Biometric device with ultrasonic transducer for fingerprint and microvasculature detection

Abstract

The invention relates to a biometric device comprising ultrasound transducer circuitry, control circuitry and data processing circuitry configured to implement:-acquiring a surface image of an element arranged against a capture surface and a series of volumetric images of the element determined from a global volumetric image (114) of its microvasculature; -determining a feature of a fingerprint of the element from the surface image and a feature of a capillary of the element from the global volumetric image (114); comparing the previously determined feature with an expected theoretical feature in order to identify or not identify the element.

Description

[0001] This application claims priority to French patent application No. 23 / 10033, filed on September 22, 2023, entitled "Biometric identification device with ultrasonic transducer for fingerprint and microvasculature detection", which is incorporated herein by reference to the fullest extent permitted by law. Technical Field

[0002] This specification relates to the field of biometric devices based on capture and detection using ultrasonic transmission. Background Technology

[0003] Biometric devices can verify or identify a user based on measurements of one or more of their biometric features, such as fingerprints, facial shape, iris patterns, retinal patterns, etc. Various techniques for measuring one or more of these features, depending on different criteria, each have advantages and disadvantages, including: more or less significant error rates, the ease of sealing and comparing the template to the measurement, the ability to detect imitation or lack thereof, proof of activity, ease of use and comfort, required sensor size, and required power consumption.

[0004] Known devices use fingerprint sensors for biometric identification, anti-spoofing, and liveness detection. The fingerprint sensor performs optical or capacitive measurements to verify whether the captured fingerprint corresponds to the expected fingerprint (authentication phase). Liveness detection and anti-spoofing algorithms are then implemented via image processing performed from the captured fingerprint or by using an infrared system. For example, it can detect whether the finger set on the device during fingerprint capture is a fake finger. Summary of the Invention

[0005] There is a need to propose a biometric device that does not have some of the drawbacks of existing solutions.

[0006] One embodiment overcomes all or part of the drawbacks of known solutions and proposes a biometric device comprising at least one ultrasonic transducer circuit, a control circuit, and a data processing circuit, configured to perform at least the following steps:

[0007] - Acquire at least one surface image of at least one element, said at least one element being configured to abut against the capture surface of the device; and at least a series of volume images of said element, from which a global volume image of the microvascular system of said element is determined;

[0008] - Determine the fingerprint features of the element from the surface image, and determine the features of the microvascular system of the element from the global volume image;

[0009] - The previously determined features (that is, those determined in the previous steps) are compared with expected theoretical features (such as features calculated during the initial steps of user registration) to identify or not identify the element.

[0010] According to a particular embodiment, the data processing circuit is configured to determine the global volume image by implementing a method for spatiotemporal filtering of a series of volume images of the element.

[0011] According to a particular embodiment, the ultrasonic transducer circuit is configured to perform the acquisition of a series of volumetric images of the element via pulse-wave Doppler imaging.

[0012] According to a particular embodiment, the data processing circuit is configured to perform a step of aligning the surface image with the global volume image between the step of acquiring the surface image and the series of volume images and the step of determining the fingerprint and vascular system features of the element.

[0013] According to a particular embodiment, the data processing circuit is configured such that the step of determining the fingerprint and the characteristics of the microvascular system of the element includes at least:

[0014] - Determine the details of the fingerprint of the element from the surface image, and determine the point of interest of the microvascular system of the element from the global volume image;

[0015] - Calculate the location of at least one detail in the surface image and / or at least one point of interest in the global volume image, and calculate the geometric relationship between the at least one detail and / or the at least one point of interest and at least some of the other details and points of interest, taking into account the depth of the point of interest in the global volume image;

[0016] The comparison step includes comparing the previously calculated position and geometry with the expected theoretical position and geometry to identify or not identify the element.

[0017] According to a specific embodiment, the point of interest corresponds to the bifurcation of a microvessel and / or the point of strong curvature of a microvessel.

[0018] According to a particular embodiment, the data processing circuit is configured to perform a step of combining the determined details and points of interest in the form of a point cloud between the step of determining the details and the points of interest and the calculation phase.

[0019] According to a particular embodiment, the data processing circuit is configured such that during the calculation step, the additional details and points of interest used to calculate the geometric relationships correspond to the details included in a region of the surface image defined by a planar geometry, and the points of interest included in a region of the global volume image included in a volume formed by the planar geometry, which extends in a dimension perpendicular to the surface image of the global volume image.

[0020] According to a particular embodiment, the planar geometry corresponds to a polygon, a circle, or an ellipse.

