of interest within the shooting area and performing Kalman filter tracking, light flow-based estimation, and feature point tracking in parallel, and suppress noise through smoothing, outlier removal, and frame drop correction, but this is merely an example and in actual implementation, it can be replaced or modified with equivalent means such as deep learning-based keypoint detection, contour centerline tracking, and template matching to match the sensor specifications, processing resources, and target delay.

[0254] Through this, the user authentication device (1) can obtain first habit data that quantifies the access path and finger movement habits that are repeatedly reproduced for each user, and can then quantitatively reflect behavioral consistency in the reliability calculation stage.

[0255] Next, the user authentication device (1) can obtain second habit data by analyzing data obtained from the light sensor (110) and calculating the entry slope of the finger relative to the reference plane based on the reference plane of the shooting area.

[0256] At this time, the reference plane refers to a reference plane that defines the shooting area of ​​the optical sensor (110). For example, in a non-contact configuration, it may be defined as a virtual plane perpendicular to the optical axis and set at a calibrated reference distance; in a contact configuration, it may be defined as the surface plane of a transparent support on which a user places their finger; and in a system using a calibration target, it may be defined as a virtual reference plane estimated to be a homography of the shooting device and a calibration board spatially fixed thereto. The specific substance and calculation method of the reference plane may vary depending on the device structure and calibration procedure.

[0257] That is, the user authentication device (1) can obtain second habit data by setting a reference plane defining the shooting area of ​​the optical sensor (110) as a reference plane, estimating the angle formed by the average direction of the finger length axis or the finger surface tangent with this reference plane on a frame-by-frame basis, and calculating the average value, variance, and rate of change for each interval.

[0258] For example, the user authentication device (1) can approximate the fingernail boundary as an ellipse and inversely calculate the gradient component according to perspective from the ratio of the major and minor axes of the ellipse, and can estimate the normal direction at a single viewpoint by combining the degree of foreshortening in the axial direction of the finger contour and the dominant direction of the skin texture isointensity line, and can directly calculate the normal vector through image matching if multiple lighting or multiple viewpoint configurations are allowed. These procedures are merely examples for illustrative purposes and can be replaced or supplemented by equivalent means, such as depth approximation models, learning-based angle regression, and single-viewpoint geometry estimation using correction patterns, in actual implementation.

[0259] Through this, the user authentication device (1) can obtain second habit data that quantifies the finger entry habit that appears repeatedly for each user into an angle indicator, and can subsequently be used as a basis for evaluating the consistency and stability of the input posture in the reliability calculation stage.

[0260] Next, the user authentication device (1) can obtain third habit data by analyzing data obtained from the light sensor (110) and detecting the positional shaking of the fingertip in consecutive frame images before and after the point of entry into the shooting area according to a preset threshold.

[0261] That is, the user authentication device (1) can obtain third habit data by defining the point in time when the fingertip first enters the boundary of the shooting area as a reference point, setting a certain time interval centered on this point, and then applying a normalized finger coordinate system and frame rate correction to calculate the inter-frame displacement of the fingertip coordinates in each frame within the interval, and calculating indicators such as amplitude, frequency components, and root mean square (RMS) shake from the displacement.

[0262] In this case, the length of the time interval and the center placement can be set differently depending on the system policy, sensor frame rate, and user scenario.

[0263] For example, the user authentication device (1) can remove low-frequency and high-frequency components, separate the fine vibration band with a bandpass filter, obtain the energy distribution by frequency using a Fourier transform or wavelet transform, and in the time domain, combine the displacement standard deviation and peak-to-peak value, and suppress noise by applying Kalman smoothing or outlier removal rules.

[0264] In addition, the level of shaking can be classified into stable, normal, and unstable grades by applying a reference threshold defined during the system initialization process or by calculating a user-specific adaptive threshold from the past measurement distribution for the same user; however, this is merely an example, and in actual implementation, it can be replaced or supplemented by equivalent methods such as time-frequency combined analysis, dynamic window size adjustment, and threshold calculation based on statistical hypothesis testing.

[0265] Through this, the user authentication device (1) can obtain third habit data that quantifies the stability and micro-tremor characteristics appearing at the moment of finger approach in a noise-robust manner, and subsequently, the user's behavioral consistency can be evaluated more precisely in the reliability calculation stage.

[0266] Next, the user authentication device (1) can register or update the user's standard behavioral habit data by storing the first habit data, the second habit data, and the third habit data in a server or database (200).

[0267] That is, the user authentication device (1) can register or update the user's standard behavioral habit data by normalizing the first habit data, second habit data, and third habit data collected from multiple sessions into the same coordinate system and time axis, assigning weights that reflect the quality and recency of each session, and then packaging them into standard behavioral habit data that includes statistical summaries and metadata such as mean, variance, covariance, quantile boundaries, and interval frequency distribution, and storing them in a server or database (200).

[0268] At this time, the user authentication device (1) can store the time axis trend and rate of change together for drift detection and apply a rule for excluding outliers to manage so that temporary noise does not excessively affect the reference value.

