Living tissue detection system

Abstract

To provide a technique of rapidly detecting a living tissue.SOLUTION: A client 100 includes: an image acquisition unit 1020 for acquiring a camera image in series; a posture estimation unit 1050 for estimating the posture of a hand from the acquired camera image; and a living tissue detection unit 1060 for determining that the hand is a living tissue when a specific action of a hand is detected from the estimated posture.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a biological detection system. [Background technology]

[0002] Biometric authentication, which verifies that a user is a person registered in advance based on an individual's physical and behavioral characteristics, is highly convenient and has attracted attention as a reliable method of identity verification because it does not require the user to memorize a password or carry an IC card or other personal item. Biometric authentication extracts feature values ​​from physical and behavioral characteristics such as fingerprints, faces, irises, veins, and palm prints, and compares these feature values ​​with those at the time of registration to determine whether the user is the actual person.

[0003] One of the vulnerabilities of biometric authentication is the risk of spoofing using artifacts. For example, if a third party obtains a person's biometric information, such as fingerprints, face, iris, veins, or palm print, and uses that information to create an artifact containing the person's biometric information, and presents that artifact to the biometric authentication system, there is a risk that the third party could successfully authenticate the person fraudulently.

[0004] In particular, in biometric authentication systems that use general-purpose cameras to acquire and authenticate biometric information, it is easy to spoof someone's identity by presenting a piece of paper with a biometric image printed on it or a tablet with a biometric image displayed on it to the biometric authentication system. For this reason, it is essential that biometric authentication systems also use liveness detection technology that detects that the presented object is a part of a living human body and not an artificial object.

[0005] There are four main approaches to achieving this liveness detection:

[0006] The first is sensor-based detection. This method uses a sensor other than a general-purpose camera to sense the electrical characteristics and 3D shape of the target and detect that it is a living body. Although this method is highly reliable, it requires a dedicated sensor, which makes it expensive and difficult to introduce into existing devices.

[0007] The second is learning-based detection. This method uses machine learning to learn from images of faces, veins, palm prints, etc., and detects by classifying them into living bodies and artificial bodies. This method prepares a large amount of images of living bodies and images of various patterns of artificial bodies as training data, and trains a classifier that can distinguish between them. In machine learning methods, performance decreases when there is a difference between the training data and the data to be classified, so accuracy decreases when the conditions under which the training data was collected (type of artificial body, environmental conditions such as lighting and white balance, sensors such as cameras) differ from the conditions at the time of live body detection.

[0008] The third is the detection of biological reactions. This method acquires biological reactions such as blinking, breathing, and pulsation, and uses them to detect live bodies. Since these biological reactions are minute features, in order to capture them reliably, the frame rate of the video, the resolution, the optical performance of the camera, etc. must be sufficiently high. If detection is attempted under conditions where any of these performances are low, there is a possibility that even a living body will be mistaken for an artificial object. For this reason, it is necessary to impose high constraints on the camera and the hardware that processes it, making it difficult to make it usable on any general-purpose camera device.

[0009] The fourth type is challenge-response type detection. This method is often used in eKYC (electronic Know Your Customer), which performs reliable identity verification via a network, and detects specific actions by requesting the user to close their eyes, turn their head, wink, etc. However, many of these methods take time to recognize, perform, and detect actions, and no method is known that can detect actions instantly during authentication.

[0010] Patent Document 1 relates to live body detection using a general-purpose camera without relying on a dedicated sensor, and performs live body detection by determining whether the movement, distortion, shape change, feature information, and three-dimensional state of the living body are each within the range of live body information. This can be considered as one of the fourth approaches. [Prior art documents] [Patent documents]

[0011] [Patent Document 1] Patent No. 6482816 Summary of the Invention [Problem to be solved by the invention]

[0012] The technology described in Patent Document 1 requires the living body to perform complex and random actions, such as "make your hand clenched, and after five seconds, raise your thumb and little finger up," and detects the living body using the finger movement speed and background color information at that time, making it difficult to detect quickly.

[0013] In addition, the technology described in Patent Document 1 is a rule-based method that performs detection by setting thresholds for each of the biological movement, distortion, shape change, feature information, three-dimensional state, etc., and therefore it is considered difficult to sufficiently reduce both the false positive rate and the non-detection rate.