[0021] According to a particular embodiment, the data processing circuit is configured to repeat the calculation steps for each determined detail and each determined point of interest, and to perform the comparison steps for the calculated positions and geometric relationships for each detail and each point of interest.

[0022] According to a particular embodiment, the data processing circuit is configured such that the geometric relationship calculated between a first detail or point of interest and a second detail or point of interest includes the distance between the first detail or point of interest and the second detail or point of interest, the orientation of the first detail or point of interest relative to the second detail or point of interest, and the orientation of the second detail or point of interest relative to the first detail or point of interest.

[0023] According to a particular embodiment, the data processing circuit is configured such that for a point of interest corresponding to a bifurcation or strong bend in a microvessel, the calculated geometric relationship includes the orientation of each of the microvessels relative to the other details and the point of interest.

[0024] According to a specific embodiment, the data processing circuit is configured as follows:

[0025] The determination of the fingerprint and the features of the microvessels of the element includes extracting at least one feature vector from the surface image and the global volume image by means of the embodiment of the deep learning algorithm;

[0026] Comparing previously determined features with expected theoretical features to identify or not identify the element includes comparing the previously determined feature vector with at least one expected theoretical feature vector using the deep learning algorithm.

[0027] According to a particular embodiment, the data processing circuit is configured such that the comparison step includes the calculation of a score whose value depends on the correlation between the determined features used to identify or not identify the element and the expected theoretical features.

[0028] According to a particular embodiment, the data processing circuit is configured to calculate a score for each determined detail and each determined point of interest during the comparison step. Attached Figure Description

[0029] These features and advantages, as well as other features and advantages, will be disclosed in detail in the following description of specific embodiments given by way of non-limiting example with reference to the accompanying drawings, wherein:

[0030] - Figure 1 and Figure 2 A biometric device according to a specific embodiment is illustrated schematically;

[0031] - Figure 3 The steps performed by a biometric device according to a specific embodiment are illustrated in schematic form;

[0032] - Figure 4 The illustration schematically shows details identified in a surface image captured by a biometric device according to a particular embodiment;

[0033] - Figure 5 An area according to a particular embodiment is schematically shown, which defines the details and points of interest considered during identification performed by a biometric device;

[0034] - Figure 6 This illustration shows an example of the calculation results obtained when the user's identity is confirmed by a biometric device;

[0035] - Figure 7 This illustration shows a first example of the calculation results obtained when the user's identity is not verified by the biometric device;

[0036] - Figure 8 This illustration shows a second example of the calculation results obtained when the user's identity is not verified by the biometric device. Detailed Implementation

[0037] Identical components are designated by the same reference numerals in different drawings. In particular, common structural and / or functional elements in different embodiments may have the same reference numerals and may have the same structure, dimensions, and material properties.

[0038] For clarity, only steps and elements that aid in understanding the embodiments are shown and described in detail. Specifically, various elements (ultrasonic transducer circuits, control circuits, processing circuits, etc.) and various implementation steps (processing of acquired images, feature determination, details of performed calculations, etc.) are not described in detail. Those skilled in the art will be able to perform these elements in detail based on the functional description provided herein.

[0039] Unless otherwise stated, when referring to two elements connected to each other, it means that there is no direct connection between them except for a conductor; when referring to two elements coupled to each other, it means that the two elements can be connected or coupled through one or more other elements.

[0040] In the following description, unless otherwise stated, when referring to absolute position qualifiers, such as the terms “front,” “back,” “up,” “down,” “left,” “right,” etc., or relative position qualifiers (such as the terms “above,” “below,” “higher,” “lower,” etc.), or orientation qualifiers (such as the terms “horizontal,” “vertical,” etc.), refer to the orientation of the graphic in the normal operating position.

[0041] Unless otherwise stated, the expressions “about,” “probably,” “basically” and “approximately” indicate within 10%, preferably within 5%.

[0042] The following is combined Figure 1 and Figure 2 A biometric device 100 according to a specific embodiment is described.

[0043] exist Figure 1 In the example shown, device 100 is configured to perform biometric identification of an element 101 from a user of device 100, which corresponds to, for example, one or more fingers of the user. Here, element 101 corresponds to an element or part of the user's body. Device 100 includes a capture surface 102 on which the element 101 of the user of device 100 is intended to be positioned during identification.

[0044] Device 100 includes at least one ultrasonic transducer circuit 104, a control circuit 106, and a data processing circuit 108. Figure 2 In the illustrative example shown, control circuit 106 can be electrically coupled to ultrasonic transducer circuit 104, thus enabling it to send signals to ultrasonic transducer circuit 103 to control the ultrasonic emission for the measurement to be performed, and also to receive electrical measurement signals sent by ultrasonic transducer circuit 104. Control circuit 106 can also be coupled to data processing circuit 108 to send measurement results obtained from ultrasonic transducer circuit 104 to data processing circuit 108.

