JP2023017923A - Biological information acquisition method, biological information acquisition model learning method, device, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a biological information acquisition method capable of acquiring biological information from measurement information acquired from a subject, a biological information acquisition model learning method, a device, and a program.

SOLUTION: A measurement information acquisition unit 25 of a biological information acquisition apparatus 10 acquires measurement information on a subject. A biological information acquisition unit 26 inputs the measurement information acquired by the measurement information acquisition unit 25 into a learned model learned beforehand for acquiring biological information indicating a state of a subject from the measurement information, and acquires the biological information on the subject.

SELECTED DRAWING: Figure 1

COPYRIGHT: (C)2023,JPO&INPIT

Description

The present disclosure relates to a biometric information acquisition method, a biometric information acquisition model learning method, an apparatus, and a program.

In recent years, the application of piezoelectric materials containing helical chiral polymers to piezoelectric devices such as sensors and actuators has been studied. Film-shaped piezoelectric bodies are used in such piezoelectric devices.

As the helical chiral polymer in the piezoelectric material, the use of optically active polymers such as polypeptides and polylactic acid polymers is drawing attention. Among them, polylactic acid-based polymers are known to exhibit piezoelectricity only by a mechanical stretching operation. It is known that a piezoelectric body using a polylactic acid-based polymer does not require poling treatment, and its piezoelectricity does not decrease over several years.

For example, as a piezoelectric body containing a polylactic acid-based polymer, a piezoelectric body having a large piezoelectric constant d14 and excellent transparency has been reported (see Patent Documents 1 and 2, for example).

Also, recently, attempts have been made to use a conductor coated with a material having piezoelectricity. For example, a piezo cable is known, which is composed of a center conductor, a piezoelectric material layer, an outer conductor and an outer jacket, which are coaxially arranged in order from the center toward the outside (see, for example, Patent Document 3).

Also, a piezoelectric unit is known in which a conductive fiber is covered with a fiber made of a piezoelectric polymer (see, for example, Patent Document 4).

Japanese Patent No. 4934235 WO2010/104196 JP-A-10-132669 WO2014/058077

By the way, for example, a piezoelectric element manufactured using a piezoelectric material as described in Patent Documents 1 to 4 is attached to a subject, and biological information of the subject is obtained from measurement information obtained from the piezoelectric element. In that case, it is necessary to apply appropriate conversion processing to the measurement information.

However, there is a problem that it is not known what kind of conversion processing should be performed on the measurement information obtained from the piezoelectric element attached to the subject to acquire the biological information of the subject.

An object of the present disclosure is to provide a biological information acquisition method, a biological information acquisition model learning method, an apparatus, and a program capable of acquiring biological information from measurement information obtained from a subject.

A biological information acquisition method of the present disclosure includes a step of acquiring measurement information obtained from a subject; and obtaining the biometric information of the subject by inputting it to a learned model.

Further, the trained model of the biological information acquisition method of the present disclosure may be a neural network trained in advance based on learning data representing a combination of the measurement information for learning and the biological information for learning. .

Further, the measurement information in the biological information acquisition method of the present disclosure is piezoelectric information representing a voltage value obtained from a piezoelectric element in contact with the subject, and the biological information is a heartbeat representing a heartbeat state of the subject. can be informational.

The piezoelectric element may be a piezoelectric element installed in a chair on which the subject sits.

The biological information acquisition model learning method of the present disclosure includes a step of acquiring learning data representing a combination of measurement information of a subject for learning and biological information representing the state of the subject for learning; a step of acquiring a trained model for acquiring the biological information from the measurement information, based on the method, wherein a computer executes a process including:

The biological information acquisition model learning method of the present disclosure corrects the drift region of the measurement information of the subject for learning, uses the corrected measurement information as the biological information, and uses the measurement information for learning and the measurement information for learning. A step of generating a combination with biometric information as the learning data may be further included.

