Biometric authentication device, biometric authentication method, and recording medium

Abstract

The present invention comprises: an extraction unit 211 for extracting a feature quantity BC to be authenticated, which is a feature quantity pertaining to data BD to be authenticated that is used for performing an authentication operation; a calculation unit 212 for calculating an unsuitability score WS indicating the degree of similarity between the feature quantity BC to be authenticated and an unsuitability feature quantity WC set on the basis of the feature quantity of each of a plurality of sample data that degrade the accuracy of the authentication operation due to having the same kind of unsuitability for authentication; and an assessment unit 213 for assessing, in accordance with the unsuitability score WS, whether or not the data BD to be authenticated is unsuitable for the authentication operation.

Description

[Technical Field]

[0001] The present disclosure relates to the technical fields of a biometric authentication device, a biometric authentication method, and a recording medium. [Background technology]

[0002] Patent Document 1 describes a technology that compares vectors extracted from selected facial regions in two facial images to be compared and obtains a facial recognition score. Patent Document 2 describes a technology that determines the positional fluctuation of a background-separated image and prohibits authentication processing if the index indicating stillness is below a certain number, or discards the image without using it as training data if the magnitude of the positional fluctuation of the background-separated image is greater than a certain threshold, thereby preventing inappropriate images from being passed on to processing after the generation of the background-separated image. Patent Document 3 describes a technology that determines an authentication error when the degree of match with authenticity information of reliability information is determined to be lower than a predetermined value, or when the degree of correlation is determined to be lower than a predetermined value, or when the amplitude of the pulse wave signal of the vitality information of reliability information is determined to be smaller than a predetermined value, and notifies a display unit, speaker, or the like of the authentication error and the reason for the authentication error. Patent Document 4 describes a technology that sets different thresholds for successful facial recognition depending on the number of registered facial images or features so that the calculation criteria for the recognition score differ depending on the number of registered facial images or features. Patent Document 5 also describes a technology that corrects the score, which is the result of face matching, based on the score distribution. Furthermore, in face recognition using cosine similarity as a matching score, Non-Patent Document 1 describes a technology that uses similar features of images that have lost their individuality to prevent erroneous recognition due to the input of inappropriate images. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Special Publication No. 2017-517076 [Patent Document 2] Japanese Patent Application Publication No. 2020-129298 [Patent Document 3] International Publication No. 2016 / 067556 [Patent Document 4] Japanese Patent Application Laid-Open No. 2013-142930 [Patent Document 5] Japanese Patent Application Laid-Open No. 2008-40874 [Non-patent literature]

[0004] [Non-Patent Document 1] Siqi Deng, “Harnessing Unrecognizable Faces for Face Recognition” (arXiv 2021) Summary of the Invention [Problem to be solved by the invention]

[0005] An object of this disclosure is to provide a biometric authentication device, a biometric authentication method, and a recording medium that aim to improve upon the techniques described in prior art documents. [Means for solving the problem]

[0006] One aspect of the biometric authentication device includes an extraction means for extracting authentication features, which are features of authentication data used to perform an authentication operation; an inappropriate feature set based on the features of each of a plurality of sample data that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness; a calculation means for calculating an inappropriateness score indicating the similarity with the authentication feature; and a determination means for determining whether the authentication data is inappropriate for the authentication operation based on the inappropriateness score.

[0007] One aspect of the biometric authentication method extracts features to be authenticated, which are features of data to be authenticated used to perform an authentication operation, calculates an inappropriateness score indicating the similarity between the features to be authenticated and inappropriate features set based on the features of each of multiple sample data that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness, and determines whether the data to be authenticated is inappropriate for the authentication operation based on the inappropriateness score.

[0008] One aspect of the recording medium causes a computer to execute a biometric authentication method that extracts authentication features, which are features of authentication data used to perform an authentication operation, calculates an inappropriateness score indicating the similarity between the authentication features and inappropriate features set based on the features of each of a plurality of sample data that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness, and determines whether the authentication data is inappropriate for the authentication operation based on the inappropriateness score. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a block diagram showing the configuration of a biometric authentication device according to the first embodiment. [Figure 2] FIG. 2 is a block diagram showing the configuration of a biometric authentication device according to the second embodiment. [Figure 3] FIG. 3 is a flowchart showing the flow of the inappropriate feature WC setting operation performed by the biometric authentication device in the second embodiment. [Figure 4] FIG. 4 is a flowchart showing the flow of the biometric authentication operation performed by the biometric authentication device in the second embodiment. [Figure 5] FIG. 5 is a flowchart showing the flow of a biometric authentication operation performed by the biometric authentication device according to the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments of a biometric authentication device, a biometric authentication method, and a recording medium will be described with reference to the drawings. [1: First embodiment]

[0011] First, a first embodiment of a biometric authentication device, a biometric authentication method, and a recording medium will be described. Hereinafter, the first embodiment of the biometric authentication device, the biometric authentication method, and the recording medium will be described using a biometric authentication device 1 to which the first embodiment of the biometric authentication device, the biometric authentication method, and the recording medium is applied. [1-1: Configuration of Biometric Authentication Device 1]

[0012] 1 is a block diagram showing the configuration of a biometric authentication device 1 according to the first embodiment. As shown in FIG. 1, the biometric authentication device 1 includes an extraction unit 11, a calculation unit 12, and a determination unit 13.

[0013] The extraction unit 11 extracts authentication target features, which are features of authentication target data BD used to perform authentication operations. The authentication target data is data related to the authentication target. The calculation unit 12 calculates an inappropriateness score indicating the similarity between the inappropriate features and the authentication target features. The inappropriate features are set based on the features of each of multiple sample data that have the same type of authentication inappropriateness and therefore deteriorate the accuracy of the authentication operation. The judgment unit 13 judges whether the authentication target data BD is inappropriate for authentication operations according to the inappropriateness score.

