WO2023089822A1

WO2023089822A1 - Wearer identification device, wearer identification system, wearer identification method, and wearer identification program

Info

Publication number: WO2023089822A1
Application number: PCT/JP2021/042778
Authority: WO
Inventors: 勇貴久保; 幸生小池
Original assignee: 日本電信電話株式会社
Priority date: 2021-11-22
Filing date: 2021-11-22
Publication date: 2023-05-25
Also published as: JPWO2023089822A1

Abstract

A wearer identification device according to an embodiment comprises a feature amount generation unit, a determining unit, and an identification unit. The feature amount generation unit receives, from a sensor worn on a part of the body of a user to be identified, a measurement signal corresponding to the vibration characteristics of the body part of the user measured by the sensor, and generates a feature amount representing the vibration characteristics from the measurement signal. The determining unit determines, on the basis of the magnitude of fluctuations in the feature amount generated by the feature amount generation unit, whether the state of the part on which the sensor is worn by the user is stable. The identification unit performs personal identification of the user on the basis of the feature amount generated by the feature amount generation unit, if the determining unit determines that the state of the part on which the sensor is worn is stable.

Description

Wearing user identification device, wearing user identification system, wearing user identification method, and wearing user identification program

Embodiments of the present invention relate to a wearer's identification device, a wearer's identification system, a wearer's identification method, and a wearer's identification program.

Using a pair of piezoelectric elements such as piezo elements, one as a speaker and the other as a microphone, measure the vibration characteristics of the target on which the speaker and microphone are installed, and recognize the state of the target based on the measured vibration characteristics There is a method called active acoustic sensing, which estimates the gripped state of an object or object (see, for example, Non-Patent Document 1). In this active acoustic sensing, sound waves, which are in the inaudible range, are transmitted from a speaker to a target, and the vibration propagated through the target is received by a microphone, and the frequency characteristics of the received signal are analyzed. It utilizes the fact that the vibration characteristics of the object on which the microphone is installed change due to changes in the internal structure and boundary conditions.

By using this active acoustic sensing method, it is also possible to identify each target.

When trying to identify a user rather than identifying static objects using active acoustic sensing, the part of the body where the speaker and microphone are attached may move, unlike static objects. obtain. When the wearing part is moved, the internal structure of the body part changes, which causes changes in the values of the vibration characteristics. This noise may cause a failure in personal identification of the user.

The present invention seeks to provide a technology that can reduce erroneous determinations when active acoustic sensing is used for personal identification of users.

In order to solve the above problems, a wearing user identification device according to one aspect of the present invention includes a feature generation unit, a determination unit, and an identification unit. The feature amount generation unit receives a measurement signal corresponding to the vibration characteristics of the user's body part measured by the sensor from the sensor attached to the body part of the user to be identified, and generates a feature representing the vibration characteristic from the measurement signal. produce quantity. The determination unit determines whether the state of the site where the sensor is attached to the user is stable, based on the magnitude of variation in the feature amount generated by the feature amount generation unit. The identification unit performs personal identification of the user based on the feature amount generated by the feature amount generation unit when the determination unit determines that the state of the wearing site is stable.

According to one aspect of the present invention, the personal identification of the user is performed only in a stable state in which the vibration characteristic value does not greatly change, thereby preventing an erroneous determination when active acoustic sensing is used for personal identification of the user. can be provided.

FIG. 1 is a block diagram showing an example configuration of a wearable user identification system including a wearable user identification device according to an embodiment of the present invention. FIG. 2 is a plan view showing the configuration of the measurement unit worn by the user. FIG. 3 is a schematic diagram showing a state in which a user wears the measurement unit. FIG. 4 is a diagram showing an example of a spectrogram. FIG. 5 is a block diagram showing an example of the hardware configuration of the wearing user identification device. FIG. 6 is a flow chart showing an example of a learning processing operation related to learning of a classification model in the wearing user identification device. FIG. 7 is a flow chart showing an example of an identification processing operation related to personal identification of a user in the wearable user identification device.

An embodiment according to the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an example of the configuration of a wearable user identification system 1 according to one embodiment of the present invention. The wearing user identification system 1 includes a wearing user identification device 10 according to one embodiment of the present invention, a measurement section 20 and an audio interface section 30 . The wearing user identification device 10 includes a signal generation unit 11 , a signal storage unit 12 , a feature amount generation unit 13 , a model learning unit 14 , a model storage unit 15 , an identification execution determination unit 16 and a user identification unit 17 . The measurement unit 20 is a sensor for measuring vibration characteristics of a measurement target and a part for attaching the sensor to the body of a target user, and includes a signal generation unit 21 and a signal reception unit 22 . The audio interface section 30 is an interface between the wearing user identification device 10 and the measurement section 20 and has a signal control section 31 and a signal amplification section 32 . Wire connection or wireless connection can be established between the signal generation unit 11 and the signal control unit 31 and between the signal control unit 31 and the signal generation unit 21 as long as they have a function of transmitting and receiving signals. However, the connection form does not matter.