[0045] Circuit 104 may include multiple ultrasonic transducers, thereby enabling image acquisition required for biometrics. For example, the ultrasonic transducers may be arranged in a matrix or in another manner suitable for the capture to be performed. The number of ultrasonic transducers in circuit 104 may depend on the size of the capture surface 102.

[0046] In the exemplary embodiment described, device 100 may be configured to acquire at least one surface image of one or more fingers of a user, the fingers being positioned on and in contact with the capture surface 102 of device 100; that is, to acquire at least one image of at least one fingerprint formed by ridges and valleys present on the skin surface of one or more fingers positioned on the capture surface 102. In device 100, the acquisition of surface images of one or more fingers of a user positioned on and in contact with the capture surface 102 is based on the fact that ultrasound emitted by transducer circuit 104 is more clearly reflected against air present in the valleys of the fingerprint of one or more fingers than against the ridges of the fingerprint.

[0047] Furthermore, device 100 can also be configured to acquire volumetric images of one or more fingers disposed on and in contact with the capture surface 102 of device 100, and from which a global volumetric image of the microvascular system of that / those fingers can be determined by processing these volumetric images; that is, an image of the microvessels of one or more fingers of the user of device 100 is captured. Therefore, the obtained global volumetric image can be obtained from the non-zero depth (along the...) of one or more fingers present on the capture surface 102. Figure 1 The dimensions of the Z-axis, which are visible in the image, are inferred from multiple volumetric images captured consecutively at a location substantially perpendicular to the capture surface 102.

[0048] For example, the ultrasonic transducer circuit 104 can be configured to perform volumetric image acquisition of one or more fingers placed on the capture surface 102 via pulse-wave Doppler imaging. In this configuration, the transducer circuit 104 is configured to emit a series of ultrasonic pulses. The response obtained in this configuration does not correspond to a change in the frequency of the received waves, but rather to a pseudo-Doppler effect, that is, a change in the time interval between received echoes relative to the time interval between emitted waves. Therefore, in device 100, pulse-wave Doppler imaging can be used to determine the spatial location of the microvascular system motion in the acquired volumetric image and, from there, to determine a microvascular system map in the global volumetric image.

[0049] For example, the ultrasonic transducer circuit 104 may include a transducer of the CMUT (“capacitive micromechanical ultrasonic transducer”) type or PMUT (“piezoelectric micromechanical ultrasonic transducer”) type, or another type of transducer.

[0050] In the exemplary embodiments described herein, the control circuit 106 may be configured to provide an electrical excitation signal to the ultrasonic transducer circuit 104, causing the ultrasonic transducer circuit 108 to emit ultrasonic waves and receive an electrical response signal generated by the ultrasonic transducer circuit 104 in response to receiving ultrasonic waves reflected by elements 101 present on the capture surface 102.

[0051] In the described exemplary embodiment, the data processing circuit 108 may be configured to analyze and process electrical response signals received by the control circuit 106 and transmitted by the ultrasonic transducer circuit 104. The processing circuit 108 may include, for example, a microprocessor coupled to a memory to perform processing on the received data.

[0052] The control circuit 106 and the data processing circuit 108 can be configured to implement the following description and Figure 3 The steps are represented in a graphical form to perform biometric identification.

[0053] During the first step 202, the device 100 may perform detection of elements present on the capture surface 102. If an element is indeed detected on the capture surface 102, the second step 204 may be implemented. If no element is detected, the detection step 202 may be repeated until an element is detected on the capture surface 102. The detection step 202 may be repeated continuously or at regular or irregular time intervals.

[0054] The device 100 can then be configured to acquire at least one surface image of the element 101 present on the capture surface 102 (step 204). For example, such a configuration of the device 100 may include sending an acquisition command or instruction sequence to the control circuitry 106 to acquire the surface image of the element 101.

[0055] During subsequent step 206, an acquisition of at least one surface image of the element 101 present on the capture surface 102 can then be performed. This acquisition may include the transducer circuit 104 emitting a series of ultrasonic signals in pulse form, receiving the echoes by the transducer circuit 104, and processing the obtained response to obtain a surface image of the element 101, i.e., an image of one or more fingerprints of the user.