The biological information acquisition model learning method of the present disclosure excludes an unmeasured area from the measurement information of the subject for learning, uses the measurement information with the unmeasured area excluded as the biological information, and uses the measurement information for learning The method may further include a step of generating a combination of the measurement information and the biometric information for learning as the learning data.

The biological information acquisition device of the present disclosure includes a measurement information acquisition unit that acquires measurement information of a subject, and acquires biological information representing the state of the subject from the measurement information acquired by the measurement information acquisition unit. and a biometric information acquisition unit that acquires the biometric information of the subject by inputting it to a learned model that has been learned in advance to acquire the biometric information.

The biological information acquisition model learning device of the present disclosure includes a learning data acquisition unit that acquires learning data representing a combination of measurement information of a subject for learning and biological information representing the state of the subject for learning; a learning unit that acquires a trained model for acquiring the biological information from the measurement information based on the learning data acquired by the data acquisition unit.

A first program of the present disclosure comprises: a computer, a measurement information acquisition unit that acquires measurement information of a subject; It is a program for functioning as a biometric information acquiring unit that acquires the biometric information of the subject by inputting to a pre-learned model for acquiring information.

A second program of the present disclosure comprises a computer, a learning data acquisition unit for acquiring learning data representing a combination of measurement information of a subject for learning and biological information representing the state of the subject for learning, and the learning It is a program for functioning as a learning unit that acquires a trained model for acquiring the biological information from the measurement information based on the learning data acquired by the data acquisition unit.

According to the biological information acquisition method, the biological information acquisition model learning method, the device, and the program of the present disclosure, it is possible to acquire the biological information from the measurement information obtained from the subject.

BRIEF DESCRIPTION OF THE DRAWINGS It is the schematic which shows an example of a structure of the biometric information acquisition apparatus of this embodiment. 1 is a schematic block diagram of a computer functioning as a biometric information acquisition device; FIG. It is an explanatory view for explaining measurement information of this embodiment. FIG. 4 is an explanatory diagram for explaining learning data according to the embodiment; FIG. 4 is an explanatory diagram for explaining learning data according to the embodiment; FIG. 2 is an explanatory diagram for explaining U-Net, which is an example of a neural network used in this embodiment; FIG. 4 is a diagram showing an example of piezoelectric information input to a trained neural network; FIG. 4 is a diagram showing an example of heartbeat information output from a trained neural network; It is a figure which shows an example of the data generation processing routine for learning performed by a biometric information acquisition apparatus. It is a figure which shows an example of the learning processing routine performed by a biometric information acquisition apparatus. It is a figure which shows an example of the biometric information acquisition processing routine performed by a biometric information acquisition apparatus. It is a figure which shows the result of an Example. It is a figure which shows the result of an Example. It is a figure which shows the result of an Example. It is a figure which shows the result of an Example. It is a figure which shows the result of an Example. It is a figure which shows the result of an Example.

The present embodiment will be described in detail below.

<System Configuration of Biological Information Acquisition Apparatus According to the Present Embodiment>

FIG. 1 is a schematic diagram showing an example of the configuration of a biological information acquisition device 10 of this embodiment. The biological information acquisition device 10 having the configuration shown in FIG. 1 can be configured by a computer including a CPU, a RAM, and a ROM storing programs and various data for executing each processing routine to be described later.

For example, the biometric information acquisition device 10 can be realized by a computer 50 shown in FIG. The computer 50 includes a CPU 51 , a memory 52 as a temporary storage area, and a non-volatile storage section 53 . The computer 50 also includes an input/output interface (I/F) 54 to which an input/output device (not shown) is connected, and a read/write (R/W) unit 55 for controlling reading and writing of data to and from a recording medium. Prepare. The computer 50 also has a network I/F 56 connected to a network such as the Internet. The CPU 51 , memory 52 , storage unit 53 , input/output I/F 54 , R/W unit 55 and network I/F 56 are connected to each other via a bus 57 .