The signal generation unit 11 of the wearing user identification device 10 generates an acoustic signal based on arbitrarily set parameters. As an example, the acoustic signal is ultrasound sweeping from 20 kHz to 40 kHz. However, the settings of the acoustic signal, such as whether or not to sweep, use of other frequency bands, etc., do not matter.

The signal control unit 31 of the audio interface unit 30 generates a drive signal based on the acoustic signal generated by the signal generation unit 11 based on the preset parameters, and vibrates the target through the signal generation unit 21 of the measurement unit 20. give. Note that in the case of a configuration in which the feature amount included in the teacher data regarding the individual to be identified is generated in advance and registered in a registration database (not shown), the vibration at this time will be the same as the vibration used at that time. As long as the included frequency is included, other frequencies may be mixed.

The signal generator 21 and the signal receiver 22 of the measurement unit 20 are composed of two piezoelectric elements that do not contact each other. A piezoelectric element can be realized by, for example, a piezo element. One piezoelectric element serves as the signal generating section 21 that generates vibration having the same frequency characteristics as the drive signal generated by the signal control section 31 of the audio interface section 30 . The other piezoelectric element serves as a signal receiving section 22 that receives vibration. The signal receiving unit 22 acquires vibrations propagating inside and on the surface of the object in which the signal receiving unit 22 is installed. Here, when the vibration given from the signal generator 21 is propagated to the signal receiver 22, the user's living body, which is the object of measurement, functions as a propagation path. The frequency characteristics of the acquired vibration change. The signal receiver 22 transmits the received vibration signal (hereinafter referred to as a reaction signal) to the signal amplifier 32 . The signal generating unit 21 and the signal receiving unit 22 may be of any form and material as long as they are mechanisms capable of propagating vibrations while being in contact with the target living body.

FIG. 2 is a plan view showing the configuration of the measurement unit 20, and FIG. 3 is a schematic diagram showing a state in which the measurement unit 20 is worn by a user who is a measurement target. Here, the measurement unit 20 is configured as a band-type sensor, but if the signal generation unit 21 and the signal reception unit 22 can be fixedly attached to the user's skin while maintaining a certain distance, the measurement unit 20 can be used for a living body. Other implementation methods such as adhesive tape may be used.

In the measuring section 20, the two piezoelectric elements, which are the signal generating section 21 and the signal receiving section 22, are attached to the fixed section 23 so as to keep a certain distance so that they do not come into contact with each other. The fixing portion 23 also functions as a reinforcing member that reinforces the strength of the signal generating portion 21 and the signal receiving portion 22 so that the signal generating portion 21 and the signal receiving portion 22 can be continuously used. A band 24 and a square can 25 are attached to positions facing the fixing portion 23 with the signal generating portion 21 and the signal receiving portion 22 interposed therebetween. A hook-and-loop fastener 26 is provided on the back surface of the band 24 . The measurement unit 20 having such a configuration is worn by the user by adjusting the length of the band 24 , wrapping it around the user's wrist, and fixing it with the hook-and-loop fastener 26 . At this time, when performing personal authentication, if it is unified for the individual, it does not matter where it is worn.

The signal amplification section 32 of the audio interface section 30 amplifies the reaction signal acquired by the signal reception section 22 of the measurement section 20 and transmits it to the wearing user identification device 10 . The reason why the signal is amplified by the signal amplifier 32 is that the vibration passing through the object to be measured is attenuated, so it is necessary to amplify the signal to a level that enables processing.

In the wearing user identification device 10 , the reaction signal transmitted from the signal amplification section 32 of the audio interface section 30 is stored in the signal storage section 12 .

The feature amount generation unit 13 extracts the reaction signal stored in the signal storage unit 12 for each fixed time interval, and performs, for example, FFT (Fast Fourier Transform) on the extracted reaction signal. , to generate a spectrogram, which is a feature quantity representing the acoustic frequency characteristics of the living body to be measured. FIG. 4 is a diagram showing an example of this spectrogram. When executing a learning processing operation related to learning of a classification model, the feature amount generation unit 13 generates teacher data including a set of the generated spectrogram and the user ID corresponding to the user who is the measurement target, and the model learning unit 14. Note that the teacher data may be generated by extracting from the registration database created in advance. Further, the feature amount generation unit 13 outputs the generated spectrogram to the identification execution determination unit 16 when executing the identification processing operation related to personal identification of the user.