[0014] It is known that using data with certain characteristics in biometric authentication can reduce the accuracy of authentication and increase the number of false positives. Such data that reduces the accuracy of authentication is referred to as data that is inappropriate for authentication. For example, when the data to be authenticated is an image of the target, an image in which factors that hinder proper authentication are captured along with the target can be cited as an example of data that is inappropriate for authentication. Specific examples of such data include images of the target itself captured in inappropriate conditions, images captured in inappropriate conditions around the target, and images captured with inappropriate settings on the imaging device itself. Other examples include images in which the target is captured in a state where its individuality is lost, or images in which the target is not actually captured. Specific examples include images in which key parts of the target are hidden, images in which the surroundings of the target are extremely bright or dark, images in which the boundary between the target and the surrounding environment is unclear, images that are significantly out of focus, images with extremely low resolution, and images captured with a camera with a dirty lens. Furthermore, if the data to be authenticated is a facial image that includes a human face, the image may be erroneously determined to be a human face and acquired as a facial image. Specific examples of such images include images of paintings such as portraits, images of statues such as sculptures, images of animals such as monkeys, and images of patterns that appear to be human faces.

[0015] If data with such authentication inappropriateness is used to authenticate an individual, it may happen that a different individual is authenticated as the actual individual. In this way, if the data to be authenticated is data with authentication inappropriateness, it is unsuitable for authentication operations, and therefore it is desirable not to use it for authentication operations.

[0016] The feature quantities of each piece of data having authentication inappropriateness are similar. Furthermore, among the authentication inappropriatenesses, the feature quantities of each piece of data having the same type of authentication inappropriateness are particularly similar. Therefore, when the feature quantities of the data to be authenticated used for performing the authentication operation are similar to the inappropriateness feature set based on the feature quantities of each piece of sample data having the same type of authentication inappropriateness, the data to be authenticated can be considered to have the corresponding type of authentication inappropriateness. Furthermore, since the data can be considered to have the possibility of deteriorating the accuracy of the authentication operation due to the authentication inappropriateness, it is desirable not to use the data to be authenticated for authentication.

[0017] For this reason, the biometric authentication device 1 in the first embodiment calculates an inadequacy score indicating the similarity between the inadequacy feature and the feature to be authenticated, and determines whether the data to be authenticated is inappropriate for the authentication operation according to the inadequacy score. This inadequacy feature is a feature set based on each feature of a plurality of sample data having the same type, i.e., one type of authentication inadequacy. Therefore, compared with a feature set based on each feature of a plurality of sample data having multiple types of authentication inadequacy, the inadequacy feature is superior in expressing one type of authentication inadequacy. [1-2: Technical Effects of Biometric Authentication Device 1]

[0018] In this way, the biometric authentication device 1 in the first embodiment can accurately determine whether or not the data to be authenticated is inappropriate for authentication operations, thereby preventing the occurrence of erroneous authentication. [2: Second embodiment]

[0019] Next, a second embodiment of the biometric authentication device, the biometric authentication method, and the recording medium will be described. Hereinafter, the second embodiment of the biometric authentication device, the biometric authentication method, and the recording medium will be described using a biometric authentication device 2 to which the second embodiment of the biometric authentication device, the biometric authentication method, and the recording medium is applied. [2-1: Configuration of Biometric Authentication Device 2]

[0020] First, the configuration of the biometric authentication device 2 in the second embodiment will be described with reference to Fig. 2. Fig. 2 is a block diagram showing the configuration of the biometric authentication device 2 in the second embodiment. In the following description, components that have already been described will be given the same reference numerals, and detailed description thereof will be omitted.

[0021] 2, the biometric authentication device 2 includes a calculation device 21 and a storage device 22. The biometric authentication device 2 may further include a communication device 23, an input device 24, and an output device 25. However, the biometric authentication device 2 does not necessarily include at least one of the communication device 23, the input device 24, and the output device 25. The calculation device 21, the storage device 22, the communication device 23, the input device 24, and the output device 25 may be connected via a data bus 26.

[0022] The arithmetic device 21 includes, for example, at least one of a central processing unit (CPU), a graphics processing unit (GPU), and a field programmable gate array (FPGA). The arithmetic device 21 reads a computer program. For example, the arithmetic device 21 may read a computer program stored in the storage device 22. For example, the arithmetic device 21 may read a computer program stored in a computer-readable, non-transitory recording medium using a recording medium reading device (e.g., an input device 24 described later) not shown in the drawings that is provided in the biometric authentication device 2. The arithmetic device 21 may acquire (i.e., download or read) the computer program from a device (not shown) located outside the biometric authentication device 2 via the communication device 23 (or another communication device). The arithmetic device 21 executes the read computer program. As a result, logical functional blocks for executing operations to be performed by the biometric authentication device 2 are realized within the arithmetic device 21. That is, the arithmetic device 21 can function as a controller for realizing logical function blocks for executing the operations (in other words, processes) that the biometric authentication device 2 should perform.

[0023] Fig. 2 shows an example of logical functional blocks realized in the arithmetic device 21 to perform biometric authentication operations. As shown in Fig. 2, the arithmetic device 21 realizes an extraction unit 211 which is a specific example of "extraction means," a calculation unit 212 which is a specific example of "calculation means," a determination unit 213 which is a specific example of "determination means," a presentation unit 214 which is a specific example of "presentation means," a correction unit 215 which is a specific example of "correction means," and an authentication unit 216.

[0024] The operations of the extraction unit 211, calculation unit 212, determination unit 213, presentation unit 214, and correction unit 215 will be described in detail later with reference to Figures 3 and 4. However, the calculation device 21 does not necessarily have to include at least one of the presentation unit 214, correction unit 215, and authentication unit 216.

[0025] The storage device 22 can store desired data. For example, the storage device 22 may temporarily store a computer program executed by the arithmetic device 21. The storage device 22 may temporarily store data that the arithmetic device 21 temporarily uses when the arithmetic device 21 is executing a computer program. The storage device 22 may store data that the biometric authentication device 2 stores long-term. The storage device 22 may include at least one of a RAM (Random Access Memory), a ROM (Read Only Memory), a hard disk device, a magneto-optical disk device, an SSD (Solid State Drive), and a disk array device. In other words, the storage device 22 may include a non-transitory recording medium.

[0026] The storage device 22 may store the feature generation model GM, the matching feature CC, and the inappropriate feature WC. However, the storage device 22 does not necessarily have to store the feature generation model GM, the matching feature CC, and the inappropriate feature WC. The feature generation model GM and the matching feature CC will be described in detail later.

