Information processing device, information processing method, and program

Abstract

An information processing device according to the present disclosure comprises: an acquiring unit for acquiring a plurality of items of data each classified in advance into any one of a plurality of classes; a calculating unit for calculating distribution data of a first Mahalanobis distance of the plurality of items of data; and a normalizing unit for normalizing a second Mahalanobis distance between input target data and the plurality of items of data, on the basis of the distribution data.

Description

[Technical Field]

[0001] This disclosure includes information processing equipment, information processing methods, and program Regarding. [Background technology]

[0002] Dimensionality reduction is sometimes performed in machine learning and other processing that handles high-dimensional data. In such applications, it is desirable that the data is appropriately separated according to class after dimensionality reduction. Patent Document 1 discloses an example of a method for generating projection matrices used for dimensionality reduction. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2010-39778 [Overview of the project] [Problems that the invention aims to solve]

[0004] This disclosure aims to improve upon the technology described in the prior art documents mentioned above. [Means for solving the problem]

[0005] According to one aspect of this disclosure, the system includes: an acquisition unit that acquires a plurality of data, each pre-classified into one of a plurality of classes; a calculation unit that calculates distribution data of a first Mahalanobis distance in the plurality of data; and a normalization unit that normalizes the second Mahalanobis distance between the input target data and the plurality of data based on the distribution data. The distribution data is an inter-class variance, and the normalization unit uses the minimum value of the first Mahalanobis distance in the inter-class variance for the normalization. An information processing device is provided.

[0006] According to another aspect of this disclosure, the method includes the steps of: acquiring a plurality of data, each pre-classified into one of a plurality of classes; calculating distribution data of a first Mahalanobis distance in the plurality of data; and normalizing a second Mahalanobis distance between an input target data and the plurality of data based on the distribution data. The computer executes Information processing methods The information processing method wherein the distribution data is an interclass variance, and in the normalization step, the minimum value of the first Mahalanobis distance in the interclass variance is used for the normalization. It will be provided.

[0007] According to another aspect of this disclosure, a program to cause a computer to perform the steps of: acquiring a plurality of data, each pre-classified into one of a plurality of classes; calculating distribution data of a first Mahalanobis distance in the plurality of data; and normalizing a second Mahalanobis distance between an input target data and the plurality of data based on the distribution data. The distribution data is an interclass variance, and in the normalization step, the minimum value of the first Mahalanobis distance in the interclass variance is used for the normalization, the program It will be provided. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing the hardware configuration of the information processing device according to the first embodiment. [Figure 2] This is a functional block diagram of the information processing device according to the first embodiment. [Figure 3A] This diagram illustrates the relationship between probability distribution data in known classes and the thresholds used for classification. [Figure 3B] This diagram illustrates the relationship between probability distribution data in known classes and the thresholds used for classification. [Figure 3C] This figure shows the normalized state of the probability distribution data shown in Figure 3A. [Figure 3D] Figure 3B shows the normalized state of the probability distribution data. [Figure 4] This is a flowchart illustrating the general process performed in the information processing device according to the first embodiment. [Figure 5] This is a functional block diagram of the information processing device according to the second embodiment. [Figure 6] It is a flowchart showing an outline of processing performed in the information processing apparatus according to the second embodiment. [Figure 7] It is a functional block diagram of an information processing apparatus according to the third embodiment. [Figure 8] It is a flowchart showing an outline of processing performed in the information processing apparatus according to the third embodiment. [Figure 9] It is a schematic diagram showing the overall configuration of an information processing system according to the fourth embodiment. [Figure 10] It is a block diagram showing an example of the hardware configuration of an earphone control apparatus according to the fourth embodiment. [Figure 11] It is a functional block diagram of an earphone and an information processing apparatus according to the fourth embodiment. [Figure 12] It is a flowchart showing an outline of biometric authentication processing performed by the information processing apparatus according to the fourth embodiment. [Figure 13] It is a functional block diagram of an information processing apparatus according to the fifth embodiment.

Modes for Carrying Out the Invention

[0009] Hereinafter, exemplary embodiments of this disclosure will be described with reference to the drawings. In the drawings, the same elements or corresponding elements are denoted by the same reference numerals, and the description thereof may be omitted or simplified.

[0010] [First Embodiment] The information processing apparatus of this embodiment is an apparatus that determines the class to which the input unknown data belongs. This data can be, for example, feature amount data extracted from biometric information. The information processing apparatus of this embodiment is assumed to be a biometric authentication apparatus that performs personal identification or the like based on biometric information, but is not limited thereto.

[0011] Figure 1 is a block diagram showing an example of the hardware configuration of the information processing device 1 according to this embodiment. The information processing device 1 may be a computer such as a PC (Personal Computer), a processing server, a smartphone, or a microcomputer. The information processing device 1 includes a processor 101, memory 102, communication I / F (Interface) 103, input device 104, and output device 105. The various parts of the information processing device 1 are interconnected via buses, wiring, drive devices, etc., which are not shown.

[0012] The processor 101 is a processing unit that includes one or more arithmetic processing circuits, such as a CPU (Central Processing Unit), GPU (Graphics Processing Unit), FPGA (Field-Programmable Gate Array), ASIC (Application Specific Integrated Circuit), or TPU (Tensor Processing Unit). The processor 101 performs predetermined calculations according to a program stored in memory 102, etc., and also has the function of controlling each part of the information processing unit 1.