[0056] The obtained response is processed to obtain a surface image of element 101, for example, signal processing corresponding to the type of "beamforming" or "microwave beamforming," also known as beamforming or spatial filtering. Alternatively, the transmission and reception of signals from transducer circuit 104 can be performed by multiplexing to perform the acquisition of a B-mode image focused on the surface of element 101. According to another variation, the transducer matrix of circuit 104 can be orthogonally addressed to transmit a series of ultrasonic pulses in the form of focused waves or plane waves, or to perform full matrix acquisition of the type RCA-OPW ("Row and Column Addressed Orthogonal Plane Wave Imaging"). For example, the paper "So you think you can DAS? A viewpoint on delay-and-sum beamforming" by V. Perrot et al., Ultrasonics 111 (2021), 106309, arXiv:2007.11960, describes an example of a DAS (delay-and-sum) beamforming method that can be used to obtain a surface image of element 101.

[0057] The acquisition of the surface image of element 101 may also include other steps not described in detail here, such as filtering, envelope detection, logarithmic compression, etc., of the acquired response signal.

[0058] The acquired surface image may correspond to the capture of a fingerprint of element 101, for example, one or more fingerprints when element 101 corresponds to one or more fingers of a user. Data associated with the acquired image is stored, for example, in the memory of device 100, or in external memory of device 100 coupled to device 100 via a communication link.

[0059] After acquiring the surface image of element 101, device 100 can be configured to continuously acquire multiple volumetric images of element 101 present on capture surface 102 over time (step 208). For example, such configuration of device 100 may include sending acquisition commands or instruction sequences to control circuitry 106 for acquiring volumetric images of element 101.

[0060] During subsequent step 210, multiple consecutive volumetric images of element 101 present on capture surface 102 can be acquired, and these volumetric images can then be processed to highlight displacement or motion associated with blood flow in the microvascular system of element 101, ultimately obtaining a global volumetric image of the microvascular system of element 101. The processing performed on the volumetric images can correspond to a spatiotemporal filtering or “clutter filtering” implementation using a high-pass filter in the slow time domain after signal demodulation. Alternatively, estimation of blood flow-related motion in the microvascular system can be performed by correlating motion between consecutive volumetric images, or by SVD decomposition (“singular value decomposition”) and excluding components including tissue motion and noise. An example of such a spatiotemporal filtering implementation is described in the paper “Spatiotemporal Clutter Filtering of Ultrafast Ultrasound Data Highly Increases Doppler and Ultrasound Sensitivity” by Demené C et al., IEEE Trans Med Imaging. November 2015, 34(11), 2271-85.

[0061] For example, volumetric image acquisition can be performed by transducer circuit 104 emitting planar ultrasound waves at multiple angles. Processing of the response signals can then include a beamforming stage, followed by coherent summation of the images corresponding to each angle to obtain a B-mode volumetric image. This process can be repeated over time to obtain multiple volumetric images, allowing monitoring of blood flow in the microvascular system of element 101, and ultimately obtaining a global volumetric image of the microvascular system of element 101 after processing the volumetric images. As an example, the paper "Coherent plane-wave compounding for very high frame rate ultrasonography and transient elastography" published by G. Montaldo et al. in the March 2009 issue of IEEE Transactions on Ultrasound, Ferroelectricity and Frequency Control, Vol. 56, No. 3, pp. 489-506, describes an example of a DAS beamforming method with coherent summation of plane waves, which can be used to obtain a global volumetric image of the microvascular system of element 101.

[0062] Therefore, the microvascular system of element 101 is determined by first identifying the movement of tissue and arterial walls due to blood flow, thanks to a series of acquired volumetric images, which allows for the localization of the microvascular system elements.

[0063] Data associated with the obtained global volumetric image can be stored in the memory of device 100, or in an external memory of device 100, for example, coupled to device 100 via a communication link.

[0064] In the embodiments described in particular, during subsequent step 212, the previously acquired surface images and global volume images can be aligned with each other, for example, by repositioning them according to a common coordinate system defined by the geometry of the transducer circuit 104.

[0065] During the subsequent step 214, the data processing circuit 108 can determine or extract details of the fingerprint of element 101 obtained from the previously acquired surface image. This detail determination can be performed by one or more image processing algorithms, which are not described in detail here but are known to those skilled in the art. R. Bansal et al. described various methods for extracting details from fingerprint images that can be used herein in “Minutiae Extraction from Fingerprint Images - a Review”, Volume 8, Issue 5, No. 3, IJCSI International Journal of Computer Science, September 2011.