The model learning unit 14 generates and learns a classification model whose input is the teacher data obtained from the feature amount generation unit 13 and whose output is the user ID. The model learning unit 14 registers the model itself or the parameters of the model obtained by this learning process in the model storage unit 15, which is a model database. As long as it is possible to perform learning so as to obtain an optimal output by performing parameter tuning or the like on teacher data, the classification model and the type of library used for its learning do not matter. For example, using a generally known machine learning library, algorithms for generating classification models such as SVM (Support Vector Machine) and neural networks obtain optimal output by performing parameter tuning, etc. on teacher data. It is also good to learn so that it can be done.

The identification execution determination unit 16 determines whether or not the user identification processing by the user identification unit 17 is to be executed. The identification execution determination unit 16 obtains the stability of the spectrogram within a set fixed time from the spectrogram acquired from the feature amount generation unit 13 . For example, in order to obtain the stability in the spectrogram for 2 seconds, obtain the average value of dB for each frequency in 2 seconds (eg: average value of 20 kHz = 0 [dB], ..., average value of 40 kHz = -5 [dB ]). Then, the identification execution determination unit 16 uses these average values to calculate the standard deviation at each frequency of the spectrogram as the degree of stability. The identification execution determination unit 16 determines whether the state of the user-worn site of the measurement unit 20 is stable based on whether the standard deviation at each frequency is higher than an arbitrary set threshold. That is, when the standard deviation at each frequency of the spectrogram is not higher than the threshold, the identification execution determination unit 16 considers that the state of the user wearing part of the measurement unit 20 is stable, and the processing in the user identification unit 17 is performed. let it happen Specifically, the identification execution determination unit 16 outputs the spectrogram acquired from the feature amount generation unit 13 to the user identification unit 17 . Conversely, when the standard deviation at each frequency of the spectrogram is higher than the threshold, the identification execution determination unit 16 determines that the user wearing the measurement unit 20 is moving and the stability is low. Do not process. That is, the identification execution determination section 16 does not output the spectrogram to the user identification section 17 . It should be noted that this processing may be performed by limiting only to values having a certain value or more in each frequency characteristic (for example, -50 [dB] or more).

The user identification unit 17 inputs the spectrogram, which is the feature amount acquired from the identification execution determination unit 16, as an input to the classification model registered in the model storage unit 15, and outputs the classification model for determining an individual user. Find the numerical value of

Here, when SVM is used as an algorithm for the model learning unit 14, the classification model outputs, as a reference value, a score indicating the degree of similarity to each label (user ID) of the classification model with respect to the input spectrogram. be done. The score can be output as '1-(similarity)', for example, if the similarity is normalized and expressed between '0' and '1'.

Also, when Random Forest is used as an algorithm for the model learning unit 14, data is randomly extracted from the teacher data and a plurality of decision trees are generated. output. Since the higher the number of determination results, the better, the classification model outputs "(the number of determinations) - (the number of determination results)" as a reference value.

Also, other classification algorithms such as DNN (Deep Neural Network) may be used as the classification algorithm. In this case, the reference value may be obtained by subtracting the normalized similarity from "1", or may be obtained by converting the similarity using the reciprocal of the similarity.

The user identification unit 17 uses the obtained reference value list to determine the user ID with the smallest (similar) reference value. In the determination process, a threshold for determination may be set for the degree of similarity, and determination may be made only when the reference value is smaller than the threshold.

FIG. 5 is a diagram showing an example of the hardware configuration of the wearing user identification device 10. As shown in FIG. As shown in FIG. 5, the wearing user identification device 10 is configured by a computer such as a microcomputer or a personal computer, and has a hardware processor 101 such as a CPU (Central Processing Unit). By using a multi-core and multi-threaded CPU, it is possible to execute a plurality of information processes at the same time. Also, the processor 101 may include multiple CPUs. In the wearable user identification device 10 , a program memory 102 , a data memory 103 , a communication interface 104 and an input/output interface 105 are connected to the processor 101 via a bus 106 . In FIG. 5, "interface" is abbreviated as "IF".

The communication interface 104 can include, for example, one or more wired or wireless communication modules. The example shown in FIG. 5 shows two

communication modules

1041 and 1042 . The communication module 1041 is a communication module using short-range wireless technology such as Bluetooth (registered trademark), and transmits and receives signals to and from the signal control section 31 and the signal amplification section 32 of the audio interface section 30 . Also, the communication module 1041 can transmit and receive signals to and from the signal control section 31 and the signal amplification section 32 of the remote audio interface section 30 via the network NW. A network consists of an IP network including the Internet and an access network for accessing this IP network. As the access network, for example, a public wired network, a mobile phone network, a wired LAN (Local Area Network), a wireless LAN, CATV (Cable Television), etc. are used. Therefore, the wearing user identification device 10 can also acquire reaction signals from the plurality of measuring units 20 via the plurality of audio interface units 30 and identify the plurality of wearing users who wear the respective measuring units 20 .