[0013] The memory 102 may include a volatile storage medium that provides a temporary memory area necessary for the operation of the processor 101, and a non-volatile storage medium that non-temporarily stores information such as data to be processed and the operating program of the information processing device 1. An example of a volatile storage medium is RAM (Random Access Memory). Examples of non-volatile storage media include ROM (Read Only Memory), HDD (Hard Disk Drive), SSD (Solid State Drive), and flash memory.

[0014] Communication I / F103 is a communication interface based on standards such as Ethernet (registered trademark), Wi-Fi (registered trademark), and Bluetooth (registered trademark). Communication I / F103 is a module for communicating with other devices such as data servers and sensor devices.

[0015] The input device 104 is a keyboard, pointing device, buttons, etc., used by the user to operate the information processing device 1. Examples of pointing devices include a mouse, trackball, touch panel, pen tablet, etc. The input device 104 may also include sensor devices such as a camera and microphone. These sensor devices can be used to acquire biometric information.

[0016] The output device 105 is, for example, a device that presents information to the user, such as a display device or a speaker. The input device 104 and the output device 105 may be integrally formed as a touch panel.

[0017] In Figure 1, the information processing device 1 is composed of a single device, but the configuration of the information processing device 1 is not limited to this. For example, the information processing device 1 may be a system composed of multiple devices. In addition, other devices may be added to the information processing device 1, and some devices may not be provided. Furthermore, some devices may be replaced with other devices having similar functions. Moreover, some functions of this embodiment may be provided by other devices via a network, and the functions of this embodiment may be implemented by distributing them across multiple devices. For example, memory 102 may include cloud storage, which is a storage device provided in another device. Thus, the hardware configuration of the information processing device 1 can be changed as appropriate.

[0018] Figure 2 is a functional block diagram of the information processing device 1 according to this embodiment. The information processing device 1 includes a first data storage unit 110, a statistical calculation unit 111, a second data storage unit 112, a minimum value calculation unit 113, a target data storage unit 114, a distance calculation unit 115, a normalization unit 116, and a determination unit 117.

[0019] The processor 101 performs predetermined arithmetic processing by executing a program stored in the memory 102. The processor 101 also controls the memory 102, communication I / F 103, input device 104, and output device 105 based on the program. Through these actions, the processor 101 realizes the functions of the statistical calculation unit 111, the minimum value calculation unit 113, the distance calculation unit 115, the normalization unit 116, and the determination unit 117. The memory 102 realizes the functions of the first data storage unit 110, the second data storage unit 112, and the target data storage unit 114.

[0020] Note that some of the functional blocks shown in Figure 2 may be provided in external devices to the information processing device 1, or may be realized through the cooperation of multiple devices.

[0021] The first data storage unit 110 stores a dataset of feature data belonging to a known first class (first dataset). In this embodiment, the first class is the class to which the feature data of the first person belongs. The feature data of the first person is also referred to as "person data".

[0022] The statistical calculation unit 111 calculates the within-class variance in the first class as a statistic from the first dataset obtained from the first data storage unit 110. Specifically, the statistical calculation unit 111 compares the personal data of the first person with the personal data of multiple individuals belonging to the first class, and calculates the distribution of Mahalanobis distances between the personal data.

[0023] The second data storage unit 112 stores a dataset of feature data (second dataset) belonging to a known second class that is different from the first class. In this embodiment, the second class is the class to which the feature data of a second person different from the first person belongs. The feature data of the second person is also called "other person data". The second data storage unit 112 stores the second class for each of the N (N is a natural number of 2 or more) second people other than the first person.

[0024] The minimum value calculation unit 113 calculates the inter-class variance in the second dataset obtained from the second data storage unit 112 based on the statistic (intra-class variance) calculated by the statistic calculation unit 111. The minimum value calculation unit 113 calculates the minimum value of the Mahalanobis distance in the inter-class variance. Specifically, the minimum value calculation unit 113 compares the first person's own data with the data of multiple other people belonging to the second class, and calculates the distribution data of the Mahalanobis distance between the person's data and the other people's data.

[0025] The target data storage unit 114 stores unknown characteristic data of the subject of judgment processing such as class classification as target data.

[0026] When the distance calculation unit 115 obtains the target data from the target data storage unit 114, it calculates the Mahalanobis distance of the target data based on the statistics (intraclass variance) calculated by the statistics calculation unit 111. Specifically, the distance calculation unit 115 compares the input target data with the first class's individual data and calculates the distribution data of the Mahalanobis distance between the target data and the individual data.

[0027] The normalization unit 116 normalizes the Mahalanobis distance of the target data based on the minimum value of the Mahalanobis distance in the interclass variance calculated by the minimum value calculation unit 113. Specific examples of the normalization method will be described later.

[0028] The determination unit 117 compares the Mahalanobis distance of the normalized target data with a predetermined threshold and makes a predetermined determination about the target data. For example, if the determination unit 117 determines a class classification, this determination determines the class to which the input target data belongs. Alternatively, if the determination unit 117 determines a person identification in biometric matching, this determination determines whether the person who acquired the target data is the same person as a registered person. The threshold is set to the same value across multiple classes.