The storage unit 53 can be implemented by a hard disk drive (HDD), solid state drive (SSD), flash memory, or the like. A program for causing the computer 50 to function is stored in the storage unit 53 as a storage medium. The CPU 51 reads out the program from the storage unit 53, develops it in the memory 52, and sequentially executes the processes of the program.

This biometric information acquisition device 10 functionally includes an input unit 12, a calculation unit 14, and an output unit 16, as shown in FIG.

The biological information acquisition device 10 of the present embodiment converts measurement information obtained from a subject into biological information of the subject. FIG. 3 shows an explanatory diagram for explaining the measurement information of this embodiment. As shown in FIG. 3, a chair C has a piezoelectric element S installed thereon. The piezoelectric element S is positioned on the back side of the thigh of the subject U sitting on the chair C. As shown in FIG.

The biological information acquisition device 10 of the present embodiment uses piezoelectric information representing the voltage obtained from the piezoelectric element S in contact with the subject U as measurement information. Then, the biological information acquisition device 10 converts the piezoelectric information obtained from the piezoelectric element S into heartbeat information representing the heartbeat state of the subject U. FIG. Heartbeat information is an example of biological information representing the condition of a subject.

The input unit 12 receives piezoelectric information as measurement information obtained from a subject. The input unit 12 also receives a combination of piezoelectric information for learning and heartbeat information for learning.

A combination of learning piezoelectric information and learning heartbeat information is data obtained in advance from a subject. Further, in the combination of the piezoelectric information for learning and the heartbeat information for learning, the piezoelectric information for learning and the heartbeat information for learning correspond to the measured times.

For example, learning piezoelectric information obtained from a piezoelectric element installed on a chair on which a subject sits and heartbeat information obtained from an electrocardiograph installed on the subject's chest are acquired in advance as a set of data. . Based on these data, learning data is generated by the process described later.

Based on the combination of learning piezoelectric information and learning heartbeat information received by the input unit 12, the calculation unit 14 generates a trained model for obtaining heartbeat information from the piezoelectric information. Further, the calculation unit 14 converts the piezoelectric information received by the input unit 12 into heartbeat information. As shown in FIG. 1, the computing unit 14 includes a learning data generation unit 17, a learning data storage unit 18, a learning data acquisition unit 20, a learning unit 22, a trained model storage unit 24, A measurement information acquisition unit 25 and a biometric information acquisition unit 26 are provided.

The learning data generation unit 17 acquires each combination of learning piezoelectric information and learning heartbeat information received by the input unit 12 . Then, the learning data generation unit 17 processes the heartbeat information for learning to generate learning data.

FIG. 4 shows an explanatory diagram for explaining the learning data. The horizontal axis of graph D1 shown in FIG. 4 represents time, and the vertical axis represents voltage value. A solid line (ECG) shown in FIG. 4 is heartbeat information for learning. A dashed line (ECG_fir) shown in FIG. 4 is data obtained by processing heartbeat information for learning.

The time-series data of heart rate information for learning indicated by the solid line (ECG) shown in FIG. 4 is entirely wavy and contains a drift component. This drift component is generated, for example, by the subject's movement while measuring heartbeat information for learning.

Therefore, the learning data generation unit 17 processes the learning heartbeat information by applying an existing high-pass filter to the learning heartbeat information including the drift component. Whether or not the heartbeat information for learning includes a drift component is determined according to the peak value of the heartbeat information for learning. For example, when the peak value of the heartbeat information for learning is equal to or greater than a predetermined threshold value, the learning data generator 17 determines that the heartbeat information for learning includes a drift component. In this embodiment, an FIR (Finite Impulse Response) filter is used as the high-pass filter.