[0269] For example, the user authentication device (1) may apply an exponentially weighted moving average to give greater weight to sessions with higher reliability based on the session quality score, and to increase the influence of more recent sessions. Additionally, the influence of extreme values ​​may be suppressed by robust estimation using the median and absolute deviation, and the reference profile may be separated into a recent profile and a conservative profile, allowing one to be operated in parallel for rapid adaptation and the other for stable judgment. Version management and change history may be recorded to re-reference reference data at a specific point in time or to restore to a previous version if necessary. These configurations and procedures are examples to aid understanding, and in actual implementation, they may be replaced or supplemented with equivalent methods tailored to storage capacity, processing delays, and service policies.

[0270] Through this, the user authentication device (1) can secure a highly reliable standard distribution to be used in the subsequent comparison step, and can stably reflect the user's behavioral habits without excessive sensitivity or insensitivity even if they gradually change, thereby increasing the consistency and accuracy of the overall authentication reliability calculation.

[0271] Finally, the user authentication device (1) can calculate the first authentication reliability according to a preset standard by comparing the user's behavioral habit data obtained during authentication based on feature points with the user's standard behavioral habit data.

[0272] That is, the user authentication device (1) can calculate the first authentication reliability by aligning and normalizing the first habit data (trajectory), second habit data (entry slope), and third habit data (position shake) to the same time axis and coordinate system, and then combining the trajectory similarity, angle deviation, and the degree of agreement of the shake statistics at the level of an indicator or score.

[0273] For example, the user authentication device (1) can calculate dynamic time distortion (DTW) distance or Frege-Hausdorff distance for the trajectory, calculate angle deviation indicators such as the average, maximum value, and coefficient of variation of frame-by-frame residuals for the entry slope, and for shake, calculate the Mahalanobis distance from the reference distribution or the p-value of the hypothesis test using statistics such as RMS displacement, frequency band energy, and variance. Then, a weighted sum score or a logistic combined score is calculated by applying weights proportional to the reliability of the three indicators (acquisition quality, illumination stability, alignment quality, etc.), and can be determined as a threshold in accordance with the operation policy. If the environment is detected as high risk, an adaptive rule can be applied to conservatively raise the threshold or temporarily increase the weight of a specific indicator (e.g., shake stability). This combination and judgment method is merely one example and can be replaced or supplemented by equivalent methods such as Bayesian combination, soft voting, and rank fusion.

[0274] Through this, the user authentication device (1) can improve the accuracy and reliability of the entire authentication by adding verification of the consistency of the user's access behavior to the primary authentication result based on biometric features, thereby increasing robustness against spoofing and proxy authentication attempts, and by providing an auxiliary reliability indicator that does not waver even with environmental changes such as lighting, posture, and fine shaking.

[0275] In another embodiment, the user authentication device (1) can calculate a second authentication reliability based on the user's gesture data.

[0276] Here, the second authentication reliability refers to a quantitative indicator derived by aligning and comparing gesture data and reference gesture data according to the same standard, and can be calculated by weighted combination of individual indicators such as trajectory similarity, the degree of agreement of time-normalized velocity, acceleration, and curvature profiles, and the degree of agreement between contact events and direction change patterns.

[0277] The second authentication reliability can be expressed as a continuous score of 0 to 1 or 0 to 100, or as a grade of pass / pending / fail, according to preset criteria, and the threshold can be set in a fixed or adaptive manner to match the security level and environmental conditions. Additionally, it can be operated to be calculated only in a registration process environment, an environment classified as high-risk, or an environment where an error is detected during the user authentication process. These definitions and methods of expression are merely examples to aid understanding and may be replaced or supplemented with other indicator configurations or expression systems that produce equivalent effects depending on the embodiment.

[0278] First, the user authentication device (1) can obtain gesture data by analyzing data obtained from the light sensor (110) and detecting the trajectory of a gesture performed by a finger within the shooting area.

[0279] At this time, gesture data refers to a data representation that includes spatiotemporal descriptors such as a time-ordered coordinate sequence of a fingertip or a pre-specified reference point in consecutive frames, and trajectory length, velocity, acceleration, angular velocity, curvature, rate of change of curvature, number of direction changes, stop section marker, entry and exit times, whether the curve is closed, and trajectory area calculated from said coordinate sequence, and may include coordinate normalization information and quality indicators as needed.

[0280] That is, the user authentication device (1) can obtain gesture data by tracking the fingertip position on a frame-by-frame basis, performing smoothing and outlier removal for noise reduction, dividing the trajectory into entry-motion-exit sections, and normalizing it into a standardized coordinate system.

[0281] For example, the user authentication device (1) can stabilize the trajectory by applying light flow-based tracking or a Kalman filter, and if necessary, combine time-series smoothing with the deep learning-based keypoint detection result to increase tracking reliability, but such implementation is merely an example and can be replaced or supplemented with techniques of equivalent effect, such as shape-centered tracking, feature point matching, and pre-trained pose estimation, depending on the embodiment.

[0282] Through this, the user authentication device (1) can obtain reliable gesture data that can be used to identify finger contact motions, finger tilting motions, circular trajectory motions or zigzag trajectory motions and to compare them with reference gesture data.

[0283] Next, the user authentication device (1) can identify at least one of a finger contact motion, a finger tilt transition motion, a circular trajectory motion, or a zigzag trajectory motion from the gesture data.