[0027] The communication device 23 is capable of communicating with devices external to the biometric authentication device 2 via a communication network (not shown).

[0028] The input device 24 is a device that accepts information input to the biometric authentication device 2 from outside the biometric authentication device 2. For example, the input device 24 may include an operation device (for example, at least one of a keyboard, a mouse, and a touch panel) that can be operated by an operator of the biometric authentication device 2. For example, the input device 24 may include a reading device that can read information recorded as data on a recording medium that can be externally attached to the biometric authentication device 2.

[0029] The output device 25 is a device that outputs information to the outside of the biometric authentication device 2. For example, the output device 25 may output information as an image. That is, the output device 25 may include a display device (a so-called display) that can display an image showing the information to be output. For example, the output device 25 may output information as sound. That is, the output device 25 may include an audio device (a so-called speaker) that can output sound. For example, the output device 25 may output information on paper. That is, the output device 25 may include a printing device (a so-called printer) that can print desired information on paper.

[0030] In the second embodiment, the authentication data BD and each of the plurality of sample data may be a biometric image in which a biometric subject to authentication is captured. The authentication inappropriateness may include at least one of (i) the inappropriateness of the biometric subject being captured in the biometric image with at least a part of the biometric subject being blurred, (ii) the inappropriateness of the biometric subject being captured in the biometric image with at least a part of the biometric subject being obscured by an obstruction, and (iii) the inappropriateness of noise being superimposed on the biometric image.

[0031] The inappropriateness feature WC includes at least a first inappropriateness feature 1WC set based on the feature of each of a plurality of first sample data items that deteriorate the accuracy of the authentication operation due to the presence of a first type of authentication inappropriateness, and a second inappropriateness feature 2WC set based on the feature of each of a plurality of second sample data items that deteriorate the accuracy of the authentication operation due to the presence of a second type of authentication inappropriateness different from the first type. For example, the first type of authentication inappropriateness may be one of the above (i) to (iii), and the second type of authentication inappropriateness may be another one of the above (i) to (iii). Furthermore, a third type of authentication inappropriateness may exist, and the third type of authentication inappropriateness may be one other than the first type of authentication inappropriateness and the second type of authentication inappropriateness (i) to (iii). Furthermore, a fourth or more types of authentication inappropriateness may exist. In the example shown below, a case in which the first type of authentication inappropriateness and the second type of authentication inappropriateness exist will be described.

[0032] Prior to the biometric authentication operation, a feature generation model GM is constructed, matching features CC are registered, and inappropriate features WC are registered. [2-2: Construction of the feature generation model GM]

[0033] The feature generation model GM is a model capable of generating features of a biometric image when the biometric image is input. The feature generation model GM may be constructed by machine learning so as to be capable of generating the same features when biometric images of the same individual are input. Specifically, the feature generation model GM may be constructed by adjusting the parameters of the feature generation model GM so that a loss function set based on the error between multiple features generated from biometric images of the same individual is reduced (preferably minimized). The feature generation model GM may be constructed as, for example, a convolutional neural network that generates features by convolution processing. The feature generation model GM may be, for example, VGG or RewNet, and may be trained using DeepFace or ArcFace. The feature generation model GM may be any model that can generate features with high accuracy, and may also be another trained engine.

[0034] For example, suppose there are three types of authentication inappropriateness: (i), (ii), and (iii). Let us consider a case where a judgment model for determining whether or not each type of inappropriateness exists is constructed by machine learning and used. In this case, it becomes necessary to construct three types of judgment models: a judgment model for authentication inappropriateness (i), a judgment model for authentication inappropriateness (ii), and a judgment model for authentication inappropriateness (iii). In addition to each judgment model, it also becomes necessary to construct a model for biometric authentication, such as a model for generating features of biometric images.

[0035] In contrast, the biometric authentication device 2 in the second embodiment only needs to prepare a feature generation model GM constructed by machine learning to generate features used for authentication. That is, the biometric authentication device 2 in the second embodiment only needs to construct one model, the feature generation model GM. This feature generation model GM generates matching features CC, which are features of the matching data CD, and generates authenticated features BC, which are features of the data to be authenticated. Furthermore, the feature generation model GM also generates features for each of a plurality of sample data having authentication inappropriateness. The inappropriateness features WC are set based on the features generated by the feature generation model GM. Therefore, the biometric authentication device 2 in the second embodiment has a simple configuration. Note that the generation of the matching features CC, the authenticated features BC, and the features of each of a plurality of sample data having authentication inappropriateness will be described later with reference to FIGS. 3 and 4. [2-3: Setting operation of inappropriate feature WC performed by biometric authentication device 2]

[0036] Next, the operation of setting the inappropriate feature WC performed by the biometric authentication device 2 in the second embodiment will be described with reference to Fig. 3. Fig. 3 is a flowchart showing the flow of the operation of setting the inappropriate feature WC performed by the biometric authentication device 2 in the second embodiment.

[0037] 3, the extraction unit 211 acquires a plurality of sample data that deteriorates the accuracy of the authentication operation (step S31). The extraction unit 211 determines whether or not there is unprocessed sample data (step S32).

[0038] If the result of the determination in step S32 is that unprocessed sample data exists (step S32: Yes), the extraction unit 211 extracts feature amounts of multiple sample data having the same type of authentication inappropriateness from the unprocessed sample data (step S33). The extraction unit 211 may extract feature amounts from multiple sample data using the feature generation model GM.

[0039] The extraction unit 211 sets an inappropriateness feature WC based on each feature of multiple sample data having the same type of authentication inappropriateness (step S34) and stores the inappropriateness feature WC in the storage device 22 (step S35). Here, each inappropriateness feature WC may be, for example, an average value of each feature of the multiple sample data. For example, if each feature of the multiple sample data can be expressed as a vector, the inappropriateness feature WC may be a normalized average vector of each feature vector of the multiple sample data. When the inappropriateness feature WC is an average value, the likelihood of authentication inappropriateness of each sample data is equally included, and an inappropriateness feature that better represents the authentication inappropriateness feature can be generated. The inappropriateness feature WC may be any feature that is representative of authentication inappropriateness and may be calculated by any method, such as calculating a weighted average value.

