Biometric authentication device and biometric authentication method for authenticating a person with reduced computational complexity

Abstract

A biometric authentication device (30) for authenticating a person, which authenticates a person by comparing the biometrics of a person's body part (4) with pre-stored biometrics of body parts of a plurality of registrants, the biometric authentication device (30) being configured to: capture image data of a person's body part (4); determine at least one characteristic geometric body part attribute (Sf, Wf, Wh) of the person's body part (4) and the biometrics of the person's body part (4) from the captured image data; and define a comparison between the biometrics of the person's body part (4) and the pre-stored biometrics of body parts of a plurality of registrants, the pre-stored biometrics having a predetermined similarity level with respect to at least one characteristic geometric body part attribute (Sf, Wf, Wh) of the person.

Description

Technical Field

[0001] This disclosure relates to a biometric authentication device and method for authenticating a person. In particular, this disclosure relates to a biometric authentication device and method for authenticating a person with reduced computational complexity. Background Technology

[0002] Biometric authentication devices are widely used for human authentication, for example, in the context of access control for resources such as buildings, rooms, computers, smartphones, electronic bank accounts, voting systems, school or university exam papers, borders, company registers, etc.

[0003] In some embodiments, the biometric authentication device is configured to capture image data of a human body part (e.g., a human hand) and determine individual and typical biometric features from the captured image data. The captured image data of a human hand or body part may involve image data captured using a near-infrared light sensor (e.g., 700 nm to 900 nm), image data captured using a visible light sensor (e.g., 400 nm to 600 nm), or a combination thereof. The biometric features determined from the image data may involve vein patterns, palm prints, life lines, etc. Image data captured in near-infrared light can determine features related to vein patterns of the hand. Image data captured in visible light can determine features associated with palm prints and life lines of the hand.

[0004] Human authentication is based on pre-stored biometrics registered under the control of a qualified and trusted institution. For example, this institution verifies a person's identity based on an identification card (e.g., a passport). Individual image data of a person's hand are captured, and biometrics of the hand or body part are determined from the captured image data. The determined biometrics are stored in a database as pre-stored biometrics. In some embodiments, the pre-stored biometrics may partially or entirely comprise the captured image data. The determined biometrics of the hand or body part, the captured image data of the hand, or a combination thereof may involve vein patterns, palm prints, life lines, etc.

[0005] Later, if human authentication is required, the biometric authentication device captures image data of the person's hand or body part. The biometrics of the person's hand or body part are determined and compared with pre-stored biometrics and / or pre-stored image data. If a match is found within the pre-stored biometrics, the person is authenticated; otherwise, authentication is rejected.

[0006] If the pre-stored biometrics involve only a small number of people (e.g., fewer than a few dozen), biometric comparisons using a "brute-force matching" method can provide a sufficiently fast response time (e.g., 1 second or less). If the pre-stored biometrics involve, for example, 1000 or more people, the computational complexity increases significantly, and biometric comparisons require sufficiently high processing power to achieve, for example, a response time of 1 second or less. However, sufficiently fast computers need, for example, parallel processing capabilities, which in turn leads to high energy consumption. Therefore, complex and expensive installations are required. For this reason, the widespread deployment of human authentication based on hand or body part biometrics is currently limited.

[0007] In existing technologies, it is known to accelerate the search for objects in a database by assigning a unique key to each object. For example, it is common practice for each person's personal card (such as an ID card or credit card) to have a unique key. A card reader that controls access to a specific resource (such as a building entrance or a credit card account balance) reads the key from that person's personal card, and if a key match is found in a pre-stored database, grants that person access to the resource. Searching for unique keys in a pre-stored database is accurate and fast.

[0008] Because matching biometrics of a person's hand or body part must be based on a measured similarity between biometrics determined using a biometric authentication device and pre-stored biometrics, it is impossible to assign a unique key to the biometrics of a person's hand or body part. For example, the difference between two n-dimensional feature vectors must be determined, and this difference must be compared to a sufficiently small threshold. Therefore, since a unique key cannot be assigned to the biometrics of a person's hand or body part, techniques known in the art for searching for unique keys in a pre-stored database cannot be used.

[0009] The paper "On Hierarchical Palmprint Coding With Multiple Features for Personal Identification in Large Databases (You et al.), IEEE TRANSACTIONS ONCIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL 14, NO.2, FEBRUARY 2004" proposes a hierarchical multi-feature coding scheme to facilitate coarse-to-fine matching for efficient palmprint verification in large databases. It defines keypoint distance (Level-1 feature) based on a global geometry basis, global texture energy (Level-2 feature), blurred "interest" lines (Level-3 feature), and local directional texture energy (Level-4 feature). Summary of the Invention

[0010] The purpose of this invention is to provide a biometric authentication device and method that do not suffer from at least some of the disadvantages of existing technologies. In particular, the purpose of this invention is to provide a biometric authentication device and method with reduced computational complexity. Specifically, the purpose of this invention is to provide a biometric authentication device and method capable of performing human authentication using current standard computing devices.