[0284] That is, the user authentication device (1) can define and determine each movement by designating a finger reference point (e.g., a pole of the fingertip or the center of a contact patch) projected onto a reference plane of the shooting area according to a consistent rule, and normalizing and analyzing the trajectory of this reference point into a finger coordinate system.

[0285] Finger contact motion is defined as an event in which actual contact occurs between a finger and a reference plane, and can be identified by detecting the point of contact by combining a sudden increase in brightness, flattening of the contour, and a sudden decrease in the estimated distance to the reference plane in a sequence of frames.

[0286] Finger tilt transition behavior is defined as an event in which the sign of the entry angle changes or its magnitude changes rapidly beyond a set threshold over time, and can be identified by detecting the transition interval by applying the rate of change of the angle time series and hysteresis rules.

[0287] Circular trajectory motion is defined as a movement in which the reference point trajectory forms a closed curve in the reference plane, the sign of curvature is constant, an estimated center exists, and the rate of change of radius remains within a limited range, and can be identified by verifying the degree of trajectory closure and radius stability.

[0288] Zigzag trajectory motion is defined as a movement in which the direction vector of the reference point trajectory is alternately reversed in each section, and endpoints or corners defined as turning points appear repeatedly at a density greater than a certain level, and can be identified by analyzing the frequency of direction changes and the distribution of straight section fragment lengths.

[0289] For example, the user authentication device (1) may apply the following intuitive judgment rules. In the case of a finger contact motion, contact can be determined when the average brightness of pixels near the reference plane increases by more than a certain percentage compared to the previous frame (e.g., more than 15%), the cross-sectional aspect ratio of the finger contour decreases rapidly (e.g., drops to 0.8 or lower), and the estimated distance to the reference plane falls below a preset threshold at the same time.

[0290] In addition, for finger tilt transition motions, after smoothing the frame-by-frame entry angle time series, a transition can be determined when the amount of angle change exceeds a threshold (e.g., 8° or more) in a short interval or when the point where the sign of the angle changes exceeds the hysteresis range.

[0291] In the case of circular trajectory motion, if least squares circle fitting is applied to the reference point trajectory and the mean square error of the radius residual is below a threshold (e.g., 5% or less of the radius) and the variance of the angular velocity per frame is low (e.g., below the variance threshold) and maintained for a certain period of time, it can be determined to be circular.

[0292] In the case of a zigzag trajectory, if the number of intervals in which the trajectory's direction vector alternately reverses is greater than a threshold (e.g., 3 times or more), the density of turning points determined as corners is greater than a certain level, and the average length of straight segments is maintained above a lower limit, it can be determined as a zigzag.

[0293] These figures and procedures are merely examples to aid understanding, and in actual implementation, the threshold, smoothing method, and feature combination may be adjusted while maintaining the same intent, or they may be replaced or supplemented with equivalent classification techniques such as hidden Markov models, conditional random fields, and CNN-RNN-based time series classifiers.

[0294] Through this, the user authentication device (1) can distinguish gesture types based on the actual movement of the finger reference point with clear and consistent standards, and can increase the accuracy of the second authentication reliability calculation by reducing the possibility of misjudgment in comparison with the reference gesture data.

[0295] Next, the user authentication device (1) can calculate a second authentication reliability by comparing the identified gesture data with the predefined standard gesture data of the corresponding user.

[0296] At this time, the reference gesture data refers to a reference template organized by type by normalizing and summarizing gesture data accumulated during the user registration or update process, and specifically, it may refer to a data set consisting of a representative trajectory for each gesture type, time-normalized velocity and acceleration profiles, curvature change profiles, statistics on the contact point and contact duration, distribution of direction change frequency and intervals, and distribution of the total duration of the gesture.

[0297] That is, the user authentication device (1) can first match the type of gesture to be compared, perform coordinate system alignment and time axis normalization to match the two gesture data so that they are compared on the same standard, and then calculate the difference in trajectory shape, the difference in speed and curvature per section, and the difference in contact and direction change patterns as individual scores and combine them to calculate the second authentication reliability.

[0298] At this time, the alignment can be designed to include spatial alignment, which applies translation, rotation, and scaling corrections after unifying the starting point and direction of travel, and temporal alignment, which corrects for speed differences between sections.

[0299] For example, the user authentication device (1) can calculate the average distance and maximum deviation after aligning two gesture trajectories to overlap, calculate the difference between the speed profile and curvature profile for each section as a summary indicator, and calculate the final score by reflecting the difference between the contact point deviation and the contact maintenance time, and the difference between the number of direction changes and the interval.

[0300] Additionally, dynamic time distortion may be used for time alignment, Frege-Hausdorff-based distance calculation for shape comparison, and statistical agreement or learning-based similarity scores for feature combination. These procedures and metric selections are merely examples to aid understanding and may be replaced or supplemented with other time alignment techniques, shape comparison techniques, or score combination rules that produce equivalent effects, depending on the embodiment.

[0301] Through this, the user authentication device (1) can secure a second authentication reliability that quantifies the consistency of the user's gesture habits, and can improve the reliability and resilience of the entire authentication by supplementing the biometric feature-based authentication result even if there are environmental changes or temporary motion fluctuations.

[0302] At this time, the user authentication device (1) can control the calculation of the second authentication reliability to be performed only in a specific environment.