[0066] During subsequent step 216, the data processing circuit 108 can determine or extract points of interest (POIs) of the microvascular system of one or more fingers present on the capture surface 102 from the previously acquired global volumetric image. In the described exemplary embodiment, these POIs of the microvessels correspond to bifurcations and / or strong bends or inflections of the microvessels. This determination of POIs can be performed by one or more image processing algorithms, for example, similar to those implemented for previously determining fingerprint details. Alternatively, the algorithm implemented for determining POIs may include the following steps:

[0067] - Segment the global volume image, for example by the properties of the Hessian matrix in each voxel of the global volume image, or by using a convolutional neural network or other methods;

[0068] - Skeletonize (or “thin”) the segmented image to reduce the thickness of elements (such as blood vessels) present in the image to voxels;

[0069] - Detect forks and / or strong inflection points;

[0070] - For example, principal component analysis (PCA) can be used to characterize the orientation of each point of interest.

[0071] According to another variant, the algorithm for determining points of interest may include the use of neural networks, which can directly detect points of interest and their type (bifurcation or strong bending) and their orientation.

[0072] Alternatively, surface image acquisition can be performed after volumetric image acquisition. Furthermore, detail determination in the surface image can be performed before or during volumetric image acquisition, or between volumetric image acquisition and the processing of these images.

[0073] In this particular embodiment, during subsequent step 218, a step of combining the previously determined details and points of interest in the form of a point cloud is performed. This combining step may include placing the fingerprint details and the points of interest of the microvascular system in the same coordinate system linked to the geometry of the transducer circuit 104. Thus, the fingerprint details can be placed in a shallow or zero-depth (Z=0) (X, Y) plane, and the points of interest of the microvascular system can be located in a volume below the volume containing the details. This combining of data from previously acquired images is implemented, for example, by the data processing circuit 108.

[0074] Then, the data processing circuit 108 can perform the calculation of the location of at least one point of interest in the surface image and / or at least one point of interest in the global volume image, and for the point of interest, take into account the depth of the point of interest in the global volume image (along...). Figure 1 The data processing circuit 108 calculates the geometric relationship between the at least one detail and / or the at least one point of interest and at least some other details and points of interest (Z-axis dimensions). In a particular embodiment, the data processing circuit 108 is configured such that during this calculation step, the other details and points of interest used to calculate the geometric relationship correspond to details included in a surface image region defined by a planar geometry (e.g., a polygon, circle, or ellipse), and points of interest included in a global volume image region defined by the volume generated by the planar geometry and extending in the depth direction (parallel to the Z-axis) of the global volume image.

[0075] The following is combined Figure 4 and Figure 5 An example of this calculation is described.

[0076] exist Figure 4 In this context, an example of a surface image acquired by device 100 is indicated and specified by reference numeral 110. This surface image 110 corresponds herein to a fingerprint present on the capture surface 102 of device 100. Details determined in this surface image 110 are represented by dots and specified by reference numerals 112 and 113.

[0077] Examples of global volumetric images acquired by device 100 include... Figure 5 As shown, and specified by reference numeral 114. Figure 5In this image, the result corresponds to the maximum intensity projection (MIP) obtained by means of the finger microvascular system present on the capture surface 102 of the device 100 along the Z-axis. The point of interest determined in this image is represented by the point specified by reference numeral 116.

[0078] In the example described here, one of the details or one of the points of interest 116 can then be selected, and then superimposed on each other, including the region of the surface image 110 of that detail or that point of interest 116 and the region of the global volume image 114, on one side bounded by the planar geometry in the surface image 110 and the volume formed by the planar geometry in the global volume image 114. Figure 4 and Figure 5 In the example, one of the details specified by reference numeral 113 is selected, and the planar geometry is defined as a circle whose center corresponds to detail 113. The calculation steps are performed by considering other details 112 included in the geometry and points of interest 116 included in the volume 120 formed by projecting the geometry along the Z-axis. That is, along the thickness direction of element 101. Figure 5 The obtained volume 120 is schematically represented, and the region of one or more fingers is defined, in which details 112, 113 and point of interest 116 are considered for this calculation. Figure 5 In the example, the volume 120 is formed into a cylinder that extends through at least a portion of the thickness of one or more fingers present on the capture surface 102.

[0079] Therefore, details 112, 113 and point of interest 116 are projected into a common 3D space from which the following computational steps are performed.

[0080] The position of detail 113 in surface image 110 can be calculated, and the geometric relationships between detail 113 and other details 112 and points of interest 116 included in volume 120 can also be calculated. In the described example, these geometric relationships may include:

[0081] - The distance between detail 113 and the points of interest 116 included in each of the other details 112 and volume 120, and

[0082] - The orientation of detail 113 relative to other details 112 and the point of interest 116 included in volume 120, and

[0083] - The orientation of the point of interest 116 included in other details 112 and volume 120 relative to detail 113.