[0011] At least one object of the present invention is achieved by the biometric authentication device and biometric authentication method as defined in the appended independent claims. The dependent claims set forth further embodiments of the invention.

[0012] At least one object of the present invention is achieved by a biometric authentication device for authenticating a person by comparing biometrics of a person's body parts with pre-stored biometrics of body parts of a plurality of registrants, wherein the biometric authentication device is configured to: capture image data of a person's body parts; determine at least one characteristic geometric body part attribute and biometrics of the person's body parts from the captured image data; limit the comparison of the biometrics of the person's body parts to pre-stored biometrics of body parts of a plurality of registrants, the pre-stored biometrics having a predetermined similarity level with respect to at least one characteristic geometric body part attribute of the person. Since the comparison of biometrics is limited to pre-stored biometrics having a predetermined similarity level with respect to at least one characteristic geometric body part attribute, the comparison is limited to a subset of the pre-stored biometrics, thus reducing computational complexity.

[0013] In an embodiment, the biometric authentication device is further configured to capture image data using at least a visible light sensor, a near-infrared light sensor, a combination of visible and near-infrared light sensors, and / or a time-of-flight camera. The visible light sensor, near-infrared light sensor, and / or time-of-flight camera may be included in the biometric authentication device. Biometrics can be determined using image data captured by the visible light sensor and / or the near-infrared light sensor. At least one characteristic geometric body part attribute can be determined using image data captured by the time-of-flight camera. The time-of-flight camera is capable of determining characteristic geometric body part attributes with high accuracy and can precisely determine pre-stored biometrics having a predetermined similarity to at least one characteristic geometric body part attribute, thereby further reducing computational complexity.

[0014] In one embodiment, the biometric authentication device is configured to provide user guidance based on image data captured by a time-of-flight camera.

[0015] In embodiments, user guidance involves moving body parts into a predetermined posture, specifically regarding one or more of the relative distance, relative orientation, and gesture of the body parts. The relative distance and / or relative orientation can be defined relative to the sensor or camera used to capture image data (e.g., a visible light sensor, a near-infrared light sensor, and / or a time-of-flight camera). By combining optical characteristics, such as the horizontal and vertical fields of view of the respective sensor or camera (especially a time-of-flight camera), absolute measurements of the body parts (e.g., the length or width of a finger or the width of a palm) can be determined. This further improves the accuracy of the determined characteristic geometric body part attributes and thus further reduces computational complexity.

[0016] In this embodiment, the body part involves one or more of the palmar side and the back side of the hand. Authentication can be based on vein patterns, palm prints, and / or the life line. Body part-based authentication has many advantages, such as ease of access and accurate authentication.

[0017] In an embodiment, at least one characteristic geometric body part attribute relates to one or more of the sum of the lengths of a set of fingers of the hand, the sum of the average widths of a set of fingers of the hand, and the width of the hand. Experimental studies in men and women of different ages have demonstrated that using the sum of the lengths of a set of fingers of the hand or the sum of the average widths of a set of fingers of the hand, along with the width of the hand, has advantageous properties compared to pre-stored biometrics, such as ease of identification and accessibility.

[0018] In this embodiment, at least one characteristic geometric body part attribute relates to one or more of the left hand, right hand, palmar side of the hand, and back of the hand. Distinction between male and female hands can also be provided. The comparison is further limited, and the computational complexity is further reduced.

[0019] In this embodiment, the similarity level involves the subdivision of at least one feature geometric body part attribute into multiple subsets. This further reduces computational complexity by requiring comparison of only one or a few relevant subsets of biometric features.

[0020] In this embodiment, the subsets overlap with each other, particularly providing a margin for the expected error based on at least one characteristic geometric body part attribute. False rejections of authentication can be avoided, while the increase in computational complexity remains limited.

[0021] At least one object of the present invention is also achieved by a biometric authentication method for authenticating a person by comparing biometric features of a person's body parts with pre-stored biometric features of body parts of multiple registrants. The biometric authentication method includes: capturing image data of a person's body parts; determining at least one characteristic geometric body part attribute and biometric features of the person's body parts from the captured image data; and limiting the comparison of the biometric features of the person's body parts to pre-stored biometric features of body parts of multiple registrants, the biometric features of the body parts of the multiple registrants having a predetermined similarity level with respect to at least one characteristic geometric body part attribute of the person.

