Personal identification device, method and program

Abstract

One aspect of this invention first generates and stores first user identification probability distribution information representing an identification probability distribution of a plurality of users to be identified. In this state, a surface electromyogram signal corresponding to a motion of an arbitrary user of the plurality of users is acquired, a feature amount is extracted from the surface electromyogram signal, and the motion of the arbitrary user is estimated from the extracted feature amount. Second user identification probability distribution information corresponding to the estimated motion is then obtained, third user identification probability distribution information corresponding to a motion transition representing the time-series change in a past plurality of motions including the estimated motion is obtained, and the first user identification probability distribution information is updated on the basis of the second user identification probability distribution information and the third user identification probability distribution information.

Description

[Technical field]

[0001] One aspect of the present invention relates to a personal identification device, method, and program for identifying an individual using, for example, a surface electromyogram of the skin. [Background technology]

[0002] Surface electromyography (EMG) is a bioelectric potential signal measured from electrodes attached to the skin, and has the characteristic that the amplitude increases with muscle activity, so it is used to measure the muscle activity of the user. Since surface electromyograms reflect muscle activity in this way, they are used as input information for intuitive control of robots, prosthetic limbs, powered suits, etc.

[0003] To control a robot or prosthetic arm based on the surface EMG, it is necessary to estimate what motion the user performed from the surface EMG. In order to identify various user motions, it is necessary to measure the surface EMG data of anticipated motions for each user in advance and build a model to estimate motions from the surface EMG.

[0004] Robots and prosthetic limbs controlled using surface electromyograms are now often used for rehabilitation purposes. However, hardware such as robots is often shared by multiple users rather than being prepared exclusively for each user. In such cases, the internal software, i.e. the model that estimates movements from surface electromyograms, needs to be switched for each user.

[0005] Therefore, a method based on a deep learning approach has been proposed to avoid having to manually switch between users. This method identifies users by estimating which user's gesture is the one from the muscle activity pattern when a specific gesture is input, and after the user is identified, a model dedicated to the identified user is used to perform processes from motion estimation to control of a robot or prosthetic hand.

[0006] For example, Non-Patent Document 1 proposes an algorithm that converts surface electromyogram signals into time-series frequency band features and uses deep learning to identify users. It has been reported that this method can achieve a user identification accuracy of 99.203%. [Prior art documents] [Non-patent literature]

[0007] [Non-Patent Document 1] L. Lu, J. Mao, W. Wang, G. Ding and Z. Zhang, "A Study of Personal Recognition Method Based on EMG Signal," in IEEE Transactions on Biomedical Circuits and Systems, vol. 14, no. 4, pp. 681-691, Aug. 2020, doi: 10.1109 / TBCAS.2020.3005148. Summary of the Invention [Problem to be solved by the invention]

[0008] However, the technology described in Non-Patent Document 1 uses the user identification result obtained from the estimation result of the user's specified gesture action performed before the start of use as it is until the end of use. For this reason, for example, when the number of users to be identified increases or the number of electrodes is reduced to make the sensor unit more compact, sufficient user identification accuracy may not be obtained only with the estimation result based on the user's specified gesture action at the start of use.

[0009] This invention has been made in light of the above-mentioned circumstances, and aims to provide a technology that can improve the accuracy of user identification even when the estimation results based on the user's specified actions do not provide sufficient accuracy in user identification. [Means for solving the problem]

[0010] In order to solve the above problem, one aspect of a personal identification device or method according to the present invention includes a first processing unit or process for generating and storing first user identification probability distribution information representing a identification probability distribution of multiple users to be identified, a second processing unit or process for acquiring a surface electromyogram signal corresponding to a motion of an arbitrary user among the multiple users and extracting a feature amount from the acquired surface electromyogram signal, a third processing unit or process for estimating the motion of the arbitrary user from the extracted feature amount, a fourth processing unit or process for determining second user identification probability distribution information corresponding to the estimated motion, a fifth processing unit or process for determining third user identification probability distribution information corresponding to a motion transition representing a time-series change of a plurality of past motions including the estimated motion, and a sixth processing unit or process for updating the first user identification probability distribution information based on the second user identification probability distribution information and the third user identification probability distribution information.