[0303] Specifically, the user authentication device (1) can be configured to perform the calculation of the second authentication reliability only when at least one of the following is detected: a first environment corresponding to the user authentication information registration procedure, a second environment identified as high risk according to a preset standard, and a third environment in which an error occurred during the process of authenticating the user based on registered feature points.

[0304] Here, the first environment refers to a state where the user authentication information registration process is in progress, the second environment refers to a state determined to be high-risk according to pre-set criteria, and the third environment refers to a state where an error occurred during the process of authenticating the user based on registered feature points.

[0305] The user authentication device (1) can refer to a registration mode flag, a registration session token, and registration procedure step information to detect a first environment, and to detect a second environment, it can combine and evaluate risk signals such as multiple failure attempts in a certain time window, abnormal terminal geographic information or time zone, pattern repetition classified as a suspected tampering signal or unrealistic matching residual, rapid deterioration of image quality indicators, and liveness check inconsistency, and to detect a third environment, it can use error signals such as failure to meet the threshold of the step-by-step matching score, instability of the alignment quality indicator, failure of the matching algorithm to converge, and breakdown of consistency between quality indicators as judgment criteria.

[0306] For example, the user authentication device (1) may define a second environment as a case in which three or more mis-authentication attempts occur out of the last five times for high-risk determination, a state in which the terminal location is excessively different from the registered area, a state in which the average error of the intersection-based matching exceeds the policy standard, or a state in which the texture entropy in the same frame sequence is abnormally low is observed simultaneously, and may define a third environment as a case in which step-by-step matching repeatedly fails to reach the termination condition or frames with low quality weights are consecutively dominant.

[0307] These criteria and thresholds are merely examples to aid understanding, and in actual implementation, detection rules, observation window lengths, weights, and thresholds can be adjusted or adaptively updated to align with the organization's security policies and operational metrics.

[0308] Through this, the user authentication device (1) can reduce delay and power consumption by omitting unnecessary gesture analysis in normal environments, and can effectively increase overall authentication reliability by selectively performing gesture-based reinforcement verification in security-sensitive environments or situations where signs of error are observed.

[0309] Furthermore, the user authentication device (1) may combine the calculated second authentication reliability with the existing authentication result to reflect it in the final judgment, and record the comparison process and result in a server or database (200).

[0310] That is, the user authentication device (1) can maintain the second authentication reliability as an independent auxiliary score or update the integrated score according to the score combination and decision combination rule, and then decide one of authentication approval, additional step request, or rejection according to the policy-specific threshold.

[0311] At this time, score combination can be implemented using weighted summation, logistic combination after normalization, Bayesian update, etc., and decision combination can integrate step-by-step judgments using majority voting, weighted voting, or sequential gating (evaluating the next condition if the priority condition is not met).

[0312] In addition, the user authentication device (1) may have a hysteresis interval and a retry cooldown time for the stability of the judgment, and may periodically adjust the threshold to match the target false positive rate (FAR), false rejection rate (FRR), or equivalent error rate (EER).

[0313] For example, the user authentication device (1) can be configured to increase the weight of the second authentication reliability in high-risk mode, and to reflect the second authentication reliability only when the result of comparing basic biometric features falls below the threshold in normal mode. Additionally, if the variability of recent sessions for the same user increases, the influence of gesture-based reliability can be temporarily expanded, and in situations where the terminal class is low or network reliability is degraded, the decision combination rule can be switched to “strict mode” to make a conservative judgment. Such policy configuration and weight / threshold settings are merely examples, and in actual operation, they can be adjusted in various ways to suit environmental conditions, user group characteristics, terminal security class, and service delay tolerance.

[0314] Through this, the user authentication device (1) can selectively combine gesture-based behavior verification with a biometric feature-based primary identification result, thereby reducing unnecessary calculations in normal environments to suppress delay and power consumption, while effectively increasing the reliability of judgment in security-sensitive situations or inputs with high uncertainty.

[0315] In addition, the user authentication device (1) can record the score, threshold version, policy key, environment flag, final judgment, and basis indicator (e.g., step score, quality indicator, number of retries) used in the combination process in an auditable format on a server or database (200), and can simultaneously ensure data security and traceability by applying integrity verification and encryption to the transmission and storage sections.

[0316] In another embodiment, the user authentication device (1) can detect an abnormal blood vessel pattern and provide a health abnormality warning.

[0317] Specifically, the user authentication device (1) can further analyze the first data (image or video of a finger vein inside the finger) to detect a vascular stenosis pattern based on a change in the cross-sectional area or blood flow velocity of the blood vessel and an abnormal blood vessel branching pattern based on a change in the branching angle or branching interval of the blood vessel, generate a user health abnormality signal based on the abnormal blood vessel pattern data that integrates these results, and then transmit warning data including the generated health abnormality signal to the user's terminal (300).

[0318] First, the user authentication device (1) can analyze the first data collected using the ultrasonic sensor (100) and detect a blood vessel stenosis pattern according to a first standard set based on the cross-sectional area of ​​the blood vessel or the change in blood flow velocity.