[0022] In an embodiment, the biometric authentication method further includes providing at least one of a visible light sensor, a near-infrared light sensor, and a time-of-flight camera to capture image data.

[0023] In one embodiment, the biometric authentication method further includes providing user guidance based on image data captured by a time-of-flight camera.

[0024] In an embodiment, at least one of the characteristic geometric body part attributes relates to the sum of the lengths of a set of fingers of the hand and one or more of the width of the hand.

[0025] In an embodiment, the similarity level involves the subdivision of at least one feature geometric body part attribute into multiple subsets.

[0026] In the embodiments, the subsets overlap each other, particularly with a margin for the expected error based on at least one characteristic geometric body part attribute. Attached Figure Description

[0027] The invention will now be described in more detail with reference to the embodiments illustrated in the accompanying drawings. The drawings show:

[0028] Figure 1 The first person's left hand is shown schematically;

[0029] Figure 2 The second person's right hand is shown schematically;

[0030] Figure 3 The venous network on the dorsal side of a third person's right hand is schematically shown;

[0031] Figure 4 The illustration schematically shows a human hand and a biometric authentication device;

[0032] Figure 5 A time-of-flight camera is shown schematically.

[0033] Figure 6 The characteristic geometric hand properties of the first person's hand are schematically shown;

[0034] Figure 7 The illustration schematically shows the subdivision of the space into subsets defined by the characteristic hand attributes;

[0035] Figure 8 The steps of a biometric authentication method using a subset of characteristic hand attributes are illustrated schematically.

[0036] Figure 9 The steps of a biometric authentication method are illustrated schematically; and

[0037] Figure 10 The steps involved in providing user guidance are illustrated schematically. Detailed Implementation

[0038] Figure 1 The palm side of the left hand 1 of the first person is schematically shown. The left hand 1 has a thumb t, an index finger i, a middle finger m, a ring finger r, and a little finger l. Figure 2 The palm side of the second person's right hand 2 is schematically shown. The right hand 2 has a thumb t, index finger i, middle finger m, ring finger r, and little finger l.

[0039] Figure 1 and Figure 2 Images of the palms of the left hand 1 and right hand 2, captured using a visible light sensor (e.g., 400 nm to 600 nm), are schematically shown. Hands 1 and 2 have palm prints P or life lines that are identifiable under visible light. Additionally or alternatively, the vein patterns of hands 1 and 2 can be determined from image data captured in near-infrared light (e.g., 700 nm to 900 nm). Figure 1 and Figure 2 No vein pattern is shown.

[0040] like Figure 1 and Figure 2As shown, the palm prints or life lines of the hands 1 and 2 of these two individuals include individual biometric features such as specific lengths, positions, curvatures, etc. Specific human authentication can be performed by comparing these features with pre-stored biometric features from body parts of the registrant, particularly by combining biometric features determined from individual vein patterns. Furthermore, human authentication can also be based on biometric features of the back of the hand determined from image data captured in visible light, near-infrared light, or a combination thereof. However, it is currently unknown whether biometric features of the back of the hand determined from image data captured with a visible light sensor are sufficient for human authentication. If relying on the back of the hand, image data captured with a near-infrared light sensor is currently considered necessary for adequate human authentication.

[0041] Figure 3 The dorsal venous network of the right hand (3) of a third person is schematically shown. The right hand (3) has a thumb (t), index finger (i), middle finger (m), ring finger (r), and little finger (l). (See diagram) Figure 3 As shown, the dorsal side of hand 3 includes veins, including a dorsal venous network 31 (dorsal venous network of hand) and dorsal metacarpal veins 32 (dorsal metacarpal joints). The vein pattern can be determined from image data captured using a near-infrared light sensor, and individual biomarkers can be determined from image data captured in near-infrared light.

[0042] Human authentication is based on pre-stored biometric data of a registered user's hand. For each registrant, the database includes biometric data determined from image data captured in visible light, near-infrared light, or combinations thereof. The image data includes image data of the hand. Image data may include image data of the palm, back of the hand, or combinations thereof. Image data may include image data of the left hand, right hand, or combinations thereof. To authenticate a person, image data is acquired from the person's hand. Image data is captured in the visible spectrum, near-infrared spectrum, or combinations thereof. Image data may be captured from the palm side of the hand, from the back side of the hand, or combinations thereof. Image data may be captured from the left hand, from the right hand, or combinations thereof. The current person's biometrics are determined from the captured image data and compared with pre-stored biometrics in the database. If they match, authentication is successful. Otherwise, authentication is rejected. Comparing a person's current biometrics with pre-stored biometrics is computationally complex, especially if the pre-stored biometrics involve, for example, more than 1000 people. However, for usability reasons, a response time of less than 1 second is required for human authentication. The following describes a technique for comparing biometric features that can reduce computational complexity by, for example, 10 to 13 times, thereby enabling sufficiently fast response times using standard computers.