[0011] According to one aspect of the present invention, for example, a first user identification probability distribution registered based on a user's specified gesture motion is updated based on the user's motion estimated during the subsequent use of the device and the transition of past motions. Therefore, even if the accuracy of the user identification probability distribution registered based on the user's specified gesture motion is low, the first user identification probability distribution can be updated. As a result, the user identification result is correctly corrected during the use of the device, and it becomes possible to correctly switch the model for estimating the user's motion. Effect of the Invention

[0012] That is, according to one aspect of the present invention, it is possible to provide a technique that can improve the accuracy of user identification even when the estimation result based on the user's prescribed action does not provide sufficient accuracy in user identification. [Brief description of the drawings]

[0013] [Figure 1]FIG. 1 is a block diagram showing an example of a hardware configuration of a personal identification device according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a block diagram showing an example of the software configuration of the personal identification device according to the embodiment of the present invention. [Diagram 3] FIG. 3 is a flowchart showing the overall processing procedure and processing contents executed by the control unit of the personal identification device shown in FIG. [Figure 4] FIG. 4 is a flowchart showing an example of the procedure and contents of a user identification process based on a prescribed action among the entire process procedures shown in FIG. [Diagram 5] FIG. 5 is a flowchart showing an example of the procedure and contents of a process for estimating a user's action after starting use, among the entire process procedures shown in FIG. [Figure 6] FIG. 6 is a flowchart showing an example of the processing procedure and processing contents of the update processing of the user identification probability distribution after the start of use, among the entire processing procedures shown in FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0015] [One embodiment] (Configuration example) (1) Equipment FIG. 1 is a block diagram showing an example of a hardware configuration of a personal identification device PD according to an embodiment of the present invention, and FIG. 2 is a block diagram showing an example of a software configuration of the personal identification device PD.

[0016] The personal identification device PD has the functions of an information processing device such as a personal computer, and is either housed within the control device of the prosthetic hand or is provided outside the control device.

[0017] The personal identification device PD has a control unit 1 that uses a hardware processor such as a central processing unit (CPU), and this control unit 1 is connected via a bus 5 to a memory unit having a program memory unit 2 and a data memory unit 3, and an input / output interface (hereinafter the interface will be abbreviated as I / F) unit 4.

[0018] A sensor unit SS is connected to the input / output I / F section 4. The sensor unit SS measures a surface electromyogram (EMG) of the skin of the user US, and is, for example, a plurality of electrodes arranged at equal intervals in the circumferential direction of the circumferential surface of the forearm of the user US. The surface EMG signals measured by the plurality of electrodes are input to the input / output I / F section 4 of the personal identification device PD via, for example, a signal cable. The input / output I / F section 4 converts each of the surface EMG signals into a digital signal that can be processed by the control section 1.

[0019] The program storage unit 2 is configured by combining a non-volatile memory such as a solid state drive (SSD) that can be written and read at any time as a storage medium, and a non-volatile memory such as a read only memory (ROM), and stores application programs required for executing various control processes according to an embodiment, in addition to middleware such as an operating system (OS). Hereinafter, the OS and each application program will be collectively referred to as a program.

[0020] The data storage unit 3 is, for example, a combination of a non-volatile memory such as an SSD that can be written to and read from at any time as a storage medium, and a volatile memory such as a RAM (Random Access Memory), and is equipped with a measurement signal storage unit 30, an action probability estimation model database (hereinafter, database will be abbreviated as DB) 31, a user identification probability distribution DB 32, a first estimation model DB 33, and a second estimation model DB 34 as main storage units required to implement one embodiment.