[0014] Therefore, an object of the present invention is to provide a technique that enables rapid live body detection. [Means for solving the problem]

[0015] In order to solve the above problems, one representative liveness detection system of the present invention is a biometric authentication system that includes an image acquisition unit that acquires images in time series, a posture estimation unit that estimates the posture of the hand from the acquired images, and a liveness detection unit that determines that the hand is live when a specific movement of the hand is detected from the estimated posture. Effect of the Invention

[0016] According to the present invention, live body detection can be performed quickly.

[0017] Problems, configurations and effects other than those described above will become apparent from the following description of the embodiments. [Brief description of the drawings]

[0018] [Figure 1] FIG. 2 is a block diagram showing an example of a configuration of a client according to the present embodiment. [Diagram 2] FIG. 2 is a block diagram showing an example of a hardware configuration of a client according to the present embodiment. [Diagram 3] 11 is a flowchart illustrating an example of a biometric registration process including a biometric detection executed by a client according to the present embodiment. [Figure 4] 10 is a diagram showing an example of feature points of a hand extracted by a feature point acquisition unit; FIG. [Diagram 5] 11A and 11B are diagrams illustrating an example of a detection situation of a gesture of rotating a palm drawn by an image output unit. [Figure 6] 11A and 11B are diagrams illustrating an example of a detection situation of a pinching action with fingers depicted by an image output unit. [Figure 7] 11A and 11B are diagrams illustrating an example of a detection situation of an action of pressing a button with a finger drawn by an image output unit. [Figure 8] 11 is a flowchart illustrating an example of a biometric authentication process including liveness detection executed by a client according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0019] Hereinafter, the embodiments will be described with reference to the drawings.

[0020] This embodiment is a biometric authentication system in which a client acquires a camera image, estimates a hand posture from the camera image, detects a specific motion from the hand posture, and performs biometric registration and biometric authentication when the detection is successful.

[0021] FIG. 1 is a block diagram showing an example of the configuration of a client according to this embodiment.

[0022] 1, a client 1000 includes a user ID input unit 1010, an image acquisition unit 1020, a feature point acquisition unit 1030, an image output unit 1040, a posture estimation unit 1050, a liveness detection unit 1060, a liveness detection unit 1070, a biometric feature extraction unit 1071, a template generation unit 1072, a biometric authentication unit 1073, and a template storage unit 1090. Each unit will be described below.

[0023] A user ID input unit 1010 acquires an input of an ID given to a user via a keyboard, a card reader, a memory on the client 1000, or the like. An image acquisition unit 1020 acquires a camera image from a camera connected to the client 1000. A feature point acquisition unit 1030 extracts feature points of a hand from the camera image acquired by the image acquisition unit 1020.

[0024] The posture estimation unit 1050 estimates the posture of the hand based on the coordinates of the feature points of the hand extracted by the feature point acquisition unit 1030. The liveness detection unit 1060 performs liveness detection based on the hand posture estimated by the posture estimation unit 1050. When this liveness detection detects a specific movement from the posture of the hand, it determines that the hand in the camera image is a live body.

[0025] The image output unit 1040 depicts the result of the live body detection on the camera image and outputs it to an output device such as a display.

[0026] The biometric detection unit 1070 detects a biometric entity from a camera image. This biometric detection is a process of identifying the position of a biometric entity in a camera image, and for example, detects the position of the face area in face authentication, detects the position of the finger area in finger vein authentication, and detects the position of the palm area in palm print authentication.

[0027] The biometric feature extraction unit 1071 extracts biometric features from the biometrics detected by the biometric detection unit 1070. In the case of face authentication, this biometric feature extraction involves extracting a face image and inputting it to a machine learning model to obtain feature quantities. In the case of finger vein authentication, an image of the finger is extracted to obtain a vein pattern with the blood vessel pattern emphasized. In the case of palm print authentication, an image of the palm is extracted to obtain a palm print pattern.

[0028] The template generating unit 1072 generates a template for enrollment based on the biometric features extracted by the biometric feature extracting unit 1071. The template storage unit 1090 stores the templates generated by the template generating unit 1072.

[0029] The biometric authentication unit 1073 performs biometric authentication by comparing the biometric features extracted by the biometric feature extraction unit 1071 with the templates registered in the template generation unit 1072 .