[0043] Figure 4 A biometric authentication device 30 is schematically shown. The biometric authentication device 30 includes a biometric sensor 10 and a processing unit 20. The biometric authentication device 30 can be connected to a user display 40, for example, to provide user guidance. Figure 4 As shown, the processing unit 20 can be attached to the biometric sensor 10. The processing unit 20 can be remotely located within a computing infrastructure, such as a host, server, cloud, etc. The processing unit 20 may include one or more processors and may store computer instructions executable by the one or more processors to enable the functions described in this disclosure. The user display 40 can be fixedly mounted near the biometric sensor 10. The user display 40 can be associated with a user device (e.g., a laptop, smartphone, smartwatch, etc.), wherein the processing unit 20 can communicate with the user display 40 via a wireless connection (e.g., Bluetooth). A biometric authentication device 30 can be included in the user device (e.g., a laptop, smartphone, smartwatch, etc.). Figure 4 As shown, the user's current hand gesture 401 and the user's desired hand gesture 402 can be displayed on the monitor 40.

[0044] The biometric sensor 10 is capable of capturing image data of a person's hand 4. The biometric sensor 10 includes a visible light sensor 101 for capturing image data in the visible spectrum, a near-infrared light sensor 102 for capturing image data in the near-infrared spectrum, and a time-of-flight camera 103 for capturing image data with three dimensions. One or more of the visible light sensor 101, the near-infrared light sensor 102, and the time-of-flight camera 103 may be included in a single sensor. Furthermore, the biometric sensor 10 includes a light source 104. Figure 4 Eight light sources 104 are shown arranged in a circle around sensors 101, 102, and time-of-flight camera 103. The light sources 104 may include different numbers of light sources and / or may be arranged in different ways. The light sources 104 may include one or more light sources that provide illumination in the visible spectrum and allow image data to be captured in the visible spectrum by the visible light sensor 101. The light sources 104 may include one or more light sources that provide illumination in the near-infrared light and allow image data to be captured in the near-infrared light by the near-infrared light sensor 102. Calibration can be provided, particularly with respect to the geometric positions of the visible light sensor 101, near-infrared light sensor 102, and time-of-flight camera 103, such as translational displacement between them. Furthermore, calibration can be provided with respect to scaling factors of the image data captured by the time-of-flight camera 103, such as the absolute size of objects in the captured image data. Calibration can be provided within the biometric sensor 10 or a combination thereof through post-processing in a dedicated computer (e.g., processing unit 20). Calibration can specify that objects in the image data captured by the visible light sensor 101, the near-infrared light sensor 102, and the time-of-flight camera 103 are aligned with each other.

[0045] Visible light sensor 101 may include a visible light-sensitive chip that provides 2D image data (2D: two-dimensional) based on the visible light intensity distribution generated from a 3D scene (3D: three-dimensional). Near-infrared light sensor 102 may include a near-infrared light-sensitive chip that provides 2D image data (2D: two-dimensional) based on the near-infrared light intensity distribution generated from a 3D scene (3D: three-dimensional). Visible light sensor 101 and near-infrared light sensor 102 may include lenses, buffers, controllers, processing electronics, etc. Visible light sensor 101 and near-infrared light sensor 102 may relate to commercially available sensors, such as the e2v Semiconductor SAS EV76C570 CMOS image sensor, which equips visible light sensor 101 with a blocking filter with a wavelength less than 500 nm and near-infrared light sensor 102 with a blocking filter with a wavelength greater than 700 nm, or, for example, the OmniVision OV4686 RGB-lr sensor, wherein visible light sensor 101 and near-infrared light sensor 102 are combined in a single chip and the sensor includes an RGB-lr filter. The light source 104 may include a visible light and / or near-infrared light generator, such as an LED (LED: light-emitting diode). The light source 104 may involve commercially available light sources, such as the high-power LED SMB1N series from Roithner Laser Technik GmbH, Vienna.