[0040] After step S35, the process proceeds to step S32. If the result of the determination in step S32 is that there is no unprocessed sample data (step S32: No), the operation of setting the inappropriate feature WC ends.

[0041] For example, if sample data having two types of authentication inappropriateness is included, the operations of steps S33 to S35 are repeated twice. [2-4: Biometric authentication operation performed by biometric authentication device 2]

[0042] Next, the biometric authentication operation performed by the biometric authentication device 2 in the second embodiment will be described with reference to Fig. 4. Fig. 4 is a flowchart showing the flow of the biometric authentication operation performed by the biometric authentication device 2 in the second embodiment.

[0043] 4, the extraction unit 211 acquires a biometric image as authentication target data BD (step S21). The extraction unit 211 extracts authentication target features BC, which are features of the biometric image (step S22). The extraction unit 211 may extract the authentication target features BC from the authentication target data BD using a feature generation model GM.

[0044] The calculation unit 212 calculates the inappropriate scores WS as a first inappropriate score 1WS indicating the similarity between the first inappropriate feature 1WC and the feature to be authenticated BC, and a second inappropriate score 2WS indicating the similarity between the second inappropriate feature 2WC and the feature to be authenticated BC (step S23). The determination unit 213 determines whether the authentication target data BD is inappropriate for the authentication operation according to the first inappropriateness score 1WS and the second inappropriateness score 2WS (step S24).

[0045] The feature quantities of sample data having the authentication inappropriateness (i) are similar to each other. The feature quantities of sample data having the authentication inappropriateness (ii) are similar to each other. The feature quantities of sample data having the authentication inappropriateness (iii) are similar to each other. On the other hand, the feature quantities of sample data having the authentication inappropriateness (i), the feature quantities of sample data having the authentication inappropriateness (ii), and the feature quantities of sample data having the authentication inappropriateness (iii) are not similar to each other. Therefore, when the data to be authenticated BD has the authentication inappropriateness (i), the authenticated feature quantity BC, which is the feature quantity of the data to be authenticated BD, is similar to the feature quantities of sample data having the authentication inappropriateness (i). When the data to be authenticated BD has the authentication inappropriateness (ii), the authenticated feature quantity BC is similar to the feature quantities of sample data having the authentication inappropriateness (ii). When the data to be authenticated BD has the authentication inappropriateness (iii), the authenticated feature quantity BC is similar to the feature quantities of sample data having the authentication inappropriateness (iii). In other words, the judgment unit 213 can judge that the authentication data BD whose features are similar to the inappropriate features WC set from the features of sample data having any type of authentication inappropriateness is inappropriate for authentication operations because of the similar authentication inappropriateness.

[0046] For example, if the feature generation model GM is expressed as a function f, Feature quantity of sample data having authentication inappropriateness=f(sample data having authentication inappropriateness) Authentication feature BC = f (authentication data BD) It can be expressed as: The sample data having authentication inappropriateness and the data to be authenticated BD have a size of, for example, W100 x H100 x 3ch. Furthermore, the feature of the sample data having authentication inappropriateness and the feature to be authenticated BC can be expressed, for example, by a 100-dimensional vector. In this case, the inappropriate feature WC set based on the feature of multiple sample data having authentication inappropriateness can also be expressed by a 100-dimensional vector. The norm of each feature may be set to 1. In this case, the calculation unit 212 calculates the inappropriateness score WS between the inappropriate feature WC and the feature to be authenticated BC by the cosine similarity=<WC,BC> The cosine similarity may be calculated using the inner product of the inappropriate feature WC and the feature to be authenticated BC. The cosine similarity corresponds to the value of cos(θ) where θ is the angle between the two vectors, and has the property that it takes a maximum value of 1 when the two vectors are the same, 0 when the two vectors are perpendicular, and a minimum value of -1 when the two vectors are pointing in opposite directions. In other words, the cosine similarity utilizes the property that the feature values of data related to the same individual tend to be similar and point in the same direction. When cosine similarity is used, the inappropriateness score WS can take a value from -1 to 1. Below, an example will be described in which the inappropriateness score WS can take a value from -1 to 1.

[0047] The user can appropriately determine the branching of processing related to the value of the inappropriateness score WS. For example, if one or more inappropriateness scores WS of the corresponding authentication target data BD are equal to or greater than a first value, the determination unit 213 may determine that the corresponding authentication target data BD is completely inappropriate for authentication. Then, the biometric authentication device 2 may terminate processing without calculating a matching score CS with the matching feature CC. For example, if the first value is 0.7, the first inappropriateness score 1WS is 0.75, and the second inappropriateness score 2WS is −0.8, the inappropriateness scores WS of 1 or more are equal to or greater than the first value, so the corresponding authentication target data BD is determined to be completely inappropriate for authentication, and processing may terminate without calculating a matching score CS with the matching feature CC.

[0048] If the result of the determination in step S24 is that the authenticated data BD is inappropriate for the authentication operation (step S24: Yes), the determination unit 213 determines the reason why the authenticated data BD is inappropriate for the authentication operation according to the first inappropriateness score 1WS and the second inappropriateness score 2WS (step S25). If the first inappropriateness score 1WS and the second inappropriateness score 2WS indicate that the similarity between the first inappropriateness feature 1WC and the authenticated feature BC is higher than the similarity between the second inappropriateness feature 2WC and the authenticated feature BC, the determination unit 213 determines that the reason why the authenticated data BD is inappropriate for the authentication operation is the first type of inappropriateness. On the other hand, when the first inappropriate score 1WS and the second inappropriate score 2WS indicate that the similarity between the second inappropriate feature 2WC and the feature to be authenticated BC is higher than the similarity between the first inappropriate feature 1WC and the feature to be authenticated BC, the judgment unit 213 determines that the reason the authenticated data BD is inappropriate for authentication operation is the second type of inappropriateness.

[0049] The presentation unit 214 presents the reason why the authentication target data BD is inappropriate for the authentication operation (step S26). On the other hand, if the result of the judgment in step S24 is that the authenticated data BD is not inappropriate for the authentication operation (step S24: No), the calculation unit 212 calculates a matching score CS indicating the similarity between the authenticated feature BC and the pre-registered matching feature CC (step S27).