Also, an input unit 107 and a display unit 108 are connected to the input/output interface 105 . The input unit 107 and the display unit 108 are so-called tablet-type inputs, in which an input detection sheet adopting an electrostatic method or a pressure method is arranged on a display screen of a display device using liquid crystal or organic EL (Electro Luminescence), for example. - using a display device can be used; Note that the input unit 107 and the display unit 108 may be configured by independent devices. The input/output interface 105 inputs operation information input from the input unit 107 to the processor 101 and displays display information generated by the processor 101 on the display unit 108 .

Note that the input unit 107 and the display unit 108 do not have to be connected to the input/output interface 105 . The input unit 107 and the display unit 108 are provided with a communication unit for connecting directly to the communication interface 104 or via the network NW, so that information can be exchanged with the processor 101 .

The input/output interface 105 may have a read/write function for a recording medium such as a semiconductor memory such as a flash memory, or may be connected to a reader/writer having a read/write function for such a recording medium. It may have functions. As a result, a recording medium detachable from the wearing user identification device 10 can be used as a model database holding classification models. The input/output interface 105 may further have a connection function with other devices.

In addition, the program memory 102 includes, as a non-temporary tangible computer-readable storage medium, for example, a non-volatile memory such as a HDD (Hard Disk Drive) or an SSD (Solid State Drive) that can be written and read at any time, and a ROM ( It was used in combination with non-volatile memory such as Read Only Memory). The program memory 102 stores programs necessary for the processor 101 to execute various control processes according to one embodiment. That is, each processing function unit of the signal generation unit 11, the feature amount generation unit 13, the model learning unit 14, the identification execution determination unit 16, and the user identification unit 17 executes the program stored in the program memory 102 as described above. It can be implemented by being read and executed by the processor 101 . Some or all of these processing functions may be implemented in various other forms, including integrated circuits such as Application Specific Integrated Circuits (ASICs) or field-programmable gate arrays (FPGAs). May be.

In addition, the data memory 103 is a tangible computer-readable storage medium, for example, a combination of the above nonvolatile memory and a volatile memory such as RAM (Random Access Memory). This data memory 103 is used to store various data acquired and created in the process of performing various processes. That is, in the data memory 103, an area for storing various data is appropriately secured in the process of performing various processes. As such areas, the data memory 103 can be provided with, for example, a signal storage section 1031, a model storage section 1032, an identification result storage section 1033, and a temporary storage section 1034. FIG.

The signal storage section 1031 stores the reaction signal transmitted from the signal amplification section 32 of the audio interface section 30 . That is, the signal storage section 12 can be configured in this signal storage section 1031 .

The model storage unit 1032 stores the classification model learned by the model learning unit 14. That is, the model storage unit 15 can be configured in this model storage unit 1032 .

The identification result storage unit 1033 stores output information obtained when the processor 101 operates as the user identification unit 17 .

Temporary storage unit 1034 stores spectrograms and teacher data acquired or generated when processor 101 performs operations as feature amount generation unit 13, model learning unit 14, identification execution determination unit 16, and user identification unit 17. , classification models, reference values, and other data.

Next, the operation of the wearing user identification device 10 will be described.
In this embodiment, prior to user identification, the wearable user identification device 10 first generates a classification model associated with the user ID using a sensor capable of measuring the state of each user to be identified, and generates the classification model. The classified model is stored in the model storage unit 15 as registration data.

For this purpose, first, the signal generator 21 and the signal receiver 22 of the measurement unit 20 are worn on the wrist of the user who is the person to be identified using the band 24 . The vibration generated by the signal generator 21 may be of any form and type as long as it has frequency characteristics like an acoustic signal. In this embodiment, an acoustic signal will be described as an example.

FIG. 6 is a flowchart showing an example of a learning processing operation related to learning of a classification model in the wearing user identification device 10. FIG. This flowchart shows the processing operation of a part of the wearing user identification device 10 , specifically, the processor 101 of the computer functioning as the signal generator 11 , the feature amount generator 13 , and the model learning unit 14 . After the measurement unit 20 is worn on the wrist of the user to be registered, when the input unit 107 instructs the start of learning via the input/output interface 105, the processor 101 starts the operation shown in this flowchart. In addition, for a remote user to be registered, when the communication interface 104 receives a predetermined installation completion notification transmitted from an information processing device such as a smartphone operated by the user via the network NW, the processor 101 The operation shown in this flow chart is started.