By applying the high-pass filter by the learning data generation unit 17, the heart rate information for learning indicated by the solid line (ECG) shown in FIG. 4 becomes time-series data indicated by the broken line (ECG_fir). The time-series data of the dashed line (ECG_fir) shown in FIG. 4 is data obtained when the number of taps, which are the parameters of the FIR filter, is set to 127 taps and the cutoff frequency is set to 1.5 Hz. Note that the number of taps is a parameter that indicates how much delay is calculated when filtering is performed.

For this reason, the learning data generation unit 17 of the present embodiment generates, as learning data, a combination of learning piezoelectric information and processed learning heartbeat information whose drift component has been corrected.

FIG. 5 shows another explanatory diagram for explaining the learning data. Similar to the graph D1 in FIG. 4, the horizontal axis of the graph D2 shown in FIG. 5 represents the time, and the vertical axis represents the voltage value.

As indicated by D2 in FIG. 5, the heartbeat information for learning may include an unmeasured region (a portion surrounded by an ellipse in the drawing). Therefore, the learning data generation unit 17 excludes the unmeasured region from the heart rate information for learning. Note that whether or not there is an unmeasured area of the heartbeat information for learning is determined according to the peak value of the heartbeat information for learning. For example, the learning data generation unit 17, when the peak value of the heartbeat information for learning is less than a value within a predetermined range (for example, -4 to +4) continues for a predetermined time or longer, the point contains an unmeasured region. Then, the learning data generation unit 17 generates, as learning data, a combination of the learning piezoelectric information and the processed heartbeat information for learning from which the unmeasured region is excluded.

Note that the learning data generation unit 17 generates the learning piezoelectric information and the learning heart rate information for the learning heart rate information that does not include the drift component and the learning heart rate information that does not include the unmeasured region. A combination with information is generated as it is as learning data.

Then, the learning data generation unit 17 stores each piece of learning data in the learning data storage unit 18 . These learning data are used for learning of a neural network, which will be described later. Thus, by correcting the drift component contained in the heartbeat information for learning and excluding the unmeasured region contained in the heartbeat information for learning, it is possible to obtain a highly accurate trained model.

Each of the learning data generated by the learning data generation unit 17 is stored in the learning data storage unit 18 . The learning data of the present embodiment is data representing a combination of the learning subject's piezoelectric information and the learning subject's heart rate information.

The learning data acquisition unit 20 acquires each piece of learning data stored in the learning data storage unit 18 .

The learning unit 22 learns a neural network, which is an example of a model for acquiring biological information from measurement information, based on the learning data acquired by the learning data acquiring unit 20 . Note that well-known deep learning or the like may be used as the learning algorithm. Then, the learning unit 22 acquires a trained neural network as an example of a trained model.

FIG. 6 shows an explanatory diagram for explaining the trained neural network of this embodiment. As shown in FIG. 6, in the present embodiment, a U-Net (for example, a reference (Olaf Ronneberger, Philipp Fischer, and Thomas Brox, "U- Net: Convolutional Networks for Biomedical Image Segmentation")). Note that the channels shown in FIG. 6 are U-Net parameters, and are set to values as shown in FIG. 6 in this embodiment.

When the subject's piezoelectric information as input data (INPUT) is input to the learned neural network as shown in FIG. 6, the subject's heart rate information is output as output data (OUTPUT).

For example, when piezoelectric information as shown in FIG. 7 is input to a trained neural network, heartbeat information as shown in FIG. 8 is obtained. The horizontal axis of the graphs in FIGS. 7 and 8 represents time, and the vertical axis represents voltage values.

The trained model storage unit 24 stores the trained neural network acquired by the learning unit 22 .

The measurement information acquisition unit 25 acquires the subject's piezoelectric information to be converted, which is received by the input unit 12 .

The biological information acquisition unit 26 acquires the piezoelectric information to be converted of the subject acquired by the measurement information acquisition unit 25 . In addition, the biometric information acquisition unit 26 reads the trained neural network stored in the trained model storage unit 24 . Then, the biological information acquisition unit 26 inputs the subject's piezoelectric information to be converted into the learned neural network to acquire the subject's heartbeat information.