[0050] Here, the extraction unit 211 may use the feature generation model GM to extract and register the matching feature CC in advance. The feature generation model GM may receive the matching data CD as input and generate the matching feature CC, which is a feature of the matching data CD. The generated matching feature CC may be stored and registered in the storage device 22.

[0051] The authentication unit 216 may determine whether or not the person is the same person as the person in question based on the value of the matching score CS for biometric authentication. For example, both the matching feature CC and the authenticated feature BC may be expressed as vectors, and the determination unit 213 may obtain the inner product of the matching feature CC and the authenticated feature BC, calculate the cosine similarity, and use it as the matching score CS. In this case, the matching score CS may take a value from -1 to 1. For example, if the matching score CS is 0.5 or higher, it may be determined that the two people are the same person, and if it is less than 0.5, it may be determined that the two people are different people. The value used for this determination can be determined arbitrarily by the user depending on the requirements.

[0052] In other words, the calculation unit 212 that calculates the inappropriateness score and the calculation unit 212 that calculates the matching score may be the same mechanism. The calculation unit 212 may calculate an inappropriateness score WS that indicates the similarity between the inappropriateness feature WC and the feature to be authenticated BC, or may calculate a matching score CS that indicates the similarity between the feature to be authenticated BC and a matching feature CC that has been registered in advance. The calculation unit 212 can calculate the scores by the same operation even if the input data is different. Note that the calculation unit 212 does not have to calculate the inappropriateness score WS and the matching score CS using cosine similarity, and may calculate the similarity of the distance between vectors, such as Euclidean space distance, or may calculate them using any known method.

[0053] When the determination unit 213 determines that the authentication-target data BD may be unsuitable for authentication, the correction unit 215 corrects the matching score CS (step S28). For example, if one or more of the inappropriateness scores WS of the authentication-target data BD are less than a first value and greater than or equal to a second value, the corresponding authentication-target data BD may be deemed unsuitable for authentication, and the calculated matching score CS may be corrected downward. For example, if the first value is 0.7, the second value is, for example, -0.7, the first inappropriateness score 1WS is 0.5, and the second inappropriateness score 2WS is -0.8, one or more of the inappropriateness scores WS are less than the first value and greater than or equal to the second value, and therefore the corresponding authentication-target data BD may be deemed unsuitable for authentication, and the calculated matching score CS may be corrected downward. On the other hand, for example, if all of the inappropriateness scores WS of the authentication-target data BD are less than the second value, the corresponding authentication-target data BD may be deemed appropriate for authentication, and the calculated matching score CS may not be corrected.

[0054] The correction unit 215 may correct the matching score CS so that the larger the value of the inappropriateness score WS, the lower the matching score CS. In other words, the correction unit 215 may correct the matching score CS so that the likelihood of the person being identified is evaluated as lower. When there are multiple types of authentication inappropriateness, the method of correcting the matching score CS may be determined based on the value of the highest inappropriateness score WS. The correction by the correction unit 215 to lower the matching score CS may be a linear transformation or a nonlinear transformation such as gamma correction. For example, the correction unit 215 may reduce the matching score by 50% when the inappropriateness score is 0.65 or greater, reduce the matching score by 30% when the inappropriateness score is 0 or greater, or reduce the matching score by 10% when the inappropriateness score is -0.65 or greater. In this case, for example, if the matching score with certain matching data CD is 0.9 but the first inappropriateness score 1WS is 0 or greater, the correction unit 215 corrects the corresponding matching score to 0.63. That is, the correction by the correction unit 215 lowers the evaluation of the likelihood that the data is the same as the collation data CD.

[0055] Alternatively, although an example has been described in which the correction unit 215 changes the reliability of authentication depending on the inappropriateness score WS, if the value of the inappropriateness score WS exceeds a threshold, the matching score CS may be uniformly corrected to be lower. The authentication unit 216 determines whether the person is the same person or not based on the corrected matching score CS (step S29). The presentation unit 214 presents the authentication result of the authentication target data BD (the determination result as to whether the person is the same person or not) (step S30). [2-4: Technical Effects of Biometric Authentication Device 2]

[0056] The biometric authentication device 2 in the second embodiment determines whether or not there are multiple types of authentication inappropriateness for each type, so it can accurately determine whether or not authenticated data is inappropriate for authentication operation for various types of authentication inappropriateness, and can prevent the occurrence of erroneous authentication. Furthermore, the biometric authentication device 2 in the second embodiment can determine which type of authentication inappropriateness the authenticated data has to be determined to be inappropriate for authentication operation by adopting which type of authentication inappropriateness data as sample data.

[0057] Furthermore, the biometric authentication device 2 according to the second embodiment calculates an inappropriateness score for each type of authentication inappropriateness, and can therefore determine the type of authentication inappropriateness contained in the data to be authenticated. If the feature of the data to be authenticated is similar to the feature of data having a certain authentication inappropriateness, it can be determined that the data to be authenticated contains that authentication inappropriateness. Therefore, the biometric authentication device 2 according to the second embodiment can not only determine that the data to be authenticated is inappropriate for authentication, but also know why the data is inappropriate for authentication.

[0058] Furthermore, the biometric authentication device 2 in the second embodiment calculates an inadequacy score for each type of different authentication inadequacy, and can therefore determine which type of authentication inadequacy and to what extent the data to be authenticated contains. That is, the biometric authentication device 2 in the second embodiment can determine the authentication inadequacy with the highest inadequacy score as the reason why the data to be authenticated is inappropriate for authentication operation.

[0059] Furthermore, since the biometric authentication device 2 in the second embodiment externally presents the reason for authentication inappropriateness, the implementer performing the biometric authentication can know the reason for the authentication inappropriateness. By knowing the type of authentication inappropriateness, the implementer may be able to consider a new authentication mechanism specialized for that type. Furthermore, by knowing the type of authentication inappropriateness, the implementer may be able to determine that the circumstances under which the authenticated data was acquired are inappropriate. For example, the authenticated data may be determined to be inappropriate for authentication operation due to a malfunction of the device, such as a dirty lens, that acquires biometric images. In this case, the implementer may adjust the device to improve the authentication inappropriateness contained in the authenticated data.