[0046] Figure 5 A time-of-flight camera 103 is schematically shown. The time-of-flight camera 103 includes a sequence controller 1031, a modulation controller 1032, a pixel matrix 1033, an analog-to-digital converter 1034 (A / D), an LED or VCSEL 1035 (LED: Light-emitting diode; VCSEL: Vertical-cavity surface-emitting laser), and a lens 1036. The sequence controller controls the modulation controller 1032 and the A / D converter 1034. The modulation controller 1032 controls the LED or VCSEL 1035 and the pixel matrix 1033. The pixel matrix 1033 provides signals to the A / D converter 1034. The sequence controller 1031 interacts with a host controller 1037, for example, via an I2C bus (I2C: I-squared-C serial data bus). The LED or VCSEL 1035 illuminates a 3D scene 1038. After time-of-flight, the lens 1036 receives the light reflected from the 3D scene 1038. The A / D converter 1034 provides raw 3D image data (3D: three-dimensional) to the host controller 1037, for example, via MIPI CSI-2 or PIF (MIPI: Mobile Industrial Processor Interface; CSI: Camera Serial Interface; PIF: Parallel Interface). The host controller performs depth map calculations and provides an amplitude image 103a and a depth image 103d of the 3D scene 1038. Figure 5 As shown, for example, because the wall behind the person is positioned at a specific distance from the time-of-flight camera 103, the background of the amplitude image 103a includes the light and shadow of the wall behind the person, while the background of the depth image 103d has a single value (e.g., black). The time-of-flight camera 103 may be an Infineon camera. TM The company's REAL3 TM It can include specifications such as: direct measurement of depth and amplitude in each pixel; highest accuracy; lean computational load; active modulation of infrared light and patented background illumination suppression (SBI) circuitry in each pixel; full operation under any lighting conditions (darkness and bright sunlight); monocular system architecture without mechanical baseline; minimal size and height for design flexibility; no close-range operation limitations; no special requirements for mechanical stability; no mechanical alignment and angle correction; no risk of recalibration or decalibration due to drops, vibration, or thermal bending; simple and very fast lifetime calibration; and cost-effective manufacturing.

[0047] As will be explained further below, such as Figure 5 The illustrated biometric authentication device 30 provides human authentication by comparing the biometric features of a person's hand 4 with pre-stored biometric features of the hands of multiple registered individuals. The biometric authentication device 30 is configured to capture image data of a person's hand 4, determine at least one characteristic geometric hand attribute Sf, Wh of the person's hand 4 and the biometric features of the person's hand 4 from the captured image data, and define a comparison between the biometric features of the person's hand 4 and the pre-stored biometric features of the hands of multiple registered individuals, wherein the pre-stored biometric features of the hands of the multiple registered individuals and the at least one characteristic geometric hand attribute Sf, Wh of the person have a predefined similarity level.

[0048] The biometric authentication device can be configured to alternately capture image data in the visible spectrum and image data in the near-infrared spectrum. The light sensor 104 can be configured to correspondingly alternately provide illumination in the visible spectrum and illumination in the near-infrared spectrum.

[0049] Figure 6 The schematic illustration shows the characteristic geometric hand attributes of a first person's hand 1, which can be determined based on image data captured, for example, by a time-of-flight camera 103. The image data captured by the time-of-flight camera 103 can be calibrated based on a geometric raster R or grid (…). Figure 6(Seen in dashed lines) to accurately determine the measurement and mapping of the first person's hand 1. Using the grid R, the lengths Li, Lm, Lr, Ll of the fingers can be determined, the width Wh of the hand can be determined, and the widths of the phalanges Wi1, Wi2, Wm1, Wm2, ... can be determined. Therefore, the sum of the lengths of the four fingers (index finger I, middle finger m, ring finger r, and little finger l) Sf = Li + Lm + Lr + Ll and the width Wh of the first person's hand 1 can be determined. In addition, the sum of the average widths of the four fingers of the hand (index finger I, middle finger m, ring finger r, and little finger l) Wf = (Wi1 + Wi2) / 2 + (Wm1 + Wm2) / 2 + ... can be determined.

[0050] like Figure 6 As shown, for each of the fingers I, m, r, l, the grid R or mesh includes reference points ri1, ri2, rm1, rm2, rr1, rr2, rl1, rl2 at the tip and root of each finger. Furthermore, in the direction from the root of the index finger i to the root of the little finger l, the grid R or mesh includes reference points ri, rim, rmr, rrl, rl. Figure 6 In the diagram, reference points ri1, ri2, rm1, ..., ri, rim, rmr, ... are marked with dark circles, but for easier reading, not all reference points include their respective reference symbols. Reference points are determined using image processing techniques (e.g., using the OpenCV function "findContours"). Based on reference points ri1, ri2, rm1, ..., ri, rim, rmr, ..., the raster R or grid can be clearly defined, and the length or width of each finger or hand can be clearly determined.