[0319] At this time, the first criterion, which is pre-set, refers to a set of measurement rules that are pre-defined to determine a section suspected of stenosis, and may consist of (i) a threshold for the cross-sectional area reduction rate, (ii) a threshold for the blood flow velocity increase rate, (iii) a threshold for a combined indicator combining cross-sectional area and velocity, (iv) a persistence condition regarding whether the abnormal condition persists for a certain length or time, and (v) a quality gate regarding whether the image quality satisfies minimum requirements.

[0320] The user authentication device (1) can calculate a cross-sectional profile along the centerline (vascular axis) of the finger coordinate system and, in a configuration where Doppler information or inter-frame speckle tracking is possible, estimate the local blood flow velocity at the same location, determine the cross-sectional area reduction rate and velocity increase rate after applying the quality gate first, and then check the comprehensive indicator and the persistence and reproducibility conditions to indicate the section suspected of stenosis.

[0321] In cases where Doppler is not provided, judgment is performed using only cross-sectional area-based indicators, while the velocity item is treated as missing and the weight of the composite indicator is automatically adjusted to achieve an equivalent effect. For sections where quality indicators fall below a standard, such as signal-to-noise ratio degradation, unclear boundaries, or skeleton disconnection, a reacquisition request flag is assigned instead of interpolation to suppress false alarms.

[0322] For example, the user authentication device (1) samples the cross-sectional area at standard intervals along the centerline, and if the cross-sectional area of ​​a specific section decreases by more than 30% compared to the reference value based on the moving average of adjacent sections, it can be marked as a primary candidate. If the estimated blood flow velocity at the same location increases by more than 40% compared to the user reference value, it is determined that the secondary condition is satisfied, and if the comprehensive indicator, such as the velocity ratio relative to the cross-sectional area, exceeds the top 10% of the reference distribution for each user, it can be confirmed as a final candidate for stenosis.

[0323] In addition, the user authentication device (1) can grant reliability by checking whether the continuous length of the candidate section is 2 mm or more, and whether it is reproduced at the same coordinates in a re-photograph of the same finger. These figures are merely examples to aid understanding, and in the actual system, the sensor resolution, finger size, sample interval, threshold, and persistence / reproducibility criteria can be adaptively adjusted according to the service policy and data characteristics.

[0324] Through this, the user authentication device (1) can detect abnormal narrowing of the lumen of the blood vessel and the accompanying change in flow rate at an early stage, and by combining quality gates and conditions of persistence and reproducibility, it can reliably detect the stenosis pattern while reducing false alarms caused by temporary noise or changes in posture.

[0325] Next, the user authentication device (1) can analyze the first data and detect an abnormal blood vessel branching pattern according to a second criterion set based on a change in the blood vessel branching angle or branching interval.

[0326] At this time, the pre-set second criterion refers to a set of rules defined in advance to determine whether there is an abnormality in the branching structure, and may consist of (i) a branching angle threshold condition (an excessively narrow or wide angle compared to individual reference values ​​or group statistics), (ii) a branching interval threshold condition (when the distance between adjacent branching points along the same blood vessel axis becomes abnormally short or excessively long), (iii) a condition for the persistence and reproducibility of abnormal patterns (continuous occurrence of abnormal length or frame and reproduction of the same location upon re-imaging), and (iv) a quality gate (satisfaction of minimum requirements such as skeleton connectivity, boundary clarity, and signal-to-noise ratio).

[0327] That is, the user authentication device (1) can calculate the branching angle (interior angle) by extracting branching points along the skeleton generated from the first data as nodes of degree 3 or higher, and estimating the local tangent vector of the segment meeting at each branching point using a least squares straight line approximation using two points separated by a fixed arc length d (e.g., 0.3 to 0.5 mm) from the branching point.

[0328] Next, the user authentication device (1) can calculate the distance between adjacent branch points based on the centerline arc length coordinate(s) along the same blood vessel axis to construct a branch interval profile, and evaluate whether the second standard is violated by comparing the branch angle distribution and branch interval distribution obtained in this way with the user-specific standard value or group statistics.

[0329] At this time, the evaluation can be performed in a fixed sequence of quality gate pre-application → angle / spacing primary judgment → consistency / reproducibility verification → (if necessary) phase auxiliary indicator verification. Branching points adjacent to the ROI boundary (where the field of view may be clipped) or points suspected of projection overlap may be excluded from the final judgment by assigning a hold or reacquisition flag. Additionally, if there is a possibility of two-dimensional angle distortion due to ultrasonic projection characteristics, angle stability can be corrected by utilizing adjacent frame averages or depth approximation information.

[0330] For example, the user authentication device (1) may determine that it is abnormal when the deviation of the branch angle relative to the same user standard distribution is outside the range of mean ± 2 standard deviations, or when an absolute angle of less than 25° (narrow angle) or greater than 135° (excessive angle) is repeated in a continuous interval of 2 mm or more. The branch interval may be determined as abnormal if the distance distribution between adjacent branch points measured along the centerline is less than the lower 5th percentile (clustered overcrowding) or greater than the upper 95th percentile (undercrowding) for n consecutive times (e.g., 3 times). Only when the main indicator is a boundary region (near the threshold), the final determination may be corrected by referring to auxiliary phase indicators (change in the ratio of the number of branch points to the number of edges, increase in the local subgraph editing distance, etc.). The threshold, length, and number of repetitions at this time are merely examples to aid understanding, and in actual implementation, they may be set differently depending on the sensor resolution (pixels / mm), finger size, and sample interval (Δs).