[0060] Moreover, the biometric authentication device 2 in the second embodiment generates features to be authenticated and inappropriate features using the same feature generation model GM, and therefore can accurately determine whether or not a feature is inappropriate for authentication. Moreover, the biometric authentication device 2 in the second embodiment performs authentication using the similarity of features generated using the same feature generation model GM, and therefore can perform biometric authentication with high accuracy.

[0061] Furthermore, the biometric authentication device 2 in the second embodiment can correct the matching score when the data to be authenticated contains authentication inappropriateness, thereby further improving the accuracy of authentication. For example, when the authentication data has an inappropriateness score equal to or greater than a certain threshold, the determination unit 213 can determine that the data to be authenticated should not be used as data to be used for matching. On the other hand, when the determination unit 213 determines that the data to be authenticated contains authentication inappropriateness but may be usable for matching, the correction unit 215 can correct the matching score, making it difficult for the authentication unit 216 to determine that the data is the actual person.

[0062] In particular, in biometric authentication, erroneous authentication due to an inappropriate image should be prevented. The household authentication device 2 in the second embodiment can authenticate biometric images with high accuracy.

[0063] In the second embodiment, the authentication data BD is described as a biometric image. Examples of biometric images include face images, iris images, fingerprint images, and other images acquired from a living body. The authentication data BD may be data other than a biometric image, and may be data other than an image, such as audio data. In the case of audio data, data that is inappropriate for authentication and that reduces the accuracy of the authentication operation may be data that does not include speech, which is prone to strong noise. [3: Third embodiment]

[0064] Next, a third embodiment of the biometric authentication device, the biometric authentication method, and the recording medium will be described. Hereinafter, the third embodiment of the biometric authentication device, the biometric authentication method, and the recording medium will be described using a biometric authentication device to which the third embodiment of the biometric authentication device, the biometric authentication method, and the recording medium is applied.

[0065] The biometric authentication device in the third embodiment may have the same configuration as the biometric authentication device 2 in the second embodiment described above. The biometric authentication device in the third embodiment differs from the biometric authentication device 2 in the second embodiment in the type of authentication inappropriateness. Other features of the biometric authentication device may be the same as other features of the biometric authentication device 2.

[0066] In the third embodiment, the authentication data BD and each of the plurality of sample data are images in which a living body that is the target of the authentication operation is captured. Furthermore, in the third embodiment, the authentication inappropriateness includes at least one of (iv) an authentication inappropriateness in which the living body is captured in the image when at least a predetermined region of the living body is obscured, and (v) an authentication inappropriateness in which the living body is captured in the image when another region different from at least a predetermined region of the living body is obscured. The predetermined region of at least a portion of the living body may include the area surrounding the living body. Specifically, the authentication data BD and each of the plurality of sample data may be a facial image in which the face of the living body that is the target of the authentication operation is captured. In this case, the authentication inappropriateness may include at least one of (iv) an inappropriateness in which the face is captured in the facial image when at least a portion of the face is obscured by a mask worn by the living body, and (v) an inappropriateness in which the face is captured in the facial image when at least a portion of the face is obscured by an obstruction other than the mask. In this specific example, the area shielded by the mask worn by the living body may be an example of at least a part of the predetermined area of the living body, including the area surrounding the living body. [3-1: Biometric authentication operation performed by biometric authentication device 3]

[0067] The flow of biometric authentication operation in the third embodiment will be described below with reference to Fig. 5. Fig. 5 is a flowchart showing the flow of biometric authentication operation in the third embodiment. Note that in the following description, processes that have already been described will be assigned the same step numbers, and detailed description thereof will be omitted.

[0068] 5, in the third embodiment, as in the second embodiment, the information processing device 3 performs operations from step S21 to step S25. In the third embodiment, the determination unit 213 determines whether the reason that the authentication data BD is inappropriate for the authentication operation, determined in step S25, is that (iv) the face is captured in the facial image while at least a part of the face is obscured by a mask worn by the living body (step S41).

[0069] If the result of the determination in step S41 is that the reason why the authentication target data BD is unsuitable for authentication operation is (iv) (step S41: Yes), the determination unit 213 determines that the authentication target data BD is not unsuitable for authentication operation. Specifically, the determination unit 213 may decide to perform face image authentication for the authentication target data BD, which is dedicated to a face image in which at least a part of the face is covered by a mask (step S42). On the other hand, if the result of the determination in step S41 shows that the reason why the authentication target data BD is inappropriate for the authentication operation is (iv) (step S41: No), the process proceeds to step S26.

[0070] A facial image in which at least a portion of the face is obscured by an obstruction is often inappropriate for use in facial recognition. For example, a facial image in which the eyes and the area around the eyes are obscured by sunglasses is considered inappropriate to be authenticated. However, because the rate of mask wearing is very high these days, there are many cases in which a facial image in which at least a portion of the face is obscured by a mask worn by a living body is obscured (hereinafter referred to as a "mask-wearing image"). Furthermore, frequently putting on and taking off a mask for biometric authentication is undesirable from the perspective of preventing the spread of infectious diseases. Note that the area obscured by the sunglasses may be an example of another area different from at least a portion of the predetermined area of the living body.

[0071] In response to this, development is underway to develop mechanisms that can accurately authenticate images of people wearing masks. For example, by performing machine learning using a large number of images of people wearing masks, it is possible to build an authentication model that can authenticate people from images of people wearing masks. [3-2: Technical Effects of Biometric Authentication Device 3]

[0072] In other words, even in an image of a person wearing a mask, accurate facial recognition can be achieved by using a dedicated mechanism. Therefore, by distinguishing between an image of a person wearing a mask and an image of a person covered by an object other than a mask, such as sunglasses, the image of a person wearing a mask can be accurately recognized.

[0073] The biometric authentication device in the third embodiment does not ultimately determine that an image of a person wearing a mask is inappropriate for authentication, but instead authenticates the person wearing a mask using a mechanism suitable for authenticating such images. This allows biometric authentication of the person to be authenticated without the person removing the mask.