[0029] Figures 3A and 3B illustrate the relationship between probability distribution data in known classes and the thresholds used for classification. Figure 3C shows the normalized version of the probability distribution data shown in Figure 3A. Figure 3D shows the normalized version of the probability distribution data shown in Figure 3B. The solid lines A1 and A2 in the figures represent the probability distribution data of the within-class variance. The dashed lines B1 and B2 represent the probability distribution data of the between-class variance. In each figure, the vertical axis represents probability, and the horizontal axis represents the Mahalanobis distance.

[0030] In Figure 3A, a threshold TH1 is set near the intersection of the probability distribution data A1 for the intraclass variance and the probability distribution data B1 for the interclass variance in the known class CL1. This allows for the comparison of the Mahalanobis distance calculated for the target data of the authenticated person with the threshold TH1 to determine whether or not the target data belongs to class CL1.

[0031] Similarly, in Figure 3B, a threshold TH2 is set near the intersection of the probability distribution data A2 for the intraclass variance and the probability distribution data B2 for the interclass variance in the known class CL2. This allows us to compare the Mahalanobis distance calculated for the target data of the authenticated person with the threshold TH2 to determine whether the target data belongs to class CL2 or not. However, if a common threshold TH1 is used as the threshold for determination in classes CL1 and CL2, it becomes impossible to perform determination in class CL2 with high accuracy.

[0032] Therefore, the information processing device 1 (normalization unit 116) according to this embodiment normalizes the probability distribution data. This makes it possible to determine the target data with high accuracy using a common threshold TH for any of the multiple classes.

[0033] Figure 3C shows that the probability distribution data A1 for the intraclass variance and the probability distribution data B1 for the interclass variance in class CL1 are normalized, and a threshold TH is set for the normalized data. The threshold TH can be set corresponding to the threshold TH1 in Figure 3A.

[0034] Similarly, Figure 3D shows that the probability distribution data A2 for the intraclass variance and the probability distribution data B2 for the interclass variance in class CL2 are normalized, and a common threshold TH is set for the normalized data.

[0035] Figure 4 is a flowchart illustrating the process performed in the information processing device according to the first embodiment. The operation of the information processing device 1 will be explained with reference to Figure 4.

[0036] In step S101, the statistical calculation unit 111 acquires the first data set from the first data storage unit 110.

[0037] In step S102, the statistics calculation unit 111 calculates statistics from the first dataset. Specifically, the statistics calculation unit 111 calculates the within-class variance in the first dataset as a statistic.

[0038] In step S103, the minimum value calculation unit 113 acquires the second data set from the second data storage unit 112.

[0039] In step S104, the minimum value calculation unit 113 calculates the minimum value of the Mahalanobis distance in the interclass variance based on the second dataset obtained in step S103 and the statistics calculated by the statistics calculation unit 111.

[0040] In step S105, the distance calculation unit 115 acquires the target data from the target data storage unit 114.

[0041] In step S106, the distance calculation unit 115 calculates the Mahalanobis distance of the target data based on the statistic (intraclass variance) calculated by the statistic calculation unit 111.

[0042] In step S107, the normalization unit 116 normalizes the Mahalanobis distance of the target data based on the minimum value calculated by the minimum value calculation unit 113.

[0043] The normalization method in the normalization unit 116 of this embodiment will be described below. Here, D is the Mahalanobis distance of the target data before normalization, normD is the Mahalanobis distance of the target data after normalization, and difMinD is the minimum Mahalanobis distance in the interclass variance of known other datasets (second class). In this case, the normalization unit 116 performs the following processing. normD = 101 - 100^(D / difMinD) if( normD < 0 ) {normD = 0}

[0044] For example, if D=300 and difMinD=150, normD is calculated as follows. normD = 101 - 100 (300 / 150) = 101 - 100 2 = -9899 Since normD < 0, normD becomes 0. In this embodiment, the normalized Mahalanobis distance can take values ​​between 0 and 100.

[0045] In step S108, the determination unit 117 makes a determination on the target data based on the normalized Mahalanobis distance.

[0046] In step S109, the determination unit 117 outputs the determination result. The destination of the determination result may be the memory 102 within the information processing device 1, or it may be another device.

[0047] The Mahalanobis distance between the data of the person being authenticated and the distribution of multiple data points within a known class differs depending on whether the person being authenticated is the person in question or not. Specifically, the Mahalanobis distance is short when the person being authenticated is the person in question and long when the person being authenticated is someone else. However, for example, if the spread of the distribution of the person's data is very small, the Euclidean distance might show a small deviation, allowing the person to be identified as the person in question, but the Mahalanobis distance would be large, potentially leading to a false identification even if the person is indeed the person in question. However, authentication systems have been using the same threshold for multiple classes to determine whether the target data belongs to a known class, i.e., to authenticate whether the person being authenticated is the person in question or not. A threshold set uniformly across classes may not be appropriate for a particular class, potentially reducing authentication accuracy.

[0048] In contrast, in this embodiment, the minimum inter-class variance in known classes is calculated, and the Mahalanobis distance of the target data is normalized based on this minimum value, making it possible to determine the target data using a common threshold across multiple classes. This provides an information processing device 1 that can improve the authentication accuracy of data across multiple classes.