[0030] Next, the hardware configuration of the client 1000 will be described with reference to FIG.

[0031] FIG. 2 is a block diagram showing an example of a hardware configuration of the client 1000 of this embodiment.

[0032] 2, the client 1000 includes a CPU (Central Processing Unit) 8010, a main memory device 8020, an auxiliary memory device 8030, an input device 8040, an output device 8050, and a communication device 8060.

[0033] The CPU 8010 is a central processing unit, and implements necessary functions by executing programs stored in the main memory device 6020. The programs include programs for realizing the user ID input unit 1010, image acquisition unit 1020, feature point acquisition unit 1030, image output unit 1040, posture estimation unit 1050, liveness detection unit 1060, liveness detection unit 1070, biometric feature extraction unit 1071, template generation unit 1072, and biometric authentication unit 1073 shown in FIG.

[0034] The primary storage device 6020 is a primary storage device used when the CPU 8010 executes processing, and is composed of a volatile storage element such as a RAM (Random Access Memory). The secondary storage device 8030 is a storage device represented by a HDD (Hard Disk Drive) or an SSD (Solid State Drive), and corresponds to the template storage unit 1090. Data stored in each unit is accumulated as data on the secondary storage device 8030.

[0035] The input device 8040 is used to read information in the user ID input unit 1010 and the image acquisition unit 1020, and is composed of a keyboard, a biosensor, a touch panel, a smart device, a scanner, a camera, etc. The output device 8050 is used to output information in the image output unit 1040, and is composed of a display, etc. The communication device 8060 is used by the client 1000 to communicate with an external device.

[0036] With the above configuration, in this embodiment, biometric registration and biometric authentication are performed after confirming that the hand is not an artificial object based on the hand movement estimated from the camera image.

[0037] Next, a procedure for performing biometric registration using the client 1000 will be described with reference to FIG.

[0038] FIG. 3 is a flowchart showing an example of a biometric registration process including a biometric detection executed by the client 1000 of this embodiment.

[0039] First, the user ID input by the user is acquired by the user ID input unit 1010 in the client 1000 (step S2010). The user ID is typically represented as a character string of any length.

[0040] Next, the camera is started (step S2020), and camera images including the user's hand are acquired in time series using the image acquisition unit 1020 (step S2030). The image acquisition unit 1020 is a general-purpose camera mounted on the client 1000, and acquires images using the camera. The acquired camera images are displayed on the output device 8050 such as a display of the client 1000 by the image output unit 1040.

[0041] Next, feature points of the user's hand are extracted from the camera image acquired in step S2030 using feature point acquisition unit 1030 (step S2040). An example of feature points of the hand is shown in FIG.

[0042] FIG. 4 is a diagram showing an example of feature points of a hand extracted by feature point acquiring unit 1030. As shown in FIG.

[0043] As shown in FIG. 4, by setting feature points 4010 at positions such as the tips of the fingers, joints, and the bottom end of the palm in a camera image and estimating these positions, it is possible to estimate the position and posture of the hand.

[0044] Returning to Fig. 3, a hand tracking process is performed based on the coordinates of the hand feature points obtained in step S2040, and it is determined whether tracking has been successful (step S2050). In this tracking process, the hand coordinates in the previous frame are compared with the hand coordinates in the current frame, and if the coordinates are sufficiently close, they are deemed to be the same hand and tracking is successful, but if the coordinates are apart by a certain value or more, they are deemed to be different hands and tracking is unsuccessful. The coordinates used for this determination can be the coordinates of a rectangular area including the entire hand or the coordinates of each feature point.

[0045] If tracking fails in step S2050, the live body detection result is initialized to the initial state, "detecting," and the process returns to step S2030 (step S2061).

[0046] On the other hand, if tracking is successful in step S2050, the process branches depending on whether the live body detection result is "detection in progress" or "detection successful" (step S2060).

[0047] If the liveness detection result is "detected" in step S2060, the hand posture is estimated based on the coordinates of the hand feature points extracted in step S2040 (step S2070).

[0048] The posture of the hand to be estimated depends on a specific motion of the hand to be detected. The specific motion of the hand may include, but is not limited to, any motion including a motion of rotating the palm, a motion of pinching with fingers, and a motion of pressing a button with fingers. The specific motion of the hand to be detected may be the same motion every time, or may be randomly changed every time liveness detection is performed.