[0021] The measurement signal storage unit 30 is used to temporarily store pre-processed surface electromyogram signals output from the sensor unit SS.

[0022] The motion probability estimation model DB 31 stores a plurality of motion probability estimation models prepared in advance corresponding to a plurality of users who share a prosthetic hand. Each motion probability estimation model has a function of outputting a corresponding motion probability as a response variable when a surface electromyogram signal is input as an explanatory variable.

[0023] The user identification probability distribution DB 32 stores a user identification probability distribution generated corresponding to each of the plurality of users. In one embodiment of the present invention, the user identification probability distribution is updated during a period of use by a user identification probability distribution update process described later.

[0024] The first estimation model DB 33 stores, as a first estimation model, a model for estimating a user identification probability distribution prepared in advance for each user, in order to estimate a user identification probability distribution from an estimated motion.

[0025] The second estimation model DB 34 stores, as a second estimation model, a model for estimating a user identification probability distribution prepared in advance for each user, in order to estimate a user identification probability distribution from past motion transitions.

[0026] The control unit 1 includes, as processing functions necessary for implementing one embodiment, a surface electromyogram acquisition processing unit 11, a preprocessing unit 12, a movement probability calculation processing unit 13, a movement estimation processing unit 14, a first user identification probability distribution calculation processing unit 15, a second user identification probability distribution calculation processing unit 16, and an update processing unit 17. All of these processing units 11 to 17 are realized by causing a hardware processor of the control unit 1 to execute an application program stored in the program storage unit 2.

[0027] The above application programs may be stored in advance in the program storage unit 2, or may be downloaded from a control device or the like when necessary and stored in the program storage unit 2.

[0028] The surface electromyogram acquisition processing unit 11 acquires the surface electromyogram signals output from each electrode of the sensor unit SS via the input / output I / F unit 4 , and passes the acquired surface electromyogram signals to the preprocessing unit 12 .

[0029] The preprocessing unit 12 performs a process of removing artifacts due to body movements on the digitized surface electromyogram signal and a process of extracting amplitude components. The preprocessed signal is then stored in the measurement signal storage unit 30. To remove the body movement artifacts, for example, a band-pass filter process of 20-500 Hz is applied. To extract the amplitude components, a filtering process is applied in which the signal is rectified and then extracted with a cutoff frequency of 4 Hz.

[0030] In addition, the processing by the pre-processing unit 12 is not limited to software processing, and may be realized by passing the surface electromyogram signal consisting of an analog signal through a band-pass filter having the above passband characteristics, rectifying it in a rectifier circuit, and then inputting it to a filter having the above cutoff frequency.

[0031] The action probability calculation processing unit 13 performs a process of estimating the action probability of each user for each predefined action type by using the action probability estimation model of each user stored in the action probability estimation model DB 31 and inputting the measurement signal for a certain period of time after the preprocessing to these action probability estimation models. An example of this action probability estimation process will be described in the operation example.

[0032] The motion estimation processing unit 14 performs processing to estimate the type of motion performed by the user, based on the motion probability of each user obtained by the motion probability calculation processing unit 13 and the user identification probability distribution of each user stored in the user identification probability distribution DB 32. An example of this motion estimation processing will also be described in the operation example.

[0033] The first user identification probability distribution calculation processing unit 15 inputs information representing the type of movement estimated by the movement estimation processing unit 14 into a first estimation model stored in a first estimation model DB 33, thereby executing a process of calculating a user identification probability distribution corresponding to the type of the estimated movement.

[0034] The second user identification probability distribution calculation processing unit 16 holds, as action transition information, time series changes of a plurality of actions performed in the past, including the above-mentioned estimated action.Then, the held action transition information is input to a second estimation model stored in a second estimation model DB 34, thereby performing a process of calculating a user identification probability distribution corresponding to the above-mentioned action transition.