[0049] Furthermore, in normalization, by using the minimum inter-class variance in known classes to which other people's data belongs, it becomes possible to minimize the rate of acceptance of other people's data when the target data is actually one's own data.

[0050] [Second Embodiment] The second embodiment will now be described. Since this embodiment is a modification of the first embodiment, the explanation of elements similar to those in the first embodiment may be omitted or simplified.

[0051] Figure 5 is a functional block diagram of the information processing device 1 according to this embodiment. As shown in Figure 5, the information processing device 1 according to this embodiment includes a first data storage unit 110, a statistical calculation unit 111, a maximum value calculation unit 118, a target data storage unit 114, a distance calculation unit 115, a normalization unit 116, and a determination unit 117. Unlike the first embodiment, the information processing device 1 according to this embodiment includes a maximum value calculation unit 118 instead of a second data storage unit 112 and a minimum value calculation unit 113.

[0052] The maximum value calculation unit 118 calculates the maximum value of the Mahalanobis distance in the statistic (intraclass variance) calculated by the statistic calculation unit 111. Specifically, the maximum value calculation unit 118 compares the personal data of the first person belonging to the first class with the personal data of other individuals belonging to the first class, and calculates the distribution data of the Mahalanobis distance between the personal data.

[0053] Figure 6 is a flowchart illustrating the process performed in the information processing device 1 according to this embodiment. The operation of the information processing device 1 will be explained with reference to Figure 6.

[0054] In step S201, the statistical calculation unit 111 acquires the first data set from the first data storage unit 110.

[0055] In step S202, the statistics calculation unit 111 calculates statistics from the first dataset. Specifically, the statistics calculation unit 111 calculates the within-class variance in the first dataset as a statistic.

[0056] In step S203, the maximum value calculation unit 118 calculates the maximum value of the Mahalanobis distance in the within-class variance of the first dataset based on the statistics calculated by the statistics calculation unit 111.

[0057] In step S204, the distance calculation unit 115 acquires the target data from the target data storage unit 114.

[0058] In step S205, the distance calculation unit 115 calculates the Mahalanobis distance of the target data based on the statistic (intraclass variance) calculated by the statistic calculation unit 111.

[0059] In step S206, the normalization unit 116 normalizes the Mahalanobis distance of the target data based on the maximum value calculated by the maximum value calculation unit 118.

[0060] The normalization method in the normalization unit 116 of this embodiment will be described below. Here, D is the Mahalanobis distance of the target data before normalization, normD is the Mahalanobis distance of the target data after normalization, and selfMaxD is the maximum value of the Mahalanobis distance in the intraclass variance of the known self dataset (first class). In this case, the normalization unit 116 performs the following processing. normD = 101 - 100^(D / selfMaxD) if( normD < 0 ) {normD = 0}

[0061] For example, if D=150 and selfMaxD=200, normD is calculated as follows. normD = 101 - 100 (150 / 200) = 101 - 100 0.75 ≈ 101 - 31.6 = 69.4

[0062] In step S207, the determination unit 117 makes a determination on the target data based on the normalized Mahalanobis distance.

[0063] In step S208, the determination unit 117 outputs the determination result. The destination of the determination result may be the memory 102 within the information processing device 1, or it may be another device.

[0064] As described above, in this embodiment, the maximum value of the intraclass variance in known classes is calculated, and the Mahalanobis distance of the target data is normalized based on this maximum value, making it possible to determine the target data using a common threshold across multiple classes. This provides an information processing device 1 that can improve the authentication accuracy of data across multiple classes.

[0065] Furthermore, in normalization, by using the maximum intraclass variance within the known class to which the personal data belongs, it becomes possible to minimize the rejection rate when the target data is personal data.

[0066] [Third Embodiment] The third embodiment will now be described. Since this embodiment is a modified version of the first or second embodiment, the explanation of elements similar to those in the first or second embodiment may be omitted or simplified.

[0067] Figure 7 is a functional block diagram of the information processing device 1 according to this embodiment. The information processing device 1 includes a first data storage unit 110, a statistical calculation unit 111, a second data storage unit 112, a minimum value calculation unit 113, a maximum value calculation unit 118, a target data storage unit 114, a distance calculation unit 115, a normalization unit 116, and a determination unit 117. The information processing device 1 according to this embodiment differs from the first embodiment in that it further includes a maximum value calculation unit 118.

[0068] Figure 8 is a flowchart illustrating the process performed in the information processing device 1 according to this embodiment. The operation of the information processing device 1 will be explained with reference to Figure 8.

[0069] In step S301, the statistical calculation unit 111 acquires the first data set from the first data storage unit 110.

[0070] In step S302, the statistics calculation unit 111 calculates statistics from the first dataset. Specifically, the statistics calculation unit 111 calculates the within-class variance in the first dataset as a statistic.

[0071] In step S303, the maximum value calculation unit 118 calculates the maximum value of the Mahalanobis distance in the within-class variance of the first dataset based on the statistics calculated by the statistics calculation unit 111.

[0072] In step S304, the minimum value calculation unit 113 acquires the second data set from the second data storage unit 112.

[0073] In step S305, the minimum value calculation unit 113 calculates the minimum value of the Mahalanobis distance in the interclass variance based on the acquired second dataset and the statistics calculated by the statistics calculation unit 111.