In this way, learning data in which learning piezoelectric information and learning heartbeat information are associated is prepared, and a neural network is trained using this learning data, thereby converting piezoelectric information into heartbeat information. Processing is done properly. Therefore, the heartbeat information of the subject can be acquired sequentially without attaching an electrocardiograph to the subject.

The output unit 16 outputs the heartbeat information of the subject acquired by the biological information acquisition unit 26 as a result.

<Action of biological information acquisition device>

Next, with reference to FIGS. 9, 10, and 11, the action of the biological information acquisition device 10 will be described. When each combination of learning piezoelectric information and learning heartbeat information is input to the biological information acquisition device 10, the input unit 12 receives each combination of learning piezoelectric information and learning heartbeat information. . Then, the biological information acquisition device 10 executes a learning data generation processing routine shown in FIG.

<Learning data generation processing routine>

In step S<b>100 , the learning data generation unit 17 acquires each combination of learning piezoelectric information and learning heartbeat information received by the input unit 12 .

In step S102, the learning data generation unit 17 applies an existing high-pass filter to the learning heartbeat information containing the drift component among the learning heartbeat information acquired in step S100. Then, the learning data generation unit 17 generates, as learning data, a combination of the learning piezoelectric information and the processed heartbeat information for learning whose drift component has been corrected.

Further, in step S102, the learning data generation unit 17 excludes the unmeasured regions from the heart rate information for learning acquired in step S100. Then, the learning data generation unit 17 generates, as learning data, a combination of the learning piezoelectric information and the processed heartbeat information for learning from which the unmeasured region is excluded.

In step S104, the learning data generation unit 17 stores each of the learning data obtained in step S102 in the learning data storage unit 18, and ends the learning data generation processing routine.

Next, when receiving the start signal of the learning process, the biological information acquiring apparatus 10 executes the learning process routine shown in FIG. 10 .

<Learning processing routine>

In step S<b>200 , the learning data acquisition unit 20 acquires each piece of learning data stored in the learning data storage unit 18 .

In step S202, the learning unit 22 learns a neural network for acquiring heartbeat information from piezoelectric information based on each of the learning data acquired in step S200. Then, the learning unit 22 acquires a trained neural network.

In step S204, the learning unit 22 stores the trained neural network acquired in step S202 in the trained model storage unit 24, and ends the learning processing routine.

Next, when receiving the input of the piezoelectric information to be converted, the biological information acquisition apparatus 10 executes the biological information acquisition processing routine shown in FIG. 11 .

<Biological information acquisition processing routine>

In step S<b>300 , the measurement information acquisition unit 25 acquires the subject's piezoelectric information to be converted, which is received by the input unit 12 .

In step S<b>302 , the biometric information acquiring unit 26 reads the trained neural network stored in the trained model storage unit 24 .

In step S304, the biological information acquisition unit 26 inputs the subject's piezoelectric information to be converted, which was acquired in step S300, into the trained neural network read out in step S302, and obtains heartbeat information of the subject. get.

In step S306, the output unit 16 outputs the heartbeat information of the subject acquired in step S304 as a result.

As described above, the biological information acquisition device 10 according to the present embodiment inputs piezoelectric information obtained from a subject to a pre-learned neural network for acquiring heartbeat information from the piezoelectric information, Acquire the heartbeat information of the subject. Thereby, the heartbeat information can be obtained from the piezoelectric information obtained from the subject. In addition, the heartbeat information of the subject can be obtained sequentially without attaching an electrocardiograph to the subject.

Further, the biological information acquiring apparatus 10 according to the present embodiment acquires heartbeat information from piezoelectric information based on learning data representing a combination of piezoelectric information of a subject for learning and heartbeat information representing the state of the subject for learning. Get a trained model for acquisition. As a result, a trained model for obtaining heartbeat information from piezoelectric information obtained from a subject can be obtained. In addition, the biological information acquiring apparatus 10 according to the present embodiment excludes the drift component and the unmeasured region included in the heart rate information for learning when generating the data for learning. model can be obtained.