[0051] Figure 7 The subdivision of the space into subset Ti is schematically illustrated by the characteristic hand attributes having Wh and the sum of the lengths of the four fingers Sf. The hand with Wh is assigned to the first coordinate, or x-coordinate, of the Cartesian coordinate system. The sum of the lengths of the four fingers Sf is assigned to the second coordinate, or y-coordinate, of the Cartesian coordinate system. The units for the x and y coordinates are millimeters [mm]. The values ​​of the width Wh and the sum Sf associated with the registrant's pre-stored biometrics are marked with small black diamonds. The Cartesian coordinate system is subdivided into overlapping subsets Ti, each of which has a rectangular shape. The overlapping portions of subsets Ti are shown in dark gray, while the non-overlapping portions of subsets Ti are marked in light gray. The overlapping regions can be selected based on the expected measurement error used to determine the width Wh and the sum Sf.

[0052] like Figure 7As shown, the hand features Wh and Sf define a two-dimensional interval of 0...18. If there are not just two but n hand features, then the interval is defined as n-dimensional. Each set of biometric features for a specific hand is assigned a corresponding geometric feature attribute. Figure 7 The two-dimensional intervals shown include at least one or at most four intervals for the feature geometric attributes. For known hand attributes, it is necessary to compare the biometric features of each interval. The size of the intervals and thus the number of intervals depends on the reproducibility of the input values ​​(i.e., the hand attributes). Assuming the hand attributes are correlated with biometric features, further optimization can be achieved by ranking candidates based on their distances and starting with the minimum distance, rather than performing a "brute-force" match within such intervals.

[0053] The width Wh and sum Sf of the person requiring authentication are marked with large white diamonds. Due to measurement errors (which make human authentication difficult based solely on width Wh and sum Sf), the large white diamonds do not match one of the black diamonds (associated with the registrant's pre-stored biometrics). Therefore, authentication based on biometric comparison is required. However, to authenticate a person, it is sufficient to compare the person's biometrics with the pre-stored biometrics of the person associated only with the corresponding subset Ti. Depending on the position of the white diamonds, the comparison is limited to a single subset Ti, or at most four subsets Ti. Figure 7 In the example shown, the comparison is limited to two subsets Ti, which include a width Wh of 70 mm and a sum Sf of 260 mm and 300 mm. Therefore, human authentication is limited to pre-stored biometrics of the hands of multiple registrants, having a predetermined similarity level with respect to at least one feature geometric hand attribute (i.e., width Wh and sum Sf). Consequently, the computational complexity for comparing biometrics is reduced (e.g., by a factor of 10 to 13), enabling sufficiently fast response times using standard computers.

[0054] In addition to the sum of the lengths of the four fingers (Sf) and the width of the hand (Wh), the feature geometry of the hand can be related to the sum of the average widths of the four fingers (Wf). Feature geometry hand attributes can include one or more of the sum Sf, width (Wh), and sum Wf. Therefore, as... Figure 7 The subdivisions shown can involve one-dimensional space, two-dimensional space, or three-dimensional space.

[0055] Besides the total length of the four fingers (Sf) and the width of the hand (Wh), the geometric hand attributes can be related to the curvature of the hand and the position of the thumb. The curvature of the hand can determine whether the image data is related to the palm or the back of the hand. Furthermore, the position of the thumb can determine whether the image data is related to the left or right hand. Therefore, the comparison of biometric features can be limited to corresponding pre-stored biometric features, such as pre-stored biometric features related to the back of the left hand, pre-stored biometric features related to the palm of the left hand, etc. This reduces the computational complexity of comparing biometric features.

[0056] Return to reference Figure 4 The biometric authentication device 30 can provide user guidance based on image data captured by the time-of-flight camera 103. The user guidance may involve moving the hand 4 into a predetermined posture. The predetermined posture may involve a posture relative to the biometric sensor 10. Figure 4 As shown, the current hand posture 401 and the desired hand posture 402 of the user can be displayed on the display 40. The size of the displayed current hand posture 401 can provide the user with guidance regarding the desired relative distance from the biometric sensor 10. For example, if the size of the displayed current hand posture 401 is smaller than the size of the displayed desired hand posture 402, the user is guided to move their hand 4 closer to the biometric sensor 10. Similar guidance regarding the relative orientation of the hand 4 can be provided. The relative orientation of the hand 4 can be related to the tilt of the palm or back of the hand relative to the plane defined by the biometric sensor 10. Similar guidance regarding hand gestures can be provided. Hand gestures can be related to finger spread, finger extension, etc.