First, the processor 101 functions as the signal generator 11 and generates an acoustic signal (driving signal) based on arbitrarily set parameters (step S101). The driving signal is, for example, an ultrasonic wave sweeping from 20 kHz to 40 kHz. However, the settings of the acoustic signal, such as whether or not to sweep, whether or not to use other frequency bands, etc., do not matter. The generated drive signal is transmitted to the audio interface section 30 via the communication interface 104 . Alternatively, a signal generation module that generates a drive signal under the control of the processor 101 may be prepared separately, and the drive signal generated there may be transmitted to the audio interface section 30 via the communication interface 104 .

The audio interface section 30 transmits this drive signal to the signal generation section 21 of the measurement section 20 . This driving signal causes the body of the user to be registered to vibrate through the signal generator 21 . The signal receiving unit 22 of the measuring unit 20 acquires the vibration that is given to the living body of the user to be registered by the signal generating unit 21 and propagates inside and on the surface of the living body. Here, when the vibration given from the piezoelectric element of the signal generating section 21 is propagated to the piezoelectric element of the signal receiving section 22, the living body of the user to be registered functions as a propagation path, and the vibration applied according to this propagation path. change the frequency characteristics of the vibration. This frequency characteristic differs from person to person. The signal receiving section 22 detects the propagated vibration and transmits a reaction signal indicated by the detected vibration to the audio interface section 30 . The signal amplifying section 32 of the audio interface section 30 amplifies the reaction signal transmitted from the signal receiving section 22 of the measuring section 20 and transmits it to the wearing user identification device 10 .

The reaction signal transmitted from the audio interface unit 30 is received by the communication interface 104. The processor 101 stores the received reaction signal in the signal storage section 1031 of the data memory 103 (step S102).

Next, the processor 101 functions as the feature generator 13 and performs the following processing operations.
First, the processor 101 extracts the reaction signal stored in the signal storage unit 1031 for each fixed time interval. The number of samples of the signal does not matter. The extracted reaction signal is stored in temporary storage section 1034 of data memory 103 . Then, the processor 101 generates a spectrogram, which is a feature quantity representing the acoustic frequency characteristics of the living body, from the extracted reaction signal stored in the temporary storage unit 1034 (step S103). The generated spectrogram is stored in temporary storage section 1034 of data memory 103 .

Then, the processor 101 assigns a user ID, which is a unique identifier, to the generated spectrogram, and generates teacher data that combines these spectrograms and the user ID (step S104). The generated teacher data is stored in temporary storage section 1034 of data memory 103 . Furthermore, the processor 101 may extract registration data created in advance from the model storage unit 15 configured in the model storage unit 1032 of the data memory 103, and use it to generate teacher data.

Next, the processor 101 functions as the model learning unit 14 and performs the following processing operations.
First, the processor 101 receives the spectrogram in the training data as an input, and generates and learns a classification model that outputs a user ID in the training data as a label and a reference value as a difference from the input (step S105).

Then, the processor 101 registers the model itself or the parameters of the classification model and the classification model obtained by this learning process in the model storage unit 15 configured in the model storage unit 1032 of the data memory 103 (step S106). ).

When the learning of one user to be registered is completed in this way, the processor 101 stops generating the drive signal and ends the transmission of the drive signal to the audio interface unit 30 by the communication interface 104 (step S107). . Then, the learning processing operation shown in this flow chart ends.

After that, other users can learn in the same way.

Next, the operation of the wearable user identification device 10 when identifying an individual user to be identified will be described. The wearing user identification device 10 inputs a spectrogram, which is a feature amount obtained from a user to be identified who wears the measurement unit 20, into a classification model registered in the model storage unit 15 or the like, and identifies an individual user. These specific processes will be described below.

FIG. 7 is a flowchart showing an example of an identification processing operation related to personal identification of the user in the wearable user identification device 10. FIG. This flowchart shows the processing operation in the processor 101 of the computer that functions as a part of the wearing user identification device 10, specifically, the signal generation unit 11, the feature amount generation unit 13, the identification execution determination unit 16, and the user identification unit 17. is shown. After the measurement unit 20 is worn on the wrist of the user to be identified, when the input unit 107 instructs the start of personal identification through the input/output interface 105, the processor 101 starts the operation shown in this flowchart. For a remote user to be identified, when a predetermined identification start notification transmitted via the network NW from an information processing device such as a smart phone operated by the user is received by the communication interface 104, the processor 101 The operation shown in this flow chart is started.

First, the processor 101 functions as the signal generator 11 and generates a drive signal based on arbitrarily set parameters (step S201). The generated drive signal is transmitted to the audio interface section 30 via the communication interface 104 . Alternatively, a signal generation module that generates a drive signal under the control of the processor 101 may be prepared separately, and the drive signal generated there may be transmitted to the audio interface section 30 via the communication interface 104 .