[0084] The calculation can be performed by taking into account all details 112 and points of interest 116 present in the volume, the region of which corresponds to the selected geometry and extends to at least a portion of the thickness of one or more fingers present on the capture surface 102.

[0085] In one particular embodiment, the data processing circuitry 108 is configured to repeat the above calculation steps for each previously determined detail 112 and each previously determined point of interest 116. When the calculation is performed on the point of interest 116 corresponding to the bifurcation or high curvature of a microvessel, the calculated geometric relationships may include the orientation of each microvessel relative to the other details 112 and points of interest 116.

[0086] For the calculations described above, the positions of details 112, 113 and point of interest 116, as well as the geometric relationships between them, are calculated in a Cartesian or spherical coordinate system.

[0087] The implementation of these / these calculation steps corresponds to Figure 3 Step 220 in the diagram.

[0088] Then, for example during the initial user registration step, the locations and geometric relationships of one or more previously calculated details 112, 113 and points of interest 116 are compared with one or more previously calculated locations and geometric relationships to confirm or not confirm the identity of the user of device 100 (step 222). This comparison can be made for the locations and geometric relationships calculated for each detail 112 and each point of interest 116.

[0089] Figure 6 An example is schematically illustrated where the calculated positions and geometries correspond to those expected positions and geometries, thereby confirming that the user identity of device 100 is indeed the expected identity. In this figure, reference numeral 122 indicates the expected positions and geometries between detail 124 and point of interest 126, which are similar to the positions and geometries calculated for details 112, 113 and point of interest 116 of a finger present on the capture surface 102 of device 100.

[0090] Figure 7 Another example is illustrated schematically, where the calculated positions and geometric relationships do not correspond to the positions and geometric relationships expected to confirm the user's identity with device 100. In this example, the determined details 112, 113, their positions, and the geometric relationships between them correspond to the expected details, but the point of interest 116 is not determined (and therefore no geometric relationship with the point of interest is calculated). For example, such an example corresponds to the case where a dummy finger is placed on the capture surface 102 of device 100. In this case, device 100 may consider that the user's identity has not been confirmed.

[0091] Figure 8 Another example is illustrated schematically, where the calculated positions and geometric relationships do not correspond to the positions and geometric relationships expected to confirm the user's identity with device 100. In this example, the determined details 112, 113, and points of interest 116, their positions, and the geometric relationships between them do not correspond to those expected. For example, such an example corresponds to a situation where the finger set on the capture surface 102 of device 100 is a real finger but not the finger of the person intended to be identified. In this case, device 100 may consider that the user's identity has not been confirmed.

[0092] In one particular embodiment, in order to compare the performed calculation with the expected result to confirm or not confirm the identity of the user of device 100, the comparison step may include a score calculation ( Figure 3 Step 224 in the diagram (its value depends on the correlation between the calculated location and geometric relationship on one hand and the theoretical location and geometric relationship on the other hand, to confirm or disconfirm the user's identity. For example, in situations such as those previously discussed...) Figure 6 In the described case, the fraction calculation method can obtain a fraction close to or equal to 1, while in cases such as those previously discussed... Figure 7 and Figure 8 In the described scenario, a score close to 0 can be obtained.

[0093] According to one example, such a score can be calculated for each previously determined detail and each previously determined point of interest. Then, some or all of the obtained scores can be selected, and a global score representing the previously calculated scores (sum, average, linear combination, etc.) can be calculated. The score calculation method is not described in detail here and may correspond to an adaptation of one of the methods known to those skilled in the art, for example, applied to identification by comparing fingerprints.

[0094] The score is then compared with a threshold to assess the correspondence between the performed biometrics and the expected biometric data, thereby confirming or denying the user's identity (step 226).

[0095] In the exemplary embodiments described above, the fingerprint features of element 101 determined from surface image 110 correspond to details 112, 113 of the fingerprint, and the microvascular system features of element 101 determined from global volume image correspond to points of interest 116 of the microvascular system. According to another exemplary embodiment, determining the fingerprint and the features of the element's microvascular system includes extracting at least one feature vector from the surface image and the global volume image, performed using a deep learning algorithm. In this case, comparing the previously determined features with expected theoretical features to identify or not identify element 101 may include comparing the previously determined feature vector with at least one expected theoretical feature vector using a deep learning algorithm. In this other example, a score calculation may also be performed, the value of which depends on the correlation between the determined feature vector used to identify or not identify element 101 and the expected feature vector.