[0035] The update processing unit 17 performs processing to update the current user identification probability distribution based on the current user identification probability distribution stored in the user identification probability distribution DB 32, the latest first user identification probability distribution obtained by the first user identification probability distribution calculation processing unit 15, and the second user identification probability distribution obtained by the second user identification probability distribution calculation processing unit 16, which reflects action transitions representing time-series changes in actions performed in the past.

[0036] (Example of operation) Next, an example of the operation of the personal identification device PD configured as above will be described.

[0037] In this example, the prosthetic hand that assists the movement of the user US's body part from the shoulder to the hand is taken as the control target, and the prosthetic hand is controlled from the position and posture of the user US in the resting state to grasp a specific object, and then the prosthetic hand is controlled until it returns to the position and posture of the resting state again. In this case, the four types of movements to be estimated are "palmar flexion" and "dorsiflexion" of the fingers, and "flexion" and "extension" of the elbow.

[0038] In addition, the users US who are expected to use the device are three persons, A, B, and C, and the explanation will be given on the assumption that any one of them will use the device.

[0039] FIG. 3 is a flowchart showing an example of the overall process executed by the control unit 1 of the personal identification device PD.

[0040] The control unit 1 of the personal identification device PD first performs a process of identifying a user based on a prescribed gesture action by the user when the user starts using the device in step S1, and then performs a process of estimating the user's actions at each stage during the subsequent period of use in step S2, and then performs a process of updating the user identification probability distribution based on the estimated actions in step S3.

[0041] (1) User identification based on specified gesture movements FIG. 4 is a flowchart showing an example of a user identification process based on the above-mentioned prescribed gesture motion, which is executed by the control unit 1 of the personal identification device PD.

[0042] (1-1) Acquisition of surface electromyogram signals After the user US attaches the electrodes of the sensor unit SS to the forearm, the user performs a prescribed gesture, for example, a so-called thumbs up with a clenched fist.

[0043] Then, surface electromyogram signals are output from each electrode of the sensor unit SS in response to the gesture movement, and these surface electromyogram signals are converted into digital signals at a fixed sampling interval by the input / output I / F unit 4 of the personal identification device PD. Under the control of the surface electromyogram acquisition processing unit 11, the control unit 1 of the personal identification device PD acquires each of the digitized surface electromyogram signals in step S11, and the acquired surface electromyogram signals are subjected to signal processing by the preprocessing unit 12. For example, as described above, body movement artifacts are removed by passing the signals through a 20-500 Hz bandpass filter, and further rectification and filtering processing with a cutoff frequency of 4 Hz are performed to extract amplitude components. Then, the preprocessing unit 12 stores the extracted amplitude components in the measurement signal storage unit 30 in chronological order.

[0044] (1-2) Calculation of user identification probability distribution Next, in step S12, the control unit 1 of the personal identification device PD reads the amplitude components from the measurement signal storage unit 30 at a constant length at a time. Then, in step S13, the read amplitude components are input to an estimation model for user identification prepared in advance in response to a thumbs-up gesture, and a user identification probability distribution P user get.

[0045] For example, the user identification probability distribution P user As P user = (0.6, 0.3, 0.1) is obtained. In this case, the probability of being estimated as user A, B, and C is 60%, 30%, and 10%, respectively. user (A), P user (B) , P user (C) can be obtained.

[0046] In step S14, the control unit 1 of the personal identification device PD calculates the user identification probability distribution P user (A), P user (B) , P user (C) is stored in the user identification probability distribution DB 32 of the data storage unit 3 in association with the user ID.

[0047] (2) Estimation of behavior after starting use FIG. 5 is a flowchart showing the procedure and contents of the motion estimation process executed by the control unit 1 of the personal identification device PD.