<Example>

Next, an example is shown. In this embodiment, the U-Net shown in FIG. 6 was learned. In this example, the U-Net model was learned. Also, when training the U-Net, 300 epochs of learning processing were executed for each. The number of learning data used for U-Net learning and the number of evaluation data used for evaluation of the trained U-Net model are as follows.

FIG. 12 shows changes in the accuracy of the learned U-Net when the learning process is performed using stationary data.

The horizontal axis of the graph shown in FIG. 12 is the epoch number representing the number of repetitions of learning, and the vertical axis is the heartbeat information output from the U-Net that has been trained and an electrocardiogram (portable electrocardiograph HCG-801 Omron Co., Ltd.) represents the difference from the actual heartbeat information (hereinafter simply referred to as "loss value").

As shown in FIG. 12, as learning progresses and the number of epochs increases, the loss value decreases, indicating that learning is being performed without problems. The dotted line (loss) shown in FIG. 12 represents the loss value between the heartbeat information output when the piezoelectric information for learning is input to the learned U-Net and the actual heartbeat information. there is The solid line (val_loss) shown in FIG. 12 indicates the difference between the heartbeat information output when the evaluation piezoelectric information different from the learning piezoelectric information is input to the learned U-Net and the actual heartbeat information. represents the loss value between

FIG. 13 shows the transition of the accuracy of the learned U-Net when the learning process is performed using the moving data. Similarly to FIG. 12, in FIG. 13 as well, as learning progresses and the number of epochs increases, the loss value decreases, indicating that learning is being performed without problems.

14, 15, 16, and 17 show waveforms of heartbeat information output when evaluation piezoelectric information different from the learning piezoelectric information is input to the learned U-Net. indicates FIGS. 14 and 15 show the results for stationary data, and FIGS. 16 and 17 show the results for moving data.

14, 15, 16, and 17, "X_test" represents the waveform of piezoelectric information for evaluation, "Y_test" represents the waveform of actually measured correct heartbeat information, and "Y_predict" represents The waveform of the heartbeat information output when the piezoelectric information for evaluation is input to the learned U-Net is shown.

As shown in FIGS. 14, 15, 16, and 17, appropriate conversion processing is performed by the learned U-Net, and heartbeat information can be accurately estimated from piezoelectric information.

The present invention is not limited to the above-described embodiments, and various modifications and applications are possible without departing from the gist of the present invention.

For example, in the above embodiment, the case of using U-Net using CNN as a model has been described as an example, but the present invention is not limited to this. Any model may be used as long as it is a trained model that can obtain heartbeat information, which is an example of biological information, from piezoelectric information, which is an example of measured information.

Further, in the above embodiment, the case of obtaining the piezoelectric information, which is an example of the measurement information, from the piezoelectric element installed on the chair has been described as an example, but the present invention is not limited to this. Any mode may be used as long as measurement information can be acquired from the subject.

Further, in the specification of the present application, an embodiment in which the program is pre-installed has been described, but it is also possible to store the program in a computer-readable recording medium and provide it.

10 biometric information acquisition device 12 input unit 14 calculation unit 16 output unit 17 learning data generation unit 18 learning data storage unit 20 learning data acquisition unit 22 learning unit 24 trained model storage unit 25 measurement information acquisition unit 26 acquisition of biometric information Department

Claims (12)