[0074] For example, when the police use facial recognition, it is essential to prevent false positives, such as identifying an unrelated person as the person of interest. However, there have been cases where false positives have led to the arrest of unrelated people, and so it is essential to take action. The above-described biometric authentication device can accurately determine whether or not the data to be authenticated is inappropriate for authentication operations, and can prevent erroneous authentication from occurring. [4: Note]

[0075] The following additional notes are provided regarding the above-described embodiment. [Appendix 1] extraction means for extracting an authentication target feature, which is a feature of authentication target data used for performing authentication operation; a calculation means for calculating an inappropriateness score indicating a degree of similarity between an inappropriate feature set based on each feature of a plurality of sample data items that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness and the feature to be authenticated; a determination means for determining whether the data to be authenticated is inappropriate for the authentication operation according to the inappropriateness score; A biometric authentication device comprising: [Appendix 2] the inappropriate feature amount includes at least a first inappropriate feature amount set based on the feature amount of each of a plurality of first sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a first type of authentication, and a second inappropriate feature amount set based on the feature amount of each of a plurality of second sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a second type of authentication that is different from the first type; the calculation means calculates, as the inappropriateness scores, a first inappropriateness score indicating a similarity between the first inappropriate feature and the feature to be authenticated, and a second inappropriateness score indicating a similarity between the second inappropriate feature and the feature to be authenticated; The determination means determines whether the authentication data is inappropriate for the authentication operation according to the first and second inappropriateness scores. 10. The biometric authentication device of claim 1. [Appendix 3] The determining means determines the reason why the authentication data is inappropriate for the authentication operation according to the first and second inappropriateness scores. 10. The biometric authentication device of claim 2. [Appendix 4] The determination means determining that the reason the data to be authenticated is unsuitable for the authentication operation is the first type of unsuitability when the first and second inappropriate scores indicate that the similarity between the first inappropriate feature and the feature to be authenticated is higher than the similarity between the second inappropriate feature and the feature to be authenticated; When the first and second inappropriate scores indicate that the similarity between the second inappropriate feature and the feature to be authenticated is higher than the similarity between the first inappropriate feature and the feature to be authenticated, it is determined that the reason why the data to be authenticated is inappropriate for the authentication operation is the second type of inappropriateness. 4. The biometric authentication device of claim 3. [Appendix 5] The authentication method further includes a presentation unit that presents a reason why the data to be authenticated is inappropriate for the authentication operation. 5. The biometric authentication device according to claim 3 or 4. [Appendix 6] the extraction means extracts the feature to be authenticated from the data to be authenticated using a feature generation model; The inappropriate feature is set based on the feature extracted from the plurality of sample data using the feature generation model. 6. A biometric authentication device according to any one of appendices 1 to 5. [Appendix 7] the calculation means calculates a matching score indicating a degree of similarity between the feature to be authenticated and a pre-registered matching feature; The method further includes a correction means for correcting the matching score when the determination means determines that the data to be authenticated is inappropriate for authentication. 5. A biometric authentication device according to any one of appendices 1 to 4. [Appendix 8] The extraction means extracts the matching feature, The method further includes a storage means for storing the matching feature amount. 6. The biometric authentication device of claim 5. [Appendix 9] each of the target data and the plurality of sample data is a biometric image including a living body that is a target of the authentication operation; The authentication inappropriateness includes at least one of the following inappropriateness: the biometrics appearing in the biometric image with at least a part of the biometrics blurred; the biometrics appearing in the biometric image with at least a part of the biometrics obscured by an obstruction; and the biometrics image having noise superimposed thereon. 9. A biometric authentication device according to any one of appendices 1 to 8. [Appendix 10] the target data and the plurality of sample data are each a face image including a face of a living body that is a target of authentication operation, The authentication inappropriateness includes at least one of the inappropriateness that the face is captured in the face image when at least a part of the face is covered by a mask worn by the living body, and the inappropriateness that the face is captured in the face image when at least a part of the face is covered by an obstruction other than the mask. 10. A biometric authentication device according to any one of appendices 1 to 9. [Appendix 11] each of the data to be authenticated and the plurality of sample data is an image in which a target of the authentication operation is captured; The authentication inappropriateness includes at least one of a third type of authentication inappropriateness in which the object is captured in the image while at least a part of a predetermined region of the object is obscured, and a fourth type of authentication inappropriateness in which the object is captured in the image while at least a part of a region of the object other than the predetermined region is obscured, When the determination means determines that the reason why the authenticated data is inappropriate for the authentication operation is the third type of authentication inappropriateness, the determination means determines that the authenticated data is not inappropriate for the authentication operation. 11. A biometric authentication device according to any one of appendices 1 to 10. [Appendix 12] the subject is a living organism, The predetermined area includes the area surrounding the living body. [Appendix 13] The inappropriate feature amount is an average value of the feature amounts of each of the plurality of sample data. 13. The biometric authentication device according to any one of Supplementary Notes 1 to 12. [Appendix 14] extracting a feature to be authenticated, which is a feature of the data to be authenticated used for performing an authentication operation; calculating an inappropriateness score indicating a degree of similarity between an inappropriate feature set based on the feature of each of a plurality of sample data items that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness and the feature to be authenticated; It is determined whether the data to be authenticated is inappropriate for the authentication operation according to the inappropriateness score. Biometric authentication methods. [Appendix 15] On the computer, extracting a feature to be authenticated, which is a feature of the data to be authenticated used for performing an authentication operation; calculating an inappropriateness score indicating a degree of similarity between an inappropriate feature set based on the feature of each of a plurality of sample data items that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness and the feature to be authenticated; It is determined whether the data to be authenticated is inappropriate for the authentication operation according to the inappropriateness score. A recording medium on which a computer program for executing a biometric authentication method is recorded.

[0076] At least some of the constituent elements of each of the above-described embodiments can be appropriately combined with at least some of the other constituent elements of each of the above-described embodiments. Some of the constituent elements of each of the above-described embodiments may not be used. Furthermore, to the extent permitted by law, the disclosures of all documents (e.g., published patent applications) cited in this disclosure are incorporated by reference as part of the description of this disclosure.