[0331] Through this, the user authentication device (1) can reliably identify structural deformations such as branching patterns (overcrowding, loss of branching, abnormal narrow angles / excessive angles) that deviate from normal physiological categories, and by combining quality gates and persistence / reproducibility conditions, it can reliably detect abnormal blood vessel branching patterns while suppressing false alarms caused by temporary noise or changes in posture.

[0332] Next, the user authentication device (1) can generate a user health abnormality signal based on abnormal blood vessel pattern data including detected blood vessel stenosis patterns and abnormal blood vessel branching patterns.

[0333] Here, user health abnormality signals refer to structured alert data that summarizes the interpretation results of abnormal blood vessel pattern data, including warning level, abnormality type, location coordinates, overall severity score, reliability indicator, occurrence and duration, and follow-up recommended actions.

[0334] That is, the user authentication device (1) first standardizes each indicator included in the abnormal blood vessel pattern data (relative reduction rate of cross-sectional area, rate of increase of blood flow velocity, deviation of branch angle, branch interval variation index, etc.) according to user reference values ​​or group statistics, and can verify position consistency (based on centerline arc length coordinates) in the same finger coordinate system.

[0335] Next, the user authentication device (1) can form a set of standardized scores with reduced noise sensitivity by weighting quality indicators such as alignment quality, skeleton connectivity, and signal-to-noise ratio, and can group abnormality candidates detected in the same blood vessel axis into spatial clusters (based on radius) to organize abnormality candidates at the actual anatomical location unit.

[0336] The user authentication device (1) can generate a user health abnormality signal by performing the following two branch judgment logics in parallel.

[0337] First, rule-based judgment allows for the immediate classification of cases where a single indicator exceeds a clear threshold as an anomaly.

[0338] For example, if a condition in which the relative reduction rate of cross-sectional area within the same blood vessel axis exceeds a threshold value persists for a predefined minimum time or if a section in which the rate of increase in blood flow velocity exceeds a threshold is repeatedly observed in consecutive frames, it can be determined as a stenosis pattern. Regarding branching structures, if branching points in which the branching angle deviation exceeds a threshold value are repeated in consecutive sections, or if changes in which the branching interval becomes significantly shorter than the threshold accumulate, it can be determined as an abnormal blood vessel branching pattern. In this case, the threshold value and duration can be set by reflecting user thresholds, device resolution, and operational policies.

[0339] Second, a composite indicator-based overall severity score can be calculated through score combination rules.

[0340] The user authentication device (1) can calculate a stenosis score by weighted summing standardized scores related to stenosis (reduction in cross-sectional area and increase in speed) and a branching score by weighted summing standardized scores related to branching (branching angle deviation and branching interval variation). Then, the two scores are combined with policy weights to derive a comprehensive severity score, and a correction coefficient reflecting frame persistence ratio, spatial cluster consistency, and alignment quality is multiplied to obtain a final score reflecting reliability indicators. If the comprehensive severity score is above a threshold, a user health abnormality signal can be generated.

[0341] Additionally, the user authentication device (1) may apply criteria for persistence, spatial consistency, and reproducibility together to distinguish between transient noise and structural anomalies. If an anomaly indicator is maintained for a set amount of time within the same location radius or is repeatedly observed in two or more sessions at different points in time, it is confirmed as an anomaly; conversely, if it is a one-time occurrence and the quality indicator is evaluated as low, it can be classified into a lower grade such as continuous observation. Spatial consistency evaluation is performed based on a radius considering the ROI length axis and width, and reproducibility evaluation can be verified by comparison with past records stored in a server or database (200).

[0342] For example, the user authentication device (1) can generate a user health abnormality signal of medium severity if the relative decrease rate of the cross-sectional area at a specific coordinate on the centerline exceeds the upper section of the personal standard distribution and the increase rate of the blood flow velocity in the same section also exceeds the standard is maintained for more than a predefined time.

[0343] As another example, the user authentication device (1) can generate a high-severity user health abnormality signal when a narrow angle pattern in which the branching angle is abnormally small in the same blood vessel axis is continuously observed for a certain length or longer, and a phenomenon in which the adjacent branching interval becomes densely packed into the individual standard sub-section appears together.

[0344] Conversely, the user authentication device (1) can be created by lowering the level to "needs observation" when only one metric slightly exceeds the threshold once and the quality metric is low. In this case, the threshold, weight, duration, and level classification are examples to aid explanation, and in actual implementation, they can be adaptively adjusted to reflect user age group, past records, sensor resolution and signal-to-noise ratio, and operating policy.

[0345] Through this, the user authentication device (1) can clearly distinguish between meaningful structural abnormalities and one-time noise from abnormal blood vessel pattern data to suppress false alarms while marking changes that require attention with rapid and consistent standards, and the generated user health abnormality signal can be provided to the user's terminal (300) in the subsequent warning data transmission stage and used as a basis to induce re-measurement or professional evaluation.

[0346] Finally, the user authentication device (1) can transmit warning data including the generated health abnormality signal to the user's terminal (300).

[0347] The user's terminal (300) is an electronic device owned and managed by the user or having usage rights, and may include, for example, a smartphone, tablet, laptop, desktop, or wearable device, but is not limited thereto.