[0074] In step S306, the distance calculation unit 115 acquires the target data from the target data storage unit 114.

[0075] In step S307, the distance calculation unit 115 calculates the Mahalanobis distance of the target data based on the statistic (intraclass variance) calculated by the statistic calculation unit 111.

[0076] In step S308, the normalization unit 116 obtains parameters that have been set in advance by the administrator or manufacturer of the information processing device 1 from a storage device such as the memory 102.

[0077] In step S309, the normalization unit 116 calculates a reference value for the normalization process based on the acquired parameters, the minimum value calculated by the minimum value calculation unit 113, and the maximum value calculated by the maximum value calculation unit 118.

[0078] The normalization method in the normalization unit 116 of this embodiment will be described below. Here, D is the Mahalanobis distance of the target data before normalization, normD is the Mahalanobis distance of the target data after normalization, difMinD is the minimum value of the inter-class variance Mahalanobis distance for known other datasets (second class), selfMaxD is the maximum value of the intra-class variance Mahalanobis distance for known self datasets (first class), SetD is the baseline value, and X is the parameter value. Note that X can be any number between 0 and 1.0. In this case, the normalization unit 116 performs the following processing. SetD = X * selfMaxD + (1-X) * difMinD normD = 101 - 100^(D / SetD) if( normD < 0 ) {normD = 0}

[0079] For example, if X=1, SetD is calculated as follows: SetD = 1 * selfMaxD + (1-1) * difMinD = selfMaxD Furthermore, if D=150 and selfMaxD=200, normD is calculated as follows. normD = 101 - 100 (150 / 200) = 101 - 100 0.75 ≈101 - 31.6 = 69.4

[0080] Furthermore, if X=0, SetD is calculated as follows. SetD = 0 * selfMaxD + (1-0) * difMinD = difMinD

[0081] Furthermore, if D=300 and difMinD=150, normD is calculated as follows. normD = 101 - 100^(D / SetD) = 101 - 100^(D / difMinD) = 101 - 100 (300 / 150) = 101 - 100 2 = -9899 Since normD < 0, normD becomes 0.

[0082] Also, when X = 0.5, SetD is calculated as follows. SetD = 0.5 * selfMaxD + (1 - 0.5) * difMinD = 0.5 * selfMaxD + 0.5 * difMinD For example, when D = 300, selfMaxD = 200, and difMinD = 150, SetD and normD are calculated as follows. SetD = 0.5 * 200 + 0.5 *​​​​​​​​​​​​​​​​​​​​​​​​As described above, in this embodiment, the maximum value of intraclass variance in a known class and the minimum value of interclass variance in other known classes are calculated, and the Mahalanobis distance of the target data is normalized based on the minimum and maximum values. This makes it possible to determine the target data using a common threshold across multiple classes. An information processing device 1 is provided that can improve the authentication accuracy of data across multiple classes.

[0086] Furthermore, the weighting of the maximum value of the intraclass variance and the minimum value of the interclass variance can be changed using parameter values. This allows for flexible adjustment of the self-rejection rate and the misapplication acceptance rate in the information processing device 1.

[0087] [Fourth Embodiment] The fourth embodiment will be described below. In the fourth embodiment, as an example of the application of the information processing device 1 of the first to third embodiments, an information processing system that performs ear acoustic matching based on acoustic characteristics acquired by earphones will be illustrated. Ear acoustic matching is a technique that determines the similarity or differentness of a person by matching the acoustic characteristics of the person's head, including the external auditory canal. Since the acoustic characteristics of the external auditory canal differ from person to person, it is suitable for biometric information used for personal identification. For this reason, ear acoustic matching is sometimes used to identify users of hearable devices such as earphones. In addition to determining the similarity or differentness of a person, ear acoustic matching can also be used to determine the wearing state of a hearable device.

[0088] Figure 9 is a schematic diagram showing the overall configuration of the information processing system according to this embodiment. The information processing system comprises an information processing device 1 and an earphone 2 that can be wirelessly connected to each other.

[0089] The earphone 2 comprises an earphone control device 20, a speaker 26, and a microphone 27. The earphone 2 is an acoustic device that can be worn on the user's head, particularly the ear, and is typically a wireless earphone, wireless headset, etc. The speaker 26 functions as a sound wave generator that emits sound waves toward the user's ear canal when worn, and is positioned on the wearing side of the earphone 2. The microphone 27 is positioned on the wearing side of the earphone 2 so that it can receive sound waves reflected in the user's ear canal, etc., when worn. The earphone control device 20 controls the speaker 26 and microphone 27 and communicates with the information processing device 1.

[0090] In this specification, "sound" including sound waves and voices includes inaudible sounds whose frequency or sound pressure level is outside the range of audible sound.

[0091] The information processing device 1 is similar to the device described in the first to third embodiments. The information processing device 1 is, for example, a computer that is communicatively connected to the earphone 2 and performs biometric verification based on acoustic information. The information processing device 1 further controls the operation of the earphone 2, transmits audio data for generating sound waves emitted from the earphone 2, and receives audio data obtained from the sound waves received by the earphone 2. Specifically, when user 3 listens to music using the earphone 2, the information processing device 1 transmits compressed music data to the earphone 2. Also, when the earphone 2 is a telephone device for giving work instructions at an event venue, hospital, etc., the information processing device 1 transmits audio data of work instructions to the earphone 2. In this case, the earphone 2 may also transmit audio data of user 3's speech to the information processing device 1.