[0096] The paper by Joshua J. et al., Engelsma, “Learning a Fixed-Length Fingerprint Representation”, IEEE Transactions on Pattern Analysis and Machine Intelligence 43 (2019): 1981–1997, describes examples of deep learning algorithms that can be implemented for this exemplary embodiment. For example, Section 3 of this document describes a CNN (convolutional neural network) deep learning algorithm for extracting features from fingerprints, using a convolutional neural network architecture as described in this document. Figure 8As shown. By training the algorithm on a fingerprint database, as described in Section 7 (consisting of Nv views of different Ni identities, e.g., Ni fingers), an algorithm is obtained that brings views of the same identity closer to each other in an Nd-dimensional space and further away from other identities in the same space, such that distance calculation in this space provides a relevant matching algorithm, as described in Equation 11 in Section 4 of this document. The method described herein can be applied to data acquired by biometric device 100, here being surface images superimposed and aligned with volumetric images. Furthermore, the architecture of the convolutional neural network can be adapted to the fact that the input image is 3D. In addition, a training database consisting of surface images and volumetric images of multiple identities (e.g., multiple fingers) and a certain number (e.g., greater than 100) and multiple views (e.g., at least 20) for each identity can be used. Thus, the obtained result corresponds to a CNN-type algorithm that generates feature vectors of size Nd (with projected coordinates in an Nd-dimensional space) from the input biometric view. Then, to compare the two vectors, the distance algorithm proposed herein can be applied and produce a matching score as described above.

[0097] The document by Y. Tang et al., “FingerNet: An Unified Deep Network for Fingerprint Minutiae Extraction”, arXiv:1709.02228, also describes examples of deep learning algorithms that can be implemented for this other exemplary embodiment. Specifically, the document describes a method for extracting fingerprint details using deep learning. Similar methods can be applied to fingerprint images obtained with biometric device 100 to extract details. Learning methods and internal databases can be used to implement this method. For example, an adaptation of the method can also achieve the extraction of points of interest in volumetric images by manually annotating points of interest in the images within a training library. This adaptation can involve modifying the neural network architecture to accept 3D images as input. Another adaptation allows for the construction of unique algorithms that extract details from surface images and extract points of interest from volumetric images.

[0098] Therefore, device 100 can form a complete biological system integrating identification, anti-spoofing, and activity detection functions based on a single compact representation of information from the captured surface fingerprint and information from the microvascular system. Device 100 can significantly achieve:

[0099] - Higher level of security compared to recognition devices that rely solely on fingerprint capture;

[0100] - Due to the reliability of data related to the microvascular system, a higher level of anti-spoofing and activity detection assurance is provided.

[0101] In the example above, device 100 performs biometrics by comparing measured data included in surface and volume images with expected data corresponding to a person whose identity is being verified. Alternatively, device 100 may compare the measured data with expected data corresponding to multiple people, in which case device 100 may verify whether the user corresponds to one of these people.

[0102] In the above description, device 100 can compare the performed measurement value with expected data stored in the memory of device 100. Alternatively, device 100 can connect to a remote database storing expected data corresponding to one or more people to verify whether the user of device 100 corresponds to that person or one of those people. In another variation, device 100 can also be connected to a smart card containing data recorded during previous registration. The data comparison step can then be performed on the smart card instead of on device 100.

[0103] In the above description, device 100 is used to directly verify or deny the identity of the user of device 100. Prior to this identification, device 100 may be used to perform initial acquisition of data (one or more surface images and volumetric images) during previous registration, which will be used as intended data during future identification of the user of device 100. Other uses of device 100 are also conceivable.

[0104] Various embodiments and variations have been described. Those skilled in the art will understand that certain features of these various embodiments and variations can be combined, and other variations will occur to those skilled in the art.

[0105] Finally, based on the functional indications given above, the actual implementation of the embodiments and variations is within the capabilities of those skilled in the art.

Claims

1. A biometric device (100) comprising at least one ultrasonic transducer circuit (104), a control circuit (106), and a data processing circuit (108), configured to perform at least the following steps: - Acquire (206, 210) at least one surface image (110) of at least one element (101) of the user's body of the device (100), the at least one element (101) being intended to be positioned against the capture surface (102) of the device (100); and at least a series of volume images of the element (101), from which a global volume image (114) of the microvascular system of the element (101) is determined. - Determine the fingerprint features of the element (101) from the surface image (110) (214, 216, 218, 220), and determine the microvascular system features of the element (101) from the global volume image (114); - The features identified in the previous steps (214, 216, 218, 220) are compared with the features calculated during the initial steps of user registration (222, 224, 226) to identify or not identify the element (101). in, The data processing circuit (108) is configured to perform a step (212) of aligning the surface image (110) with the global volume image (114) between the steps of acquiring the surface image (110) and the series of volume images (206, 210) and the steps of determining the fingerprint and microvascular system features of the element (101) (214, 216, 218, 220).