[0048] (2-1) Acquisition of surface electromyogram signals After the initial registration of the user identification probability distribution by the above-mentioned specified gesture motion, it is assumed that the user moves his / her fingers to operate the prosthetic hand. Then, in step S21, the control unit 1 of the personal identification device PD, under the control of the surface electromyogram acquisition processing unit 11, acquires the surface electromyogram signals output from each electrode of the sensor unit SS via the input / output I / F unit 4, and performs preprocessing by the preprocessing unit 12 to extract the amplitude components. Note that the contents of the preprocessing are the same as those at the time of the initial registration of the user identification probability distribution, so a description thereof will be omitted here.

[0049] (2-2) Calculating the probability of action Next, in step S22, the control unit 1 of the personal identification device PD reads the amplitude components of the surface EMG signal from the measurement signal storage unit 30 at a fixed length under the control of the movement probability calculation processing unit 13. Then, in step S23, the movement probability calculation processing unit 13 reads the movement probability estimation model of each user from the movement probability estimation model DB 31, and in step S24 inputs the amplitude components of the surface EMG signal to the movement probability estimation model as explanatory variables. The movement probability estimation model outputs the movement probability corresponding to the input amplitude components as a response variable.

[0050] For example, let the amplitude component of the surface electromyogram signal just obtained be S. Let this amplitude component S be the motion probability estimation model M of user A. A When input to this, the motion probability estimation model M A from M A (S) = (P motion_1 ,…,P motion_k ) will be output.

[0051] Here, k in the above formula indicates the number of types of motions to be estimated. Specifically, the motions to be estimated include four types: "palmar flexion," "dorsal flexion of fingers," "elbow flexion," and "elbow extension," so in this case k = 4.

[0052] Similarly, the motion probability calculation processing unit 13 converts the amplitude component S into motion probability estimation models M B , M C, and the estimated action probability for users B and C is M B (S), M C (S) is obtained. An example of the result is shown below. M A (S) = (0.2, 0.5, 0.2, 0.1) M B (S) = (0.4, 0.3, 0.2, 0.1) M C (S) = (0.2, 0.2, 0.2, 0.4) (2-3) Motion Estimation Next, under the control of the motion estimation processing unit 14, the control unit 1 of the personal identification device PD first obtains the user identification probability distribution P user (A), P user (B) , P user Then, in step S26, the motion estimation processing unit 14 reads the motion probability estimation result M A (S), M B (S), M C (S) is the above user identification probability distribution P user (A), P user (B) , P user (C) The motion is estimated by weighting and adding the data.

[0053] For example, the estimation result M A (S), M B (S), M C (S) is the user identification probability distribution P user If we add the weighted sum of =(0.6, 0.3, 0.1), we get m^=arg max motion Σ u={A,B,C} P user (u) M u (S) =arg max motion (0.6 × [0.2, 0.5, 0.2, 0.1] + 0.3 × [0.4, 0.3, 0.2, 0.1] + 0.1 × [0.2, 0.2, 0.2, 0.4]) =arg max motion [0.26, 0.41, 0.2, 0.13] It becomes.

[0054] Among the values ​​[0.26, 0.41, 0.2, 0.13] obtained by the weighted addition, the index with the maximum value is the second element (0.41). Therefore, the motion estimation processor 14 estimates that the motion “dorsiflexion of fingers” corresponding to the second element is the motion m^ performed by the user.

[0055] (3) Update the user identification probability distribution The control unit 1 of the personal identification device PD then executes a process of updating the user identification probability distribution based on the estimated motion m̂ as follows. FIG. 6 is a flowchart showing an example of the procedure and processing contents of a user identification probability distribution update process executed by the control unit 1 of the personal identification device PD.

[0056] (3-1) Estimation of the first user identification probability distribution The control unit 1 of the personal identification device PD, under the control of the first user identification probability distribution calculation processing unit 15, first reads in a first estimation model for estimating a user identification probability distribution based on a current estimated motion from the first estimation model DB 33 in step S31. Then, in step S32, the first user identification probability distribution calculation processing unit 15 inputs the motion m^ estimated by the motion estimation processing unit 14 as an explanatory variable to the read first estimation model, and calculates a user identification probability distribution P corresponding to the estimated motion m^ from the first estimation model. user m^ get.