[0348] The user's terminal (300) can receive and display authentication progress status and warning data through communication with the user authentication device (1).

[0349] That is, the user authentication device (1) can structure warning data including a warning level, summary cause (suspected stenosis, suspected branching abnormality, etc.), approximate coordinates or description of the affected location, measurement time and reliability, and follow-up recommendations (guidance on re-measurement, recommendation for professional consultation, etc.) and safely transmit it via a server or database (200) or directly to the user's terminal (300).

[0350] For example, the user authentication device (1) can define the packaged warning data in JSON or protocol buffer format to include mandatory items (warning identifier, warning level, time of occurrence, location summary, severity score, confidence indicator) and optional items (recent trend, reproducibility indicator, reference snapshot hash, remeasurement guide link), and in the network segment, apply Transport Layer Security (TLS) and integrity tags to transmit the message through an encrypted and verified channel.

[0351] The user's terminal (300) can present a warning to the user through a dedicated application or web-based interface, such as a warning pop-up, push notification, or badge display, and provide a summary of the cause of the warning, a description of the location (e.g., near the first joint, relative coordinates on the fingertip reference length axis), severity and reliability, and recommended measures on a detailed screen.

[0352] Additionally, the user authentication device (1) can increase the reliability of warning delivery even in the event of a temporary communication failure by operating a transmission acknowledgment (ACK), retransmission, and delay queue. It can also apply differential notifications by grade, such as displaying a re-measurement workflow along with an immediate notification for high-grade warnings, and recommending re-verification during the next authentication for low-grade warnings. These data items, notification methods, and security mechanisms are merely examples to aid understanding, and in actual implementation, the items, frequency, and display methods may be adjusted according to the organization's policy, the scope of user consent, and regional regulations.

[0353] Through this, the user authentication device (1) can quickly and safely notify the user of potential abnormalities detected from internal blood vessel information derived during the authentication process, and the user can check the warning content on their terminal (300) and select follow-up measures such as re-measurement or professional consultation. At this time, the warning data is provided as reference information rather than a diagnostic result and can be utilized as an auxiliary means to enhance user safety and induce early response.

[0354] Although embodiments of the present invention have been described in more detail with reference to the attached drawings, the present invention is not necessarily limited to these embodiments and may be modified in various ways within the scope of the technical spirit of the present invention. Accordingly, the embodiments disclosed in the present invention are intended to explain, not limit, the technical spirit of the present invention, and the scope of the technical spirit of the present invention is not limited by these embodiments. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. The scope of protection of the present invention shall be interpreted by the claims below, and all technical spirits within an equivalent scope shall be interpreted as being included within the scope of rights of the present invention.

[0355] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below. Explanation of the symbols

[0357] 1: Finger vein and finger joint pattern-based user authentication device 100: Ultrasonic sensor 110: Optical sensor 120: Region of Interest Extraction Unit 130: Data Processing Department 140: Data Matching Section 150: Feature point extraction unit 160: Feature Point Register 170: User Authentication Section 200: Server or database 300: User's terminal