[0049] The action of rotating the palm is to rotate it horizontally to turn it over, but since images printed on paper or displayed on a display cannot be turned over, this detection cannot be successful. In this action, the hand posture that needs to be estimated is the rotation angle of the palm relative to the camera. When the feature points of the hand are expressed in three-dimensional coordinates, this rotation angle can be obtained by calculating the angle at which the hand is tilted with respect to the line connecting the camera and the hand. In addition to the method of directly estimating the rotation angle of the hand, a method can also be adopted in which the palm becomes longer vertically when rotated, and the rotation angle can be indirectly estimated from the accuracy obtained from the two-dimensional coordinates of the feature points that make up the palm, or the aspect ratio of the rectangular area containing the palm.

[0050] The action of pinching with fingers involves placing an object to be pinched at an arbitrary position on an image captured by a camera, and then pinching the object with the fingers. Images printed on paper or images displayed on a display cannot be detected successfully because they do not allow for three-dimensional pinching movements. In the case of this action, the hand posture that needs to be estimated is the coordinates of the fingertips and the distance between the pinching fingers. When pinching with the thumb and index finger, the coordinates of the tip of the thumb and the tip of the index finger and the distance between them are calculated from the feature point coordinates.

[0051] The action of pressing a button with a finger involves placing the button to be pressed at an arbitrary position on an image captured by a camera and pressing the button on that position, and it is difficult to successfully detect this action with an image printed on paper or an image displayed on a display. There is a possibility of successful detection by bringing an artificial object closer to the camera, but in that case, the background area other than the hand included in the artificial object will also move in the same way. Therefore, it is also possible to analyze the movement between frames of the hand area and the background area and detect that the movement of the background is not linked to the hand. In this action, the hand posture that needs to be estimated is the depth coordinate of the feature point of the tip of the pressing finger (for example, the index finger) and the angle of the finger with respect to the camera. Note that instead of the coordinates and angle of the finger, it is also possible to determine that the button has been pressed by placing the finger on the coordinates of the button for a certain period of time. These can be obtained from the feature point coordinates of the finger.

[0052] Based on the hand posture estimated in step S2070, if a specific hand movement is detected, the liveness detection result is updated to "detection successful", and if a specific hand movement cannot be detected, the result remains as "detection in progress" (step S2072).

[0053] For example, if the specific hand motion is a palm rotation motion, liveness detection is successful when the palm is rotated a predetermined angle, and "detection successful" is substituted for the liveness detection result. In this case, if the user is requested to rotate the palm facing the camera, turn the back of the hand once toward the camera, and then rotate again to face the palm toward the camera, the palm will be facing the camera when liveness detection is successful, and the user can be guided to the optimal posture for the finger vein authentication and palm print authentication that are performed afterwards. Therefore, this method not only performs liveness detection, but also has the potential to contribute to the speed and accuracy of subsequent biometric authentication.

[0054] In addition, if the specific hand movement is a pinching movement with the fingers, the liveness detection is deemed successful when the distance between the tips of the two fingers on the object to be pinched is less than a certain value, and the liveness detection result is assigned the value "detection successful."

[0055] In addition, if the specific hand movement is the movement of pressing a button with a finger, the liveness detection is deemed successful when the coordinates of the fingertip or the angle of the finger on the object to be pressed (e.g. a button) reach or exceed a certain level, and the liveness detection result is assigned "detection successful."

[0056] If the above conditions are not met, it is determined that live body detection has not been completed, and the fine detection result remains as "detection in progress."

[0057] The hand posture estimated in step S2070 and the detection status of the specific hand motion in step S2072 are drawn by the image output unit 1040 superimposed on the camera image displayed on the output device 8050 such as a display of the client 1000 (step S2090).

[0058] Here, the detection situation depicted in step S2090 will be described with reference to FIGS.

[0059] FIG. 5 is a diagram showing an example of a detection situation of a gesture of rotating a palm drawn by the image output unit 1040. In FIG.

[0060] 5, the rotation direction of the hand is shown in the camera image by an arrow 5010 divided into multiple parts, and the color of parts of the arrow changes depending on the amount of rotation of the hand. This makes it possible to provide feedback to the user on the rotation direction and current amount of rotation of the hand.