When generating learning data representing a combination of the measurement information of the subject for learning and the biometric information representing the state of the subject for learning, the drift component of the biometric information of the subject for learning is corrected. generating a combination of the measured information and the corrected biological information for learning as the learning data;
obtaining the generated learning data;
a step of acquiring a trained model for acquiring the biological information from the measurement information based on the acquired learning data;
A biological information acquisition model learning method in which a computer executes processing including When generating learning data representing a combination of the measured information of the subject for learning and the biological information representing the state of the subject for learning, excluding unmeasured regions of the biological information of the subject for learning, generating, as the learning data, a combination of the measurement information for learning and the biometric information for learning from which the unmeasured region is excluded;
obtaining the generated learning data;
a step of acquiring a trained model for acquiring the biological information from the measurement information based on the acquired learning data;
A biological information acquisition model learning method in which a computer executes processing including obtaining measurement information obtained from a subject;
A step of inputting the acquired measurement information into a learned model that has been pre-learned by the biometric information acquisition model learning method according to claim 1 or claim 2 to acquire the biometric information of the subject;
A biological information acquisition method in which a computer executes a process including The trained model is a neural network trained in advance based on learning data representing a combination of the measurement information for learning and the biological information for learning.
The biometric information acquisition method according to claim 3 . The measurement information is piezoelectric information representing a voltage value obtained from a piezoelectric element in contact with the subject,
The biological information is heartbeat information representing the heartbeat state of the subject,
The biometric information acquisition method according to claim 3 or 4. The piezoelectric element is a piezoelectric element installed in a chair on which the subject sits,
The biometric information acquisition method according to claim 5 . When generating learning data representing a combination of the measurement information of the subject for learning and the biometric information representing the state of the subject for learning, the drift component of the biometric information of the subject for learning is corrected. a learning data generation unit that generates a combination of the measurement information and the corrected learning biological information as the learning data;
a learning data acquisition unit that acquires the learning data generated by the learning data generation unit;
a learning unit that acquires a trained model for acquiring the biological information from the measurement information based on the learning data acquired by the learning data acquisition unit;
biometric information acquisition model learning device. When generating learning data representing a combination of the measured information of the subject for learning and the biological information representing the state of the subject for learning, excluding unmeasured regions of the biological information of the subject for learning, a learning data generation unit that generates, as the learning data, a combination of the measurement information for learning and the biological information for learning from which the unmeasured region is excluded;
a learning data acquisition unit that acquires the learning data generated by the learning data generation unit;
a learning unit that acquires a trained model for acquiring the biological information from the measurement information based on the learning data acquired by the learning data acquisition unit;
biometric information acquisition model learning device. a measurement information acquisition unit that acquires measurement information of a subject;
The measured information acquired by the measured information acquisition unit is input to a trained model that has been trained in advance by the biological information acquisition model learning device according to claim 7 or claim 8, and the biological information of the subject is obtained. a biometric information acquisition unit to acquire;
A biometric information acquisition device including. When generating learning data representing a combination of measurement information of a subject for learning and biological information representing the state of the subject for learning, the computer corrects a drift component in the biological information of the subject for learning, a learning data generation unit that generates a combination of the measurement information for learning and the corrected biological information for learning as the learning data;
a learning data acquisition unit that acquires the learning data generated by the learning data generation unit; and a learning data acquisition unit that acquires the biological information from the measurement information based on the learning data acquired by the learning data acquisition unit. A program that functions as a learning part that acquires a trained model for acquisition. When generating learning data representing a combination of the measured information of the subject for learning and the biometric information representing the state of the subject for learning, the computer excludes the unmeasured region in the biometric information of the subject for learning. and a learning data generation unit that generates, as the learning data, a combination of the measurement information for learning and the biological information for learning from which the unmeasured region is excluded,
a learning data acquisition unit that acquires the learning data generated by the learning data generation unit; and a learning data acquisition unit that acquires the biological information from the measurement information based on the learning data acquired by the learning data acquisition unit. A program that functions as a learning part that acquires a trained model for acquisition. A computer comprising: a measurement information acquisition unit that acquires measurement information of a subject; and inputting the measurement information acquired by the measurement information acquisition unit to a learned model pre-learned by the program according to claim 10 or claim 11. and functioning as a biological information acquisition unit that acquires the biological information of the subject.

Images