[0092] This overall configuration is merely an example; for instance, the information processing device 1 and the earphone 2 may be connected by a wire. Furthermore, the information processing device 1 and the earphone 2 may be configured as a single integrated device, and the information processing system may include other devices.

[0093] Figure 10 is a block diagram showing an example of the hardware configuration of the earphone control device 20. The earphone control device 20 includes a processor 201, memory 202, speaker I / F 203, microphone I / F 204, communication I / F 205, and battery 206. The various parts of the earphone control device 20 are interconnected via buses, wiring, drive devices, etc., which are not shown.

[0094] The descriptions of the processor 201, memory 202, and communication I / F 205 are omitted as they overlap with those of the first embodiment.

[0095] The speaker interface 203 is an interface for driving speaker 26. The speaker interface 203 includes a digital-to-analog conversion circuit, an amplifier, etc. The speaker interface 203 converts audio data into an analog signal and supplies it to speaker 26. As a result, speaker 26 emits sound waves based on the audio data.

[0096] The microphone interface 204 is an interface for acquiring signals from the microphone 27. The microphone interface 204 includes an analog-to-digital conversion circuit, an amplifier, etc. The microphone interface 204 converts the analog signal generated by the sound waves received by the microphone 27 into a digital signal. As a result, the earphone control device 20 acquires audio data based on the received sound waves.

[0097] Battery 206 is, for example, a rechargeable battery and supplies the power necessary for the operation of earphone 2. This allows earphone 2 to operate wirelessly without being connected to an external power source via a wire. If earphone 2 is connected via a wire, battery 208 may not be provided.

[0098] The hardware configuration shown in Figure 10 is illustrative, and other devices may be added, or some devices may be omitted. Furthermore, some devices may be replaced with other devices having similar functions. For example, the earphone 2 may be further equipped with input devices such as buttons to accept operation by user 3, and may be further equipped with display devices such as a display or indicator light to provide information to user 3. Thus, the hardware configuration shown in Figure 10 can be modified as appropriate.

[0099] Figure 11 is a functional block diagram of the earphone 2 and information processing device 1 according to this embodiment. The information processing device 1 includes a target data storage unit 114, a distance calculation unit 115, a normalization unit 116, a determination unit 117, a distance information storage unit 119, and an acoustic characteristics acquisition unit 120. The configuration of the block diagram of the earphone 2 is the same as in Figure 10, so its description is omitted. Among the functional blocks of the information processing device 1, the functions of the target data storage unit 114, the distance calculation unit 115, the normalization unit 116, and the determination unit 117 are the same as those described in the first embodiment.

[0100] The distance information storage unit 119 stores distance information such as the intraclass variance, interclass variance, the maximum Mahalanobis distance in the intraclass variance, and the minimum Mahalanobis distance in the interclass variance, calculated from known datasets. For this reason, in Figure 11, the block related to the function of normalizing the probability distribution for each Mahalanobis distance of known class datasets across classes is omitted from the diagram.

[0101] The acoustic characteristics acquisition unit 120 acquires frequency-domain acoustic characteristics based on sound waves propagating through the user's head. These acoustic characteristics may be, for example, a frequency spectrum obtained by converting a time-domain acoustic signal to the frequency domain using an algorithm such as the Fast Fourier Transform.

[0102] In Figure 11, some or all of the functions of the functional blocks described within the information processing device 1 may be provided in the earphone control device 20 instead of the information processing device 1. That is, each of the above functions may be implemented by the information processing device 1, by the earphone control device 20, or through the cooperation of the information processing device 1 and the earphone control device 20. In the following description, unless otherwise specified, each functional block related to the acquisition and determination of acoustic information will be provided within the information processing device 1, as shown in Figure 11.

[0103] Figure 12 is a flowchart illustrating the biometric verification process performed by the information processing device 1 according to this embodiment. The operation of the information processing device 1 will be explained with reference to Figure 12.

[0104] The biometric verification process shown in Figure 12 is executed, for example, when user 3 starts using the earphone 2 by operating it. Alternatively, the biometric verification process shown in Figure 12 may be executed at predetermined intervals when the power of the earphone 2 is turned on.

[0105] In step S401, the acoustic characteristics acquisition unit 120 instructs the earphone control device 20 to emit a test sound. The earphone control device 20 transmits a test signal to the speaker 26, and the speaker 26 emits a test sound generated based on the test signal into the ear canal of the user 3.

[0106] The test signal may include signals containing frequency components within a predetermined range, such as chirp signals, M-sequence (Maximum Length Sequence) signals, white noise, and impulse signals. This allows for the acquisition of an acoustic signal containing frequency information within a predetermined range. The test sound may be an audible sound with a frequency and sound pressure level within the audible range. In this case, the user 3 can be informed that a comparison is being made by making them perceive the sound wave during the comparison. Alternatively, the test sound may be an inaudible sound with a frequency or sound pressure level outside the audible range. In this case, the sound wave can be made less perceptible to the user 3, improving user comfort.