The audio interface section 30 transmits this drive signal to the signal generation section 21 of the measurement section 20 . This driving signal causes the body of the user to be registered to vibrate through the signal generator 21 . The vibration at this time includes the frequency included in the vibration used to generate the feature quantity included in the registered data regarding the learned user registered in the model storage unit 15 (hereinafter referred to as a registered user). It does not matter if other frequencies are mixed as long as it is The signal receiving unit 22 of the measuring unit 20 acquires the vibration that is given to the living body of the user to be identified by the signal generating unit 21 and propagates inside and on the surface of the living body. Here, when the vibration given from the piezoelectric element of the signal generating section 21 is propagated to the piezoelectric element of the signal receiving section 22, the body of the user to be identified functions as a propagation path, and the vibration is applied according to this propagation path. change the frequency characteristics of the vibration. The signal receiving section 22 detects the propagated vibration and transmits a reaction signal indicated by the detected vibration to the audio interface section 30 . The signal amplifying section 32 of the audio interface section 30 amplifies the reaction signal transmitted from the signal receiving section 22 of the measuring section 20 and transmits it to the wearing user identification device 10 .

The reaction signal transmitted from the audio interface unit 30 is received by the communication interface 104. The processor 101 stores the received reaction signal in the signal storage section 1031 of the data memory 103 (step S202).

Next, the processor 101 functions as the feature amount generation unit 13 and extracts the reaction signal stored in the signal storage unit 1031 for each fixed time interval. The number of samples of the signal does not matter. The extracted reaction signal is stored in temporary storage section 1034 of data memory 103 . Then, the processor 101 performs, for example, FFT on the extracted reaction signal stored in the temporary storage unit 1034 to generate a spectrogram, which is a feature quantity representing the acoustic frequency characteristics of the living body (step S203). The generated spectrogram is stored in temporary storage section 1034 of data memory 103 .

Next, the processor 101 functions as the identification execution determination unit 16 and performs the following processing operations.
First, the processor 101 calculates the stability of the spectrogram within a set fixed time from the spectrogram stored in the temporary storage unit 1034 as follows. First, the processor 101 determines whether spectrograms for a certain period of time, for example, 2 seconds, have been generated (step S204). When determining that the spectrogram for 2 seconds has not yet been generated (NO in step S204), the processor 101 proceeds to the process of step S201 and repeats the above-described processing operations.

On the other hand, if it is determined that the spectrogram for 2 seconds has been generated (YES in step S204), the processor 101 calculates the degree of stability (step S205). For example, the processor 101 obtains the average value of dB for each frequency for 2 seconds from the spectrogram for 2 seconds stored in the temporary storage unit 1034, and uses this average value to obtain the standard deviation at each frequency of the spectrogram, Calculated as stability.

Then, the processor 101 determines whether the state of the user-mounted site of the measurement unit 20 is stable based on whether the standard deviation at each frequency of these spectrograms is equal to or less than the set arbitrary threshold value (step S206). ).

When the standard deviation at each frequency of the spectrogram is higher than the threshold, the processor 101 determines that the user wearing the measurement unit 20 is moving and not stable (NO in step S206). In this case, the processor 101 deletes the oldest spectrogram from among the spectrograms stored in the temporary storage unit 1034 within a certain period of time (step S207). After that, the processor 101 shifts to the process of step S201 and repeats the above-described processing operations.

On the other hand, when the standard deviation at each frequency of the spectrogram is equal to or less than the threshold, the processor 101 determines that the state of the user-mounted site of the measurement unit 20 is stable (YES in step S206). In this case, the processor 101 functions as the user identification unit 17 and performs the following processing operations.
First, the processor 101 performs personal identification of the user (step S208). That is, the processor 101 stores the classification model registered in the model storage unit 15 configured in the model storage unit 1032 of the data memory 103 with the spectrogram, which is the feature amount generated in step S203 and stored in the temporary storage unit 1034. Enter one, say the latest spectrogram, and get a list of reference values for the classification model. A list of the acquired reference values is stored in the temporary storage unit 1034 of the data memory 103 . Next, processor 101 identifies the smallest reference value from the list of reference values stored in temporary storage unit 1034 . The processor 101 determines, in the model storage unit 15, that registered users associated with the same feature amount as the specified reference value are similar users. Then, processor 101 stores the determined user ID of the registered user in identification result storage section 1033 of data memory 103 as the personal identification result of the user to be identified. Note that in this determination process, a threshold for determination may be set for the degree of similarity, and similar users may be determined only when the specified reference value is smaller than this threshold.