Claims

Claim 1 An ultrasonic sensor for acquiring an image or video of a finger vein inside the finger (first data); an optical sensor for acquiring an image or video of the external shape of the finger (second data); a region of interest extraction unit for extracting a region of interest for each of the first and second data; a data processing unit for performing segmentation and skeleton processing on the extracted region of interest; a data matching unit for aligning the first and second data after the segmentation and skeleton processing is completed; a feature point extraction unit for extracting a plurality of feature points based on the processed first and second data; a feature point registration unit for storing the extracted plurality of feature points in a server or database and registering them for each user; and a user authentication unit for authenticating a user based on the registered feature points.The method includes, wherein the feature point extraction unit is controlled to extract relative position information of the intersection point of the finger vein pattern and the finger joint pattern, and the finger vein branch point detected around the intersection point, using the plurality of feature points; the user authentication unit performs matching between the registered feature point and the extracted feature point, and simultaneously analyzes data acquired using the optical sensor to calculate a first authentication reliability based on first habit data, which is the user's finger movement trajectory; second habit data, which is the finger entry inclination relative to the reference plane of the shooting area; and third habit data, which is the fingertip position shake; analyzes data acquired from the optical sensor to detect the user's gesture trajectory, and identifies at least one of a finger contact motion, a finger tilt transition motion, a circular trajectory motion, or a zigzag trajectory motion from the gesture trajectory to calculate a second authentication reliability, wherein the calculation of the second authentication reliability is performed only when at least one of a first environment corresponding to the user's authentication information registration procedure, a second environment identified as high risk according to a preset standard, and a third environment in which an error occurred during the matching process between the feature points is detected. A user authentication device based on finger vein and finger joint patterns, characterized by being controlled and configured to determine whether to grant final authentication approval to a user by combining the matching result between the feature points and the calculated first authentication reliability and second authentication reliability. Claim 2 A user authentication device based on finger vein and finger joint patterns according to claim 1, wherein the region of interest extraction unit is controlled to extract a first ROI from the finger vein image, the region including the first joint of the finger closest to the fingernail and a portion separated from therein by a predetermined distance (22~24mm) in the vertical direction, and to extract a second ROI from the finger image, the region including the first joint of the finger and a portion separated from therein by a predetermined distance (22~24mm) in the vertical direction, and wherein the data processing unit is controlled to perform segmentation and skeleton processing on the extracted first ROI and second ROI, respectively. Claim 3 A user authentication device based on finger vein and finger joint pattern according to claim 1, wherein the feature point extraction unit is controlled to extract the intersection point of the finger vein pattern and the finger joint pattern as a first feature point, extract the feature of the finger joint pattern itself as a second feature point, and extract the feature of the finger vein itself as a third feature point. Claim 4 A user authentication device based on finger vein and finger joint pattern according to claim 3, wherein the second feature point comprises at least one of the number of skeleton lines of the joint pattern, the total length of the joint pattern, the ratio of the number of branch points to the number of endpoints of the joint pattern, the skeleton width of the joint pattern, and the position coordinates of the branch points or endpoints of the joint pattern expressed on a normalized finger coordinate system. Claim 5 A user authentication device based on a vein and finger joint pattern according to claim 3, wherein the third feature point comprises at least one of the endpoint, branch point, intersection point of a vein skeleton line, line length of an individual vessel skeleton, distance between the endpoint and the branch point, branch angle, curvature of the vein skeleton curve, and distance between the centerlines of the individual vessel skeleton lines. Claim 6 A user authentication device based on a vein and finger joint pattern according to claim 3, wherein the third feature point comprises at least one of the number of skeleton lines of a vein, the total length of the vein skeleton lines, the ratio of the number of branch points to the number of endpoints of the vein, the width of the vein skeleton, and the position coordinates of the branch points or endpoints of the vein expressed on a normalized finger coordinate system. Claim 7 A user authentication device based on a finger vein and finger joint pattern according to claim 3, wherein the user authentication unit is controlled to identify and authenticate a user based on a plurality of authentication steps, and the plurality of authentication steps include a first authentication step that identifies and authenticates a user based on a first feature point, a second authentication step that identifies and authenticates a user based on a second feature point, and a third authentication step that identifies and authenticates a user based on a third feature point. Claim 8 A user authentication device based on a vein and finger joint pattern according to claim 7, wherein the user authentication unit is controlled to identify and authenticate a user by sequentially performing the first authentication step, the second authentication step, and the third authentication step. Claim 9 A user authentication device based on a vein and finger joint pattern according to claim 7, wherein the user authentication unit is controlled to identify and authenticate a user by performing at least two selected steps among the first authentication step, the second authentication step, and the third authentication step simultaneously or sequentially. Claim 10 A step of acquiring an image or video of the finger veins inside the finger using an ultrasonic sensor (first data); a step of acquiring an image or video of the external shape of the finger using an optical sensor (second data); a step of extracting an area corresponding to the first to 1.5 joints from the fingertip as a first ROI from the first data, and extracting an area adjacent to the first joint and a predetermined distance therefrom as a second ROI from the second data; a step of generating first processed data and second processed data by performing segmentation and skeleton processing on each of the first and second ROIs; a step of aligning and matching the positions of the first and second processed data; a step of extracting a plurality of feature points based on the first and second processed data; a step of storing the extracted plurality of feature points in a server or database and registering them for each user; The method includes a step of authenticating a user based on the registered feature points; wherein the plurality of feature points include the intersection points of a finger vein pattern and a finger joint pattern, and relative position information of finger vein branch points detected around the intersection points; and after the step of authenticating a user based on the registered feature points, a step of calculating a first authentication reliability based on the user's behavioral habit data;The method further includes the step of calculating a first authentication reliability based on the user's behavioral habit data, comprising: a step of obtaining first habit data by analyzing data obtained using the optical sensor to extract the trajectory of the user's finger moving within the shooting area of ​​the optical sensor; a step of obtaining second habit data by analyzing data obtained from the optical sensor to calculate the entry slope of the finger relative to the reference plane based on the reference plane of the shooting area of ​​the optical sensor; a step of obtaining third habit data by analyzing data obtained from the optical sensor to detect the positional shaking of the fingertip in consecutive frame images before and after the point of entry into the shooting area of ​​the optical sensor according to a preset threshold value; a step of storing the first habit data, second habit data, and third habit data in a server or database to register or update the user's reference behavioral habit data; and a step of calculating a first authentication reliability according to a preset standard by comparing the user's behavioral habit data obtained during authentication based on the feature point with the user's reference behavioral habit data; and a step of calculating a second authentication reliability based on the user's gesture data.A user authentication method based on finger vein and finger joint patterns, comprising: a step of calculating a second authentication reliability based on the user’s gesture data, wherein the step of acquiring gesture data by analyzing data acquired from the optical sensor to detect a gesture trajectory performed by a finger within the shooting area of ​​the optical sensor; a step of identifying at least one motion among a finger contact motion, a finger tilt transition motion, a circular trajectory motion, or a zigzag trajectory motion from the gesture data; and a step of calculating a second authentication reliability according to a preset standard by comparing the identified gesture data with a preset reference gesture data of the corresponding user, wherein the step of calculating the second authentication reliability is performed only when at least one environment is detected among a first environment corresponding to a user authentication information registration procedure, a second environment identified as high risk according to a preset standard, and a third environment in which an error occurred during the process of authenticating the user based on the registered feature points.

Images