2. The biometric device (100) according to claim 1, wherein, The data processing circuit (108) is configured to determine the global volume image (114) by implementing a method for spatiotemporal filtering of a series of volume images of the element (101).

3. The biometric device (100) according to any one of the preceding claims, wherein the ultrasonic transducer circuit (104) is configured to perform the acquisition of a series of volumetric images of the element (101) by means of pulse wave Doppler imaging.

4. The biometric device (100) according to any one of the preceding claims, wherein the data processing circuit (108) is configured such that the step (214, 216, 218, 220) of determining the fingerprint and the characteristics of the microvascular system of the element (101) includes at least: - Determine (214, 216) the details of the fingerprint of the element (101) from the surface image (110) (112, 113), and determine the point of interest (116) of the microvascular system of the element (101) from the global volume image (114). - Calculate (220) the position of at least one detail (113) in the surface image (110) and / or at least one point of interest (116) in the global volume image (114), and calculate the geometric relationship between the at least one detail (113) and / or the at least one point of interest (116) and at least a portion of the other details (112) and points of interest (116) for the point of interest (116) taking into account the depth of the point of interest (116) in the global volume image (114); And the comparison steps (222, 224, 226) include comparing the previously calculated position and geometric relationship with the expected theoretical position and geometric relationship to identify or not identify the element (101).

5. The biometric device (100) according to claim 4, wherein, The point of interest (116) corresponds to the bifurcation of a microvessel and / or the point of strong curvature of a microvessel.

6. The biometric device (100) according to any one of claims 4 or 5, wherein, The data processing circuit (108) is configured to perform a step (218) between the step (214, 216) of determining the details (112, 113) and the points of interest (116) and the calculation step (220) to combine the determined details (112, 113) and points of interest (116) in the form of a point cloud.

7. The biometric device (100) according to any one of claims 4 to 6, wherein, The data processing circuit (108) is configured such that during the calculation step (220), the other details (112) and points of interest (116) used to calculate the geometric relationship correspond to details (112) included in the region of the surface image (110) defined by the planar geometry, and points of interest (116) included in the region of the global volume image (114) included in the volume (120) formed by the planar geometry, which extends in the dimension of the global volume image (114) perpendicular to the surface image (110).

8. The biometric device (100) according to claim 7, wherein, The planar geometric figure (118) corresponds to a polygon, a circle, or an ellipse.

9. The biometric device (100) according to any one of claims 7 or 8, wherein the data processing circuit (108) is configured to repeat the calculation step (220) for each determined detail (112, 113) and each determined point of interest (116), and to perform the comparison step (222, 224, 226) for the calculated position and the geometric relationship for each detail (112, 113) and each point of interest (116).

10. The biometric device (100) according to any one of claims 4 to 9, wherein, The data processing circuit (108) is configured such that the geometric relationship calculated between the first detail (113) or point of interest (116) and the second detail (112) or point of interest (116) includes the distance between the first detail (113) or point of interest (116) and the second detail (112) or point of interest (116), the orientation of the first detail (113) or point of interest (116) relative to the second detail (112) or point of interest (116), and the orientation of the second detail (112) or point of interest (116) relative to the first detail (113) or point of interest (116).

11. The biometric device (100) according to any one of claims 4 to 10, wherein, The data processing circuit (108) is configured such that for a point of interest (116) corresponding to a bifurcation or strong bend of a microvessel, the calculated geometric relationship includes the orientation of each microvessel in the microvessel relative to the other details and points of interest.

12. The biometric device (100) according to any one of claims 1 to 3, wherein, The data processing circuit (108) is configured such that: - The determination (214, 216, 218, 220) of the fingerprint and the features of the microvascular system of the element (101) includes extracting at least one feature vector from the surface image (110) and the global volume image (114) by implementing a deep learning algorithm; Comparing previously determined features with expected theoretical features (222, 224, 226) to identify or not identify the element (101) includes comparing the previously determined feature vector with at least one expected theoretical feature vector by the deep learning algorithm.

13. The biometric device (100) according to any one of the preceding claims, wherein the data processing circuit (108) is configured such that the comparison step (222, 224, 226) includes the calculation (224) of a score whose value depends on the correlation between the determined feature for identifying or not identifying the element (101) and the expected theoretical feature.

14. The biometric device (100) according to any one of claims 4 to 11 and according to claim 13, wherein, The data processing circuit (108) is configured to calculate a score for each determined detail (112, 113) and each determined point of interest (116) during the comparison steps (222, 224, 226).

Images