[0057] Figure 8 The steps of a biometric authentication method for authenticating a person by comparing the biometric features of a person's hand 4 with pre-stored biometric features of the hands of multiple registered persons are illustrated schematically. In step S1, image data of the person's hand 4 is captured. In step S2, at least one characteristic geometric hand attribute Sf, Wh of the person's hand 4, and the biometric features of the person's hand 4 are determined from the captured image data. Specifically, based on the combination... Figure 7 The number of subsets is determined by the defined interval. In step S3, the comparison of the biometric features of a person's hand 4 is limited to the pre-stored biometric features of the hands of multiple registered persons, which have a predetermined similarity level with at least one characteristic geometric hand attribute Sf, Wh of the person, specifically based on one or more determined subsets.

[0058] Figure 9The building blocks of the algorithm chain used for authenticating a person are schematically illustrated. In step Sa1, image data is acquired using a visible light sensor 101 and a near-infrared light sensor 102, wherein the visible light sensor 101, the near-infrared light sensor 102, the time-of-flight camera 103, and the light source 104 are controlled respectively. The image data may be stored in shared memory or a file system. In step Sa2, the background in the image data acquired from the visible light sensor 101 and the near-infrared light sensor 102 is removed. In step Sa3, a region of interest (ROI) is determined in the image data acquired from the visible light sensor 101 and the near-infrared light sensor 102. In step Sa4, image enhancement or filters are applied to the image data acquired from the visible light sensor 101 and the near-infrared light sensor 102. The enhancement or filters may be related to unsharpening masks, adaptive thresholding, Laplacian, etc. In step Sa5, biometric features are extracted from the image data acquired from the visible light sensor 101 and the near-infrared light sensor 102. Image feature extraction can be related to ORB (Oriented Fast and Rotated BRIEF), SIRF (Scale Invariant Feature Transform), SURF (Speed ​​Robust Feature Transform), etc. In step Sa6, biometric matching is performed based on biometric features extracted from image data acquired from visible light sensor 101 and near-infrared light sensor 102. Biometric matching can involve BF (Brute Force Matching), FLANN (Fast Nearest Neighbor Search), etc. In step Sa7, the individual's identity and the terminal's identity are verified, and a yes / no decision is provided accordingly. Depending on the type of feature descriptor, Euclidean distance (or L2 norm) is typically used to compare and match n-dimensional feature vectors, while Hamming distance is used to compare and match binary-coded features.

[0059] Figure 10The steps involved in providing user guidance, which involves guiding a person to place their hand 4 in a desired posture relative to the biometric authentication device 30, are illustrated schematically. In step Sg1, the time-of-flight camera 103 verifies whether the subject is within, for example, a distance of 40 cm. If not, step Sg1 is repeated after a predetermined delay, for example, preferably 200 milliseconds. In step Sg2, the time-of-flight camera 103 verifies whether the subject has the shape of a hand. If not, step Sg1 is repeated after a predetermined delay, for example, between one and five seconds (preferably three seconds). In step Sg3, the absolute size of the hand is determined based on image data captured by the time-of-flight camera 103. In step Sg4, a representation of the hand at a desired location (optimally for the visible light sensor 101 and the near-infrared light sensor 102) is displayed on the display 40, at a predetermined distance, for example, approximately k cm, where k is within the depth of focus area of ​​the visible light sensor 101 and / or the near-infrared light sensor 102. The representation of the hand is shown with the fingers slightly spread. In step Sg5, the current position of the hand is continuously determined using the time-of-flight camera 103, and a representation of the current position of the hand is continuously displayed on the display 40, particularly along with a superimposed level indicating the rotation or tilt of the hand relative to the desired position. In step Sg6, it is verified whether the hand is in the desired position, particularly with respect to the desired predetermined distance from the visible light sensor 101 and the near-infrared light sensor 102. If not, step Sg5 continues. Otherwise, in step Sg7, the representation of the hand at the desired position is changed to a hand with fingers spread (i.e., a hand with the desired gesture), and the representation of the hand at the current position is continuously displayed based on the image data captured by the time-of-flight sensor 103. In step Sg8, if the hand is at an optimal distance (e.g., within ±3 cm), and if the hand is within the optimal rotation range (e.g., within ±12°) and the optimal tilt range (e.g., within ±12°), for example according to the combination Figure 9 The specified steps involve using a visible light sensor 101 and a near-infrared light sensor 102 to capture image data for user authentication. The user guidance ends here. If image data cannot be captured in step Sg8, the user guidance is interrupted after a predetermined time period (e.g., preferably four seconds), whereby corresponding information is displayed on the display 40, and the user guidance restarts in step Sg1.