Then, the processor 101 outputs the user ID, which is the personal identification result stored in the identification result storage unit 1033 (step S209). For example, the processor 101 displays the user ID on the display unit 108 via the input/output interface 105 . The processor 101 can also provide the user ID to applications and the like that require personal authentication of the user.

When personal identification for one user to be identified is completed in this manner, the processor 101 stops generating the drive signal and ends transmission of the drive signal to the audio interface unit 30 by the communication interface 104 (step S210). ). Then, the identification processing operation shown in this flow chart ends.

In the wearable user identification device 10 according to the embodiment described above, the feature amount generation unit 13 measures the A response signal, which is a measurement signal corresponding to the vibration characteristics of a user's body part, is received, a feature quantity representing the vibration characteristics is generated from the reaction signal, and an identification execution determination unit 16 as a determination unit determines the feature quantity generation unit 13. Based on the magnitude of variation in the generated feature amount, it is determined whether the state of the site where the measurement unit 20 is worn by the user is stable. The unit 17 performs personal identification of the user based on the feature amount generated by the feature amount generation unit 13 . In this way, by performing personal identification of the user only in a stable state where the vibration characteristic value does not change significantly, the effect of disturbance such as moving the wearing part or receiving an external stimulus to the wearing part. It is possible to reduce erroneous determination in personal identification. That is, the wearable user identification device 10 according to one embodiment can reduce erroneous determinations when active acoustic sensing is used for personal identification of the user.

Note that the reaction signal received from the measurement unit 20 is a vibration signal obtained by detecting vibration propagating inside the user's body, which is the part where the measurement unit 20 is attached. It can be a spectrogram representing the frequency characteristics of the vibration signal generated by, for example, performing FFT (Fast Fourier Transform) on the signal. In this way, a spectrogram can be generated as a feature quantity.

Then, the identification execution determination unit 16 obtains the standard deviation at each frequency of the spectrogram from the spectrogram for a certain period of time using the average value of dB for each frequency in the certain period of time, and the standard deviation at each frequency is the set threshold In the following cases, it is determined that the state of the attachment site is stable. In this manner, the stability can be easily calculated, and based on the stability, a stable state in which the vibration characteristic value does not change significantly can be determined.

In addition, the wearing user identification device 10 according to one embodiment further includes a model storage unit 15 which is a database in which feature amounts for each of a plurality of users to be registered are registered in advance. A user having a feature amount corresponding to the feature amount generated by the feature amount generation unit 13 from the reaction signal received from the measurement unit 20 from among a plurality of registration target users registered in the identification target user identify as In this way, by registering the feature amount for each of a plurality of registration target users in the model storage unit 15, the identification target user can be easily identified based on the feature amount.

Here, the model storage unit 15 inputs a feature amount, and stores a value based on the difference between the feature amount of at least one user to be registered and the input feature amount as an identifier uniquely given to the user to be registered. This model is a feature amount generated by the feature amount generation unit 13 from the reaction signal received from the measurement unit 20 for each of a plurality of registered users. The user identification unit 17 inputs a spectrogram, which is a feature amount generated for a user to be identified, to the model stored in the model storage unit 15, and the user identification unit 17 selects the values output from the model. , is determined as the user ID of the user to be identified, thereby identifying the user to be identified. Therefore, it is possible to appropriately identify the user to be identified using the spectrogram, which is the feature amount of the registered user.

In addition, the wearable user identification system 1 according to one embodiment includes the wearable user identification device 10 according to one embodiment, and the piezoelectric element generates the first vibration to be applied to the wearing part of the user's body, and gives it to the user's body. a measurement unit 20 that acquires, as a measurement signal, a vibration signal corresponding to the second vibration propagated inside the body among the first vibrations received. Therefore, by having each user to be identified wear the measurement unit 20, each user can be individually identified.

[Other embodiments]
In the above embodiment, the user's individual identification is performed based on the feature quantity only in a stable state where the vibration characteristic value does not change significantly. However, the model learning unit 14 may similarly perform learning based on the feature amount only in a stable state, not limited to the case of personal identification of the user. As a result, only stable learning data can be used for data used for learning.

Also, the audio interface unit 30 is arranged between the wearing user identification device 10 and the measuring unit 20, but the audio interface unit 30 may be incorporated in either the wearing user identification device 10 or the measuring unit 20. good.

Also, in the above embodiment, the processing function unit of the wearable user identification device 10 has been described as being composed of one computer, but it may be composed of a plurality of computers by arbitrary division. For example, the model learning unit 14 and the model storage unit 15 may be configured in a computer or a server device that can communicate via the network NW via the communication interface 104 and that is separate from the computer that configures the wearing user identification device 10. .