[0057] As a result, for example, P user m^ =(0.1, 0.8, 0.1) is obtained.

[0058] (3-2) Estimation of the second user identification probability distribution The control unit 1 of the personal identification device PD then reads in a second estimation model for estimating a user identification probability distribution based on a motion transition from the second estimation model DB 34 in step S33 under the control of the second user identification probability distribution calculation processing unit 16. Then, in step S34, the second user identification probability distribution calculation processing unit 16 inputs, as an explanatory variable, the motion transition, which is a time-series change of the motion m^ estimated by the motion estimation processing unit 14 at a certain time in the past, to the read second estimation model, and calculates a user identification probability distribution P user motion transition get.

[0059] As a result, for example, P user motion transition = (0.2, 0.5, 0.3) is obtained.

[0060] With the above estimation process, for example, if user A has a strong tendency to transition from elbow flexion to palmar flexion of the fingers, while other users have a strong tendency to transition from palmar flexion of the fingers to elbow flexion, then by having the second estimation model learn this characteristic in advance, the second estimation model will obtain an estimation result that indicates a high probability that user A will transition from elbow flexion to palmar flexion of the fingers.

[0061] (3-3) Update of user identification probability distribution Next, in step S35, under the control of the update processing unit 17, the control unit 1 of the personal identification device PD calculates the user identification probability distribution P corresponding to the estimated motion m̂ obtained by the first user identification probability distribution calculation processing unit 15. user m^ and the user identification probability distribution P corresponding to the motion transition obtained by the second user identification probability distribution calculation processing unit 16. user motion transition Based on this, the latest user identification probability distribution P^ user Calculate.

[0062] for example, P^ user =αP user +(1-α) / 2(P user m^ + P user motion transition ) The latest user identification probability distribution P^ user For example, if α=0.5, P^ user =(0.375, 0.475, 0.15) Thus, the probability of user B being the highest.

[0063] The update processing unit 17 updates the user identification probability distribution P user The calculated latest user identification probability distribution P^ user In addition, the above user identification probability distribution P^ user The calculation process is repeated so that α converges to 1. For example, 0.01 is added each time the calculation step is repeated until α becomes 1.

[0064] (Action and effect) As described above, in one embodiment, the user identification probability distribution registered based on the user's specified gesture movements when the device is first used is updated based on the user's movements during the subsequent period of use, and on the estimated movements and transitions of past movements.

[0065] Therefore, even if the accuracy of the user identification probability distribution registered based on the user's specified gesture motion at the start of use is low, the user identification probability distribution is updated based on the user's motion estimated during the subsequent period of use and past motion transitions. As a result, the user identification result is correctly corrected during the period of use, and it becomes possible to correctly switch the model for estimating the user's motion to a model corresponding to the user.

[0066] [Other embodiments] As each model described in the embodiment, for example, in addition to a Convolutional Neural Network (CNN), a machine learning model to which other architectures such as a Recurrent Neural Network (RNN) can be applied can be used.

[0067] The hardware and software configuration of the personal identification device, the processing procedures and processing contents of the processing for registering a user identification probability distribution based on specified actions, the processing for estimating actions during the period of use and the processing for updating the user identification probability distribution, the areas to be measured for surface electromyograms, etc. can be modified in various ways without departing from the spirit of this invention.

[0068] Although the embodiment of the present invention has been described in detail above, the above description is merely an example of the present invention in every respect. It goes without saying that various improvements and modifications can be made without departing from the scope of the present invention. In other words, in implementing the present invention, specific configurations according to the embodiment may be appropriately adopted.