[0061] FIG. 6 is a diagram showing an example of a detection situation of a pinching action with fingers drawn by the image output unit 1040. In FIG.

[0062] In the example of Figure 6, a toggle switch 6010, which is operated by pinching it with your fingers, is displayed in a random location on the camera image. If the user pinches the toggle switch 6010 and live presence is successfully detected, an image is displayed indicating that the position of the toggle switch 6010 has changed and it is now ON, making it possible to provide feedback to the user that the toggle switch 6010 has been pinched.

[0063] FIG. 7 is a diagram showing an example of a detection situation of an action of pressing a button with a finger drawn by the image output unit 1040. In FIG.

[0064] In the example of Figure 7, a virtual button 7010 is displayed at a random location on the camera image, and when the user presses the button 7010 with his or her finger and live presence is successfully detected, the button 7010 is changed to the pressed version, making it possible to provide feedback to the user that the button 7010 has been pressed.

[0065] After the resultant image is drawn in step S2090, it is determined whether or not a timeout has occurred (step S2092). If a certain amount of time has passed since the start of processing and a timeout has occurred in step S2092, the registration is considered unsuccessful and the processing is terminated. If the elapsed time is still less than the certain amount of time and a timeout has not occurred in step S2092, the processing returns to step S2030 and continues.

[0066] On the other hand, if the liveness detection result is "detection successful" in step S2060, the liveness detection process is completed and the liveness registration process begins. First, liveness detection is performed from the camera image (step S2080). This liveness detection is a process of identifying the position of the liveness in the camera image. For example, in face authentication, the position of the face area is detected, in finger vein authentication, the position of the finger area is detected, and in palm print authentication, the position of the palm area is detected.

[0067] Next, biometric feature extraction is performed on the biometrics detected in step S2080 (step S2082). In the case of face authentication, this biometric feature extraction involves extracting an image of the face and inputting it into a machine learning model to obtain features. In the case of finger vein authentication, an image of the finger is extracted and a vein pattern with the blood vessel pattern emphasized is obtained. In the case of palm print authentication, an image of the palm is extracted and a palm print pattern is obtained.

[0068] Based on the biometric features extracted in step S2082, a template for enrollment is generated (step S2084), and it is determined whether all of the processes in steps S2080, S2082, and S2084 have been successful and the template for enrollment has been successfully generated (step S2086).

[0069] If it is determined in step S2086 that the template generation has failed, the process proceeds to step S2090, where an attempt is made to generate the template again.

[0070] On the other hand, if it is determined in step S2086 that the template generation was successful, the generated template is linked to the user ID entered in step S2010 and registered in template storage unit 1090 (step S2088), and the processing is completed, with the biometric registration being deemed successful.

[0071] 3, biometric registration is performed by detecting biometrics and confirming that the biometrics are not artificial, and then the biometrics are registered and the template is registered in the template storage unit 1090. This completes the initial registration of the biometric authentication system, and biometric authentication can be used thereafter.

[0072] Next, a procedure for performing biometric authentication using the client 1000 will be described with reference to FIG.

[0073] Fig. 8 is a flowchart showing an example of biometric authentication processing including liveness detection executed by the client 1000 of this embodiment. In Fig. 8, the same processes as those in Fig. 3 are denoted by the same reference numerals, and the description thereof will be omitted.

[0074] 8, after the user ID is obtained in step S2010, a template registered in template storage unit 1090 is obtained (step S3012). Since the template is associated with the user ID and stored in template storage unit 1090 in step S2088, the template can be obtained using the user ID as a key. Note that steps S2010 and S3012 are required in 1:1 authentication in which authentication is performed by specifying a user, but steps S2010 and S3012 are not required in 1:N authentication in which the user is identified without specifying a user.

[0075] The steps from acquiring a camera image to live body detection (steps S2030, S2040, S2050, S2060, S2061, S2070, S2072, S2090, and S2092) are the same as those in the live body registration process of Fig. 3. However, if live body detection is performed only during registration and not during authentication, it is also possible to skip steps S2030 to S2092 and proceed to live body detection step S2080 after camera startup step S2020.

[0076] The biometric detection step S2080 and the biometric feature extraction step S2082 are the same processes as those for biometric registration.

[0077] Next, biometric authentication is performed by comparing the biometric features extracted in step S2082 with the template associated with the user ID input in step S2010 and registered in template generating unit 1072 (step S3084).