[0107] In step S402, the microphone 27 receives reflected sound (otoacoustics) from the ear canal and converts it into a time-domain electrical signal. This electrical signal is sometimes called an acoustic signal. The microphone 27 transmits the acoustic signal to the earphone control device 20, and the earphone control device 20 transmits the acoustic signal to the information processing device 1.

[0108] In step S403, the acoustic characteristic acquisition unit 120 acquires frequency-domain acoustic characteristics based on sound waves propagating through the user's head. These acoustic characteristics may be, for example, a frequency spectrum obtained by converting a time-domain acoustic signal to the frequency domain using an algorithm such as the Fast Fourier Transform.

[0109] In step S404, the target data storage unit 114 stores the acoustic characteristics acquired from the earphone control device 20 as target data for feature extraction.

[0110] In step S405, the distance calculation unit 115 retrieves target data from the target data storage unit 114. This target data is unknown data that is the subject of determination in this determination process.

[0111] In step S406, the distance calculation unit 115 calculates the Mahalanobis distance of the target data.

[0112] In step S407, the normalization unit 116 normalizes the Mahalanobis distance of the target data based on the distance information stored in the distance information storage unit 119. The distance information includes the intraclass variance, interclass variance, the maximum value of the Mahalanobis distance in the intraclass variance, and the minimum value of the Mahalanobis distance in the interclass variance, all calculated from known datasets. For example, if the maximum value of the Mahalanobis distance in the intraclass variance is used, normalization can be performed in the same manner as in the first embodiment.

[0113] In step S408, the determination unit 117 performs a determination process on the target data based on the normalized Mahalanobis distance. For example, if the determination in the determination unit 117 is class classification, this determination process determines the class to which the input target data belongs. Alternatively, for example, if the determination in the determination unit 117 is person identification in biometric matching, this determination process determines whether the person who acquired the target data is the same person as a registered person.

[0114] In step S409, the determination unit 117 outputs the determination result from step S408. This determination result is used, for example, to control whether the use of earphone 2 is permitted or not. The output destination may be the memory 102 in the information processing device 1, or it may be another device.

[0115] Although this embodiment describes an example of ear acoustic matching, it can be similarly applied to biometric matching using other biometric information. Examples of applicable biometric information include face, iris, fingerprints, palm prints, veins, voice, auricle, gait, etc.

[0116] According to this embodiment, since normalization is performed in the same manner as in the first to third embodiments, it becomes possible to determine target data using a common threshold across multiple classes. This provides an information processing device 1 that can improve the authentication accuracy of data across multiple classes in biometric authentication such as ear acoustic matching.

[0117] [Fifth Embodiment] Figure 13 is a functional block diagram of the information processing device 4 according to the fifth embodiment. The information processing device 4 comprises an acquisition unit 401, a calculation unit 402, and a normalization unit 403. The acquisition unit 401 acquires multiple data, each pre-classified into one of several classes. The calculation unit 402 calculates distribution data of the first Mahalanobis distance for the multiple data. The normalization unit 403 normalizes the second Mahalanobis distance between the input target data and the multiple data based on the distribution data.

[0118] According to this embodiment, an information processing device 4 is provided that can improve the authentication accuracy of data in multiple classes.

[0119] [Modified Embodiment] This disclosure is not limited to the embodiments described above and can be modified as appropriate without departing from the spirit of this disclosure. For example, an example in which some components of one embodiment are added to another embodiment, or in which some components of another embodiment are replaced, is also an embodiment of this disclosure.

[0120] In the embodiments described above, intra-class variance or inter-class variance is used exemplarily as an indicator of data dispersion, but any statistic other than variance may be used as an indicator of dispersion.

[0121] Each embodiment also includes a method for recording a program on a storage medium that operates the configuration of the embodiment to realize the functions of the embodiment described above, reading the program recorded on the storage medium as code, and executing it on a computer. In other words, a computer-readable storage medium is also included in the scope of each embodiment. Furthermore, not only the storage medium on which the above-mentioned program is recorded, but also the program itself is included in each embodiment. In addition, one or more components included in the above-described embodiment may be circuits such as ASICs or FPGAs configured to realize the functions of each component.

[0122] Examples of storage media that can be used include floppy disks, hard disks, optical disks, magneto-optical disks, CD (Compact Disk)-ROMs, magnetic tapes, non-volatile memory cards, and ROMs. Furthermore, the scope of each embodiment is not limited to programs that perform processing on the storage medium alone, but also includes programs that operate on an OS (Operating System) in cooperation with other software and the functions of expansion boards to perform processing.

[0123] The services realized by the functions of each of the above embodiments can also be provided to users in the form of SaaS (Software as a Service).

[0124] Furthermore, the embodiments described above are merely examples of concrete implementations of this disclosure, and the technical scope of this disclosure should not be interpreted as being limited by them. In other words, this disclosure can be implemented in various ways without departing from its technical concept or its main features.

[0125] Some or all of the embodiments described above may also be described as follows, but are not limited to the following.

[0126] (Note 1) An acquisition unit that acquires multiple data items, each of which is pre-classified into one of several classes, A calculation unit that calculates the distribution data of the first Mahalanobis distance in the aforementioned plurality of data, A normalization unit that normalizes the second Mahalanobis distance between the input target data and the multiple data points based on the distribution data, An information processing device equipped with the following features.