In addition, the method described in the above embodiment can be executed by a computer (computer) as a program (software means), such as a magnetic disk (floppy (registered trademark) disk, hard disk, etc.), an optical disk (CD-ROM, DVD , MO, etc.), a semiconductor memory (ROM, RAM, flash memory, etc.), or the like, or may be transmitted and distributed via a communication medium. The programs stored on the medium also include a setting program for configuring software means (including not only execution programs but also tables and data structures) to be executed by the computer. A computer that realizes this apparatus reads a program recorded on a recording medium, and optionally constructs software means by a setting program. The operation is controlled by this software means to execute the above-described processes. The term "recording medium" as used herein is not limited to those for distribution, and includes storage media such as magnetic disks, semiconductor memories, etc. provided in computers or devices connected via a network.

In short, the present invention is not limited to the above embodiments, and can be modified in various ways without departing from the gist of the invention at the implementation stage. Moreover, each embodiment may be implemented in combination as much as possible, and in that case, the effect of the combination can be obtained. Furthermore, the above-described embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements.

Reference Signs List 1 wearing user identification system 10 wearing user identification device 11 signal generation unit 12 signal storage unit 13 feature amount generation unit 14 model learning unit 15 model storage unit 16 identification execution determination unit 17 user identification unit 20 Measurement unit 21 Signal generation unit 22 Signal reception unit 23 Fixing unit 24 Band 25 Square can 26 Velcro fastener 30 Audio interface unit 31 Signal control unit 32 Signal amplification unit 101 Processor 102 Program memory 103... Data memory 1031... Signal storage unit 1032... Model storage unit 1033... Identification result storage unit 1034... Temporary storage unit 104...

Communication interface

1041, 1042... Communication module 105... Input/output interface 106... Bus 107... Input unit 108... Display Part NW ... network

Claims

A measurement signal corresponding to the vibration characteristics of the user's body part measured by the sensor is received from a sensor attached to the body part of the user to be identified, and a feature quantity representing the vibration characteristic is obtained from the measurement signal. a feature generation unit that generates
a judgment unit that judges whether the state of the part where the sensor is attached to the user is stable based on the magnitude of variation in the feature amount generated by the feature amount generation unit;
an identification unit that performs personal identification of the user based on the feature amount generated by the feature amount generation unit when the determination unit determines that the state of the wearing site is stable;
A wearable user identification device comprising:
The measurement signal received from the sensor is a vibration signal obtained by detecting vibration propagating inside the body of the user, which is the attachment site of the sensor, and
2. The wearing user identification device according to claim 1, wherein the feature amount generated by the feature amount generation unit is a spectrogram representing frequency characteristics of the vibration signal.
The determination unit obtains the standard deviation at each frequency of the spectrogram from the spectrogram for a certain period of time using the average value of dB for each frequency in the certain period of time, and the standard deviation at each frequency is set 3. The wearing user identification device according to claim 2, wherein, if it is equal to or less than a threshold, it is determined that the state of said wearing part is stable.
Further comprising a database in which the feature amount for each of a plurality of registration target users is registered in advance,
The identification unit has the feature amount corresponding to the feature amount generated by the feature amount generation unit from the measurement signal received from the sensor, among the plurality of registered users registered in the database. identifying the user as being the user to be identified;
A wearable user identification device according to any one of claims 1 to 3.
The database inputs the feature amount, and uniquely assigns to the user to be registered a value based on the difference between the feature amount of at least one user to be registered and the input feature amount. It remembers the model to be output in association with the identifier that
The model is learned based on the feature amount generated by the feature amount generation unit from the measurement signal received from the sensor for each of the plurality of registered users,
The identification unit inputs the feature amount generated for the user to be identified to the model, and determines, among the values output from the model, the relevance to the feature amount of the user to be identified. Identifying the user by determining the identifier output in association with the highest value as the identifier of the user to be identified;
5. A wearable user identification device according to claim 4.
a wearable user identification device according to any one of claims 1 to 5;
A piezoelectric element generates a first vibration applied to the attachment site of the body of the user, and a second vibration propagated inside the body among the first vibration applied to the body of the user. the sensor that acquires a vibration signal corresponding to the measurement signal as the measurement signal;
A wearable user identification system comprising:
A wearing user identification method in a wearing user identification device that includes a processor and identifies a user who is an identification target wearing a sensor on a part of the body, comprising:
generating, by the processor, a feature quantity representing the vibration characteristics from a measurement signal according to the vibration characteristics of the user's body part measured by the sensor;
determining, by the processor, whether or not the state of the site where the sensor is attached to the user is stable based on the magnitude of variation in the generated feature quantity;
When the processor determines that the state of the attachment site is stable, personal identification of the user is performed based on the generated feature amount.
Wearable user identification method.
A wearable user identification program that causes a processor to function as each part of the wearable user identification device according to any one of claims 1 to 5.