[0077] This disclosure may be modified as appropriate within the scope of the claims and the technical idea that can be read from the entire specification. The biometric authentication device, biometric authentication method, and recording medium that involve such modifications are also included in the technical idea of this disclosure. [Explanation of symbols]

[0078] 1,2 Biometric authentication device 11,211 Extraction part 12,212 Calculation section 13,213 Judgment section 214 Presentation section 215 Correction Unit GM feature generation model BD Authentication Data BC Authentication feature WC Inappropriate features WS Inappropriate Score 1WC First inappropriate feature 2WC Second inappropriate feature 1WS First Inappropriate Score 2WS Second Inappropriate Score CD matching data CC matching feature CS Matching Score

Claims

1. extraction means for extracting from the data to be authenticated features which are features used in the authentication operation; a calculation means for calculating an inappropriateness score indicating a degree of similarity between an inappropriate feature set based on each feature of a plurality of sample data items that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness and the feature to be authenticated; a determination means for determining whether the data to be authenticated is inappropriate for the authentication operation according to the inappropriateness score; Equipped with the inappropriate feature amount includes at least a first inappropriate feature amount set based on the feature amount of each of a plurality of first sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a first type of authentication, and a second inappropriate feature amount set based on the feature amount of each of a plurality of second sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a second type of authentication that is different from the first type; the calculation means calculates, as the inappropriateness scores, a first inappropriateness score indicating a similarity between the first inappropriate feature and the feature to be authenticated, and a second inappropriateness score indicating a similarity between the second inappropriate feature and the feature to be authenticated, The determination means determines whether the authentication data is inappropriate for the authentication operation according to the first and second inappropriateness scores. Biometric authentication device.

2. The determining means determines the reason why the authentication data is inappropriate for the authentication operation according to the first and second inappropriateness scores. The biometric authentication device according to claim 1 .

3. The determination means determining that the reason the data to be authenticated is unsuitable for the authentication operation is the first type of unsuitability when the first and second inappropriate scores indicate that the similarity between the first inappropriate feature and the feature to be authenticated is higher than the similarity between the second inappropriate feature and the feature to be authenticated; When the first and second inappropriate scores indicate that the similarity between the second inappropriate feature and the feature to be authenticated is higher than the similarity between the first inappropriate feature and the feature to be authenticated, it is determined that the reason why the data to be authenticated is inappropriate for the authentication operation is the second type of inappropriateness. The biometric authentication device according to claim 2 .

4. The authentication method further comprises: a presentation unit for presenting a reason why the data to be authenticated is inappropriate for the authentication operation. The biometric authentication device according to claim 2 or 3.

5. the extraction means extracts the feature to be authenticated from the data to be authenticated using a feature generation model; The inappropriate feature is set based on the feature extracted from the plurality of sample data using the feature generation model. The biometric authentication device according to claim 1 .

6. the calculation means calculates a matching score indicating a degree of similarity between the feature to be authenticated and a pre-registered matching feature; The method further includes a correction means for correcting the matching score when the determination means determines that the data to be authenticated is inappropriate for authentication. The biometric authentication device according to claim 1 .

7. The extraction means extracts the matching feature, The method further includes a storage means for storing the matching feature amount. The biometric authentication device according to claim 6.

8. Extracting features to be authenticated, which are features to be used in authentication operations, from the data to be authenticated; calculating an inappropriateness score indicating a degree of similarity between an inappropriate feature set based on the feature of each of a plurality of sample data items that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness and the feature to be authenticated; It is determined whether the data to be authenticated is inappropriate for the authentication operation according to the inappropriateness score.

1. A computer-implemented biometric authentication method comprising: the inappropriate feature amount includes at least a first inappropriate feature amount set based on the feature amount of each of a plurality of first sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a first type of authentication, and a second inappropriate feature amount set based on the feature amount of each of a plurality of second sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a second type of authentication that is different from the first type; calculating, as the inappropriateness scores, a first inappropriateness score indicating a similarity between the first inappropriate feature and the feature to be authenticated, and a second inappropriateness score indicating a similarity between the second inappropriate feature and the feature to be authenticated; determining whether the authentication data is inappropriate for the authentication operation according to the first and second inappropriateness scores; Biometric authentication methods.

9. Extracting features to be authenticated, which are features to be used in authentication operations, from the data to be authenticated; calculating an inappropriateness score indicating a degree of similarity between an inappropriate feature set based on the feature of each of a plurality of sample data items that deteriorate the accuracy of the authentication operation due to having the same type of authentication inappropriateness and the feature to be authenticated; It is determined whether the data to be authenticated is inappropriate for the authentication operation according to the inappropriateness score. A biometric authentication method, comprising: the inappropriate feature amount includes at least a first inappropriate feature amount set based on the feature amount of each of a plurality of first sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a first type of authentication, and a second inappropriate feature amount set based on the feature amount of each of a plurality of second sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a second type of authentication that is different from the first type; calculating, as the inappropriateness scores, a first inappropriateness score indicating a similarity between the first inappropriate feature and the feature to be authenticated, and a second inappropriateness score indicating a similarity between the second inappropriate feature and the feature to be authenticated; determining whether the authentication data is inappropriate for the authentication operation according to the first and second inappropriateness scores; A computer program for causing a computer to execute a biometric authentication method.

10. extraction means for extracting from the data to be authenticated features which are features used in the authentication operation; a calculation means for calculating a similarity between an inappropriate feature set based on a feature of sample data having authentication inappropriateness and the feature to be authenticated; a determination means for determining whether the data to be authenticated has the authentication inappropriateness based on the similarity; Equipped with the inappropriate feature amount includes at least a first inappropriate feature amount set based on the feature amount of each of a plurality of first sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a first type of authentication, and a second inappropriate feature amount set based on the feature amount of each of a plurality of second sample data items that deteriorates the accuracy of the authentication operation due to the inappropriateness of a second type of authentication that is different from the first type; the calculation means calculates, as the similarity, a first similarity which is a similarity between the first inappropriate feature and the feature to be authenticated, and a second similarity which is a similarity between the second inappropriate feature and the feature to be authenticated; The determining means determines whether the data to be authenticated is inappropriate for the authentication operation according to the first and second similarities. Biometric authentication device.

Images