[0069] In short, this invention is not limited to the above-mentioned embodiment as it is, and in the implementation stage, the components can be modified and embodied without departing from the gist of the invention. In addition, various inventions can be formed by appropriately combining multiple components disclosed in the above-mentioned embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, components from different embodiments may be appropriately combined. [Explanation of symbols]

[0070] PD...Personal Identification Device SS: Sensor unit 1...Control unit 2...Program memory section 3. Data storage section 4... Input / Output I / F section 5. Bus 11...Surface electromyogram acquisition processing unit 12...Pretreatment section 13...Motion probability calculation processing unit 14...Motion estimation processing unit 15...first user identification probability distribution calculation processing unit 16...Second user identification probability distribution calculation processing unit 17...Update processing section 30...Measurement signal storage unit 31...Movement Probability Estimation Model DB 32...User identification probability distribution DB 33…First estimation model DB 34…Second estimation model DB

Claims

1. a first processing unit that generates and stores first user identification probability distribution information representing an identification probability distribution of a plurality of users to be identified; a second processing unit that acquires a surface electromyogram signal corresponding to a motion of an arbitrary user among the plurality of users and extracts a feature amount from the acquired surface electromyogram signal; a third processing unit that estimates the motion of the arbitrary user from the extracted feature amount; a fourth processing unit for determining second user identification probability distribution information corresponding to the estimated motion; a fifth processing unit that calculates third user identification probability distribution information corresponding to a motion transition that represents a time-series change of a plurality of past motions including the estimated motion; a sixth processing unit that updates the first user identification probability distribution information based on the second user identification probability distribution information and the third user identification probability distribution information; A personal identification device comprising:

2. 2. The personal identification device according to claim 1, wherein the first processing unit acquires the surface electromyogram signals corresponding to specified movements by the multiple users to be identified, and generates and stores the first user identification probability distribution information based on the acquired surface electromyogram signals.

3. 2. The personal identification device according to claim 1, wherein the second processing unit removes artifacts due to body movement from the acquired surface electromyogram signal, rectifies the signal, extracts an amplitude component, and sets the extracted amplitude component as the feature.

4. The third processing unit includes: an action probability estimation processing unit that estimates, for each of the plurality of users, an action probability corresponding to the feature amount by using an action probability estimation model that is created in advance corresponding to each of the plurality of users; a motion estimation processing unit that estimates the motion probability by reflecting the first user identification probability distribution information in the estimated motion probability; The personal identification device of claim 1 .

5. the fourth processing unit estimates the second user identification probability distribution information corresponding to the estimated motion by using a first user identification probability distribution estimation model created in advance corresponding to a plurality of types of basic motions; 2. The personal identification device according to claim 1, wherein the fifth processing unit estimates the third user identification probability distribution information corresponding to the action transitions using a second user identification probability distribution estimation model created in advance corresponding to a plurality of types of basic action transitions.

6. 2. The personal identification device according to claim 1, wherein the sixth processing unit updates the first user identification probability distribution information by weighting and adding the second user identification probability distribution information and the third user identification probability distribution information to the first user identification probability distribution information.

7. A personal identification method executed by an information processing device, comprising: A first processing step of acquiring and storing first user identification probability distribution information representing an identification probability distribution of a plurality of users to be identified; a second processing step of acquiring a surface electromyogram signal corresponding to a motion of any one of the plurality of users, and extracting a feature amount from the acquired surface electromyogram signal; a third processing step of estimating the motion of the arbitrary user from the extracted feature amount; a fourth process of determining second user identification probability distribution information corresponding to the estimated motion; a fifth process of calculating third user identification probability distribution information corresponding to a motion transition representing a time-series change of a plurality of past motions including the estimated motion; a sixth processing step of updating the first user identification probability distribution information based on the second user identification probability distribution information and the third user identification probability distribution information; 13. A method for identifying an individual comprising:

8. 7. A program for causing a processor included in a personal identification device to execute processes executed by the first to sixth processing units included in the personal identification device according to claim 1.

Images