[0127] (Note 2) A determination unit determines which of the multiple classes the target data belongs to by comparing the normalized second Mahalanobis distance with a threshold set for the multiple classes. An information processing device as described in Appendix 1, further comprising the following:

[0128] (Note 3) The aforementioned distribution data is the inter-class variance, The normalization unit uses the minimum value of the first Mahalanobis distance in the interclass variance for the normalization. The information processing device described in Appendix 1.

[0129] (Note 4) The aforementioned distribution data is within-class variance, The normalization unit uses the maximum value of the first Mahalanobis distance in the intraclass variance for the normalization. The information processing device described in Appendix 1.

[0130] (Note 5) The aforementioned distribution data consists of within-class variance and between-class variance. The normalization unit uses the maximum value of the first Mahalanobis distance in the intraclass variance and the minimum value of the first Mahalanobis distance in the interclass variance for the normalization. The information processing device described in Appendix 1.

[0131] (Note 6) The normalization unit modifies the weighting of the maximum value and the minimum value in the normalization. The information processing device described in Appendix 5.

[0132] (Note 7) The aforementioned data consists of biometric information of a person. An information processing device as described in any of the appendices 1 to 6.

[0133] (Note 8) The aforementioned biometric information indicates the acoustic characteristics of the human ear canal. The information processing device described in Appendix 7.

[0134] (Note 9) The steps include: obtaining multiple data points, each pre-classified into one of several classes; The steps include calculating the distribution data of the first Mahalanobis distance in the aforementioned multiple data sets, The steps include: normalizing the second Mahalanobis distance between the input target data and the plurality of data based on the distribution data; An information processing method comprising the following:

[0135] (Note 10) On the computer, The steps include: obtaining multiple data points, each pre-classified into one of several classes; The steps include calculating the distribution data of the first Mahalanobis distance in the aforementioned multiple data sets, The steps include: normalizing the second Mahalanobis distance between the input target data and the plurality of data based on the distribution data; A storage medium containing a program to execute something.

[0136] This application claims priority based on Japanese Patent Application No. 2023-7782, filed on 23 January 2023, and incorporates all of its disclosures herein. [Explanation of Symbols]

[0137] 1.4 Information Processing Devices 2 earphones 3 users 20 Earphone control device 26 speakers 27 Microphone 101, 201 processors 102, 202 memory 103, 205 Communication I / F 104 Input device 105 Output device 110 First data storage unit 111 Statistics calculation section 112 Second Data Storage Unit 113 Minimum Value Calculation Unit 114 Target data storage unit 115 Distance Calculation Unit 116 Normalization section 117 Judgment section 118 Maximum value calculation unit 119 Distance information storage unit 120 Acoustic Characteristics Acquisition Unit 203 Speaker I / F 204 Microphone Interface 206 Battery

Claims

1. An acquisition unit that acquires multiple data items, each of which is pre-classified into one of several classes, A calculation unit that calculates the distribution data of the first Mahalanobis distance in the aforementioned plurality of data, A normalization unit that normalizes the second Mahalanobis distance between the input target data and the multiple data points based on the distribution data, Equipped with, The aforementioned distribution data is the interclass variance, The normalization unit is an information processing device that uses the minimum value of the first Mahalanobis distance in the interclass variance for the normalization.

2. A determination unit determines which of the multiple classes the target data belongs to by comparing the normalized second Mahalanobis distance with a threshold set for the multiple classes. The information processing apparatus according to claim 1, further comprising:

3. The aforementioned distribution data is within-class variance, The normalization unit uses the maximum value of the first Mahalanobis distance in the intraclass variance for the normalization. The information processing apparatus according to claim 1.

4. The aforementioned distribution data consists of within-class variance and between-class variance. The normalization unit uses the maximum value of the first Mahalanobis distance in the intraclass variance and the minimum value of the first Mahalanobis distance in the interclass variance for the normalization. The information processing apparatus according to claim 1.

5. The normalization unit modifies the weighting of the maximum value and the minimum value in the normalization. The information processing apparatus according to claim 4.

6. The aforementioned data consists of biometric information of a person. The information processing apparatus according to any one of claims 1 to 5.

7. The aforementioned biometric information indicates the acoustic characteristics of the human ear canal. The information processing apparatus according to claim 6.

8. The steps include: obtaining multiple data points, each pre-classified into one of several classes; A step of calculating the distribution data of the first Mahalanobis distance in the aforementioned plurality of data, The steps include: normalizing the second Mahalanobis distance between the input target data and the plurality of data based on the distribution data; An information processing method performed by a computer equipped with the following features: The aforementioned distribution data is the inter-class variance, An information processing method in which, in the normalization step, the minimum value of the first Mahalanobis distance in the interclass variance is used for the normalization.

9. On the computer, The steps include: obtaining multiple data points, each pre-classified into one of several classes; A step of calculating the distribution data of the first Mahalanobis distance in the aforementioned plurality of data, The steps include: normalizing the second Mahalanobis distance between the input target data and the plurality of data based on the distribution data; A program to execute, The aforementioned distribution data is the inter-class variance, A program that, in the normalization step, uses the minimum value of the first Mahalanobis distance in the interclass variance for the normalization.

Images