[0060] List of reference numerals

[0061] 1, 2, 3 The hands of the first, second, and third people

[0062] t, i, m, r, l: thumb, index finger, middle finger, ring finger, little finger

[0063] P Palm print or life line

[0064] 31, 32 Dorsal venous network of the hand, dorsal palmar bony veins

[0065] 4 people's hands

[0066] 10 Biometric Sensors

[0067] 101 Visible Light Sensor

[0068] 102 Near-infrared light sensor

[0069] 103 Time-of-Flight Cameras

[0070] 104 Light Source

[0071] 20 processing units

[0072] 30 Biometric authentication devices

[0073] 40 monitors

[0074] 401, 402 The user's current hand posture, the user's desired hand posture

Claims

1. A biometric authentication device (30) for authenticating a person by comparing a biometric feature of a body part (4) of the person with pre-stored biometric features of body parts of a plurality of registered persons, the biometric authentication device (30) comprising: a biometric sensor configured to capture image data of a body part (4) of the person; and a processing unit configured to determine from the captured image data: at least one characteristic geometric body part attribute of the body part (4) of the person, the at least one characteristic geometric body part attribute comprising a width of a hand of the person, and a biometric feature of the body part (4) of the person; wherein the processing unit is further configured to either accept or reject the authentication of the person by: using a space defined by the at least one characteristic geometric body part attribute, the space being subdivided into a plurality of pre-defined value subsets for the at least one characteristic geometric body part attribute, and comparing the biometric feature of the body part (4) of the person only with the pre-stored biometric features of the body parts of each registered person whose at least one characteristic geometric body part attribute is in the same pre-defined value subset as the at least one characteristic geometric body part attribute of the person.

2. The biometric authentication device (30) according to claim 1, wherein the biometric sensor comprises at least one of a visible light sensor (101), a near-infrared light sensor (102), a combined visible light sensor and near-infrared light sensor, and a time-of-flight camera (103).

3. The biometric authentication device (30) according to claim 1, wherein the biometric sensor comprises a time-of-flight camera (103), and the processing unit is further configured to provide user guidance based on the image data captured with the time-of-flight camera (103).

4. The biometric authentication device (30) according to claim 3, wherein The user guidance involves moving the body part (4) into a predetermined pose.

5. The biometric authentication device (30) according to claim 3, wherein The user guidance involves moving the body part (4) into a predetermined pose, the predetermined pose being related to one or more of a relative distance, a relative orientation, and a hand gesture of the body part (4) of the person.

6. The biometric authentication device (30) according to any one of claims 1-5, wherein, The body part (4) involves one or more of a palm side of a hand and a back side of a hand.

7. The biometric authentication device (30) according to any one of claims 1-5, wherein, The at least one characteristic geometric body part attribute further comprises one or more of a sum of lengths of a set of fingers of the hand and a sum of average widths of the set of fingers of the hand.

8. The biometric authentication device (30) according to any one of claims 1-5, wherein, The at least one characteristic geometric body part attribute involves one or more of a left hand, a right hand, a palm side of a hand, and a back side of a hand.

9. The biometric authentication device (30) according to claim 1, wherein The pre-defined value subsets overlap each other with a margin according to an expected error of the at least one characteristic geometric body part attribute.

10. A biometric authentication method for authenticating a person by comparing a biometric feature of a body part (4) of the person with pre-stored biometric features of body parts of a plurality of registered persons, the biometric authentication method comprising: capturing image data of a body part (4) of the person via a biometric sensor; determining from the captured image data, via a processing unit: at least one characteristic geometric body part attribute of the body part (4) of the person, the at least one characteristic geometric body part attribute comprising a width of a hand of the person, and a biometric feature of the body part (4) of the person; biometric features of the person's body part (4); and authenticating the person by or rejecting the person via the processing unit by: using a space defined by the at least one characteristic geometric body part attribute, the space being subdivided into subsets for a plurality of predefined values of the at least one characteristic geometric body part attribute, and comparing the biometric features of the person's body part (4) only with pre-stored biometric features of body parts of each registered person whose at least one characteristic geometric body part attribute is in the same subset of predefined values as the at least one characteristic geometric body part attribute of the person.

11. The biometric authentication method of claim 10, wherein the biometric sensor comprises at least one of a visible light sensor (101), an infrared light sensor (102), and a time-of-flight camera (103) to capture image data.

12. The biometric authentication method of claim 10 or 11, wherein the biometric sensor comprises a time-of-flight camera (103), the method further comprising providing user guidance via the processing unit based on image data captured with the time-of-flight camera (103).

13. The biometric authentication method according to claim 10 or 11, wherein, The at least one characteristic geometric body part attribute further comprises one or more of a sum of lengths of a set of fingers of a hand and a sum of average widths of a set of fingers of a hand.

14. The biometric authentication method according to claim 10, wherein The subsets of predefined values overlap each other with a margin according to an expected error of the at least one characteristic geometric body part attribute.

Images