[0078] Next, it is determined whether the biometric features extracted in step S2082 match the template associated with the user ID input in step S2010 and registered in template generation unit 1072, and whether the biometric authentication has been successful (step S3086).

[0079] If the biometric authentication is successful in step S3086, the authentication is deemed successful and the process ends. On the other hand, if the biometric authentication is unsuccessful in step S3086, the process proceeds to step S2090 and continues.

[0080] According to the biometric registration process in FIG. 8, biometric detection is performed to confirm that the subject is not an artificial object, and then biometric authentication can be performed.

[0081] In addition, in the case of biometric authentication using other parts of the body, such as face authentication, parts of the body other than the hands can also be detected and tracked.

[0082] Furthermore, when attempting spoofing using an artificial object in a system that uses a body part other than the hand for authentication, there are two possible attacks: one in which the hand and another body part are both artificial, and one in which only the body part used for authentication is an artificial object.

[0083] In the former case, it is necessary to present a hand and another body part as an artificial object at the same time and perform a specific action. This is much more difficult than an attack that presents another body part alone as an artificial object. For example, it is technically difficult to artificially generate a hand and a face at the same time using Deep fake technology to perform a specific action in real time.

[0084] On the other hand, in the latter case, only the artificial body parts are used and the hand is human, so the technical difficulty is lower than in the former case. However, it is difficult to break through liveness detection and successfully authenticate because the hand and artificial object must be presented properly.

[0085] Furthermore, by detecting that the hand and another body part are the same living body by observing the synchronization of movements and the agreement of pulsation periods, it is possible to perform even more reliable living body detection.

[0086] According to this embodiment, feature points of the hand are extracted from a camera image, the posture of the hand is estimated from the feature points, and specific movements of the hand are detected from the posture of the hand to determine that the hand is a living body, thereby enabling rapid live body detection.

[0087] Furthermore, according to this embodiment, a general-purpose camera can be used to acquire camera images, so there is no need to prepare dedicated sensors or high-performance cameras, reducing costs and allowing this embodiment to be introduced into existing devices.

[0088] Furthermore, according to this embodiment, the specific movement of the hand to be detected is large, so that it is less affected by the frame rate and resolution, and liveness detection can be performed stably.

[0089] Furthermore, according to this embodiment, since machine learning is not used for live body detection, it is less susceptible to changes in environmental conditions or changes in camera, and can also handle unknown man-made objects.

[0090] Furthermore, according to this embodiment, since live body detection is performed by detecting only the hand movements, live body detection can be performed without capturing an image of the face.

[0091] The present invention is not limited to the above-described embodiments, and includes various modified examples. For example, the above-described embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those including all of the configurations described. It is also possible to replace a part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

Claims

1. In a biometric authentication system that performs identity authentication using pre-registered biometric information, an image acquisition unit that acquires images in time series; a posture estimation unit that estimates a posture of the hand from the acquired image; a biometric detection unit that, when detecting a specific movement of the hand from the estimated posture, determines that the hand is a biometric.

2. 2. The biometric authentication system according to claim 1, A biometric authentication system in which the specific action is pressing a button with a finger.

3. 2. The biometric authentication system according to claim 1, A biometric authentication system, wherein the specific motion is a palm rotation motion.

4. 2. The biometric authentication system according to claim 1, A biometric authentication system, wherein the specific action is a pinching action with fingers.

5. 2. The biometric authentication system according to claim 1, a living body detection unit that detects a living body from the acquired image; a biometric feature extraction unit that extracts biometric features from the detected biometric information; a registration unit that registers the extracted biometric feature when the biometric detection unit determines that the hand is a biometric feature.

6. 6. The biometric authentication system according to claim 5, A biometric authentication system including a biometric authentication unit that, when the biometric detection unit determines that the hand is a living body, performs identity authentication by comparing the biometric features extracted by the biometric feature extraction unit with the biometric features registered in the registration unit.

7. 2. The biometric authentication system according to claim 1, A biometric authentication system comprising an image output unit that displays the acquired image while superimposing a detection status of the specific motion on the acquired image.

8. 2. The biometric authentication system according to claim 1, a feature point extraction unit that extracts feature points of the hand from the acquired image; The posture estimation unit estimates the posture of the hand from the extracted feature points.

Images