JP6943287B2 - Biometric information processing equipment, biometric information processing systems, biometric information processing methods, and programs - Google Patents

Description

The present disclosure relates to a technique for processing biometric information of a person.

There are problems of employee turnover and leave due to stress in the workplace, and the resulting increase in the burden on the company. In order to solve this problem, there are techniques and researches for estimating the degree of stress of the subject by measuring biological information (also called biological signal) such as heartbeat, sweating, and body temperature.

Among the biological information, the heartbeat information is closely related to stress and is useful information for estimating the degree of stress (hereinafter, also referred to as “stress estimation”). For example, Patent Document 1 describes an invention relating to a method for evaluating a degree of fatigue from information on a pulse wave generated by a heartbeat.

However, it is difficult to estimate stress with high accuracy simply by using biometric information. This is because the biological information varies depending on factors other than stress, for example, the degree of exercise of the person to be measured (hereinafter, also referred to as "exercise level" and "exercise state").

Therefore, there is a study to analyze the biological information and at the same time to estimate the motor state of the person to be measured at the time when the biological information is obtained. For example, Non-Patent Document 1 discloses a study on the relationship between electrodermal activity (EDA) and mental state. In this study, the number of occurrences of EDA peaks (ie, EDA peak appearance frequency) obtained from subjects by wearable sensors was measured every 30 seconds. In addition, the exercise level (sitting state, walking state, etc.) of the subject in each period was determined by the accelerometer, and the peak appearance frequency of EDA was tabulated for each exercise level. This study showed that the higher the exercise level, the higher the EDA amplitude and the frequency of EDA peak occurrence.

Non-Patent Document 2 also discloses a study in which a subject's skin temperature (ST) and skin conductance (SC) and other biological information as well as hand acceleration (Accelerometer data; ACC) are measured. ..

In addition, as a technique for processing information regarding exercise and biological information, a technique for calculating "physical strength age" from a pulse after exercise is described in Patent Document 2. However, the technique disclosed in Patent Document 2 does not include a process of estimating the exercise state because it is premised that the person to be measured performs a predetermined exercise. Therefore, it is not possible to estimate the degree of stress of a freely moving subject by the technique disclosed in Patent Document 2.

Further, Patent Document 3 discloses a technique of estimating an exercise frequency at a predetermined time interval from an acceleration sensor mounted on a person and extracting a scene (a group of the same operation contents) based on a change in the exercise frequency. ing. However, the technique disclosed in Patent Document 3 is not a technique for measuring a person's physical and mental condition. Although Patent Document 3 describes that body temperature is measured, body temperature is only used for calculating the average value of body temperature at predetermined time intervals.

International Publication No. 2005/000119 Japanese Unexamined Patent Publication No. 2011-161079 International Publication No. 2010/032579

A. Sano, “Measuring College Students ’Sleep, Stress, Mental Health and Wellbeing with Wearable Sensors and Mobile Phones,” Massachusetts Institute of Technology, 2015, pp.87-106. A. Sano et al., “Recognizing academic performance, sleep quality, stress level, and mental health using personality traits, wearable sensors and mobile phones,” 2015 IEEE 12th International Conference on Wearable and Implantable Body Sensor Networks (BSN), 2015, pp.1-6.

Among the index values based on biological information (hereinafter referred to as "features") used for stress estimation, the longer the measurement data, the more accurately the features are detected, and the more accurately they are detected. There are features that require sufficient length of measurement data.

For example, as feature amounts related to heartbeat, feature amounts called HF (High Frequency) component and LF (Low Frequency) component are known. The HF and LF components are waves with relatively long periods that appear in the time series data of heart rate variability. The longer the time series data used in the analysis, the more accurate these components will be detected. In particular, since the LF component is a component having a frequency in the range of about 0.05 Hz to 0.15 Hz, it is difficult to detect it from time series data of about 30 seconds.

In addition, the correlation dimension in the time-series data of heart rate variability is also one of the features that requires a relatively long time-series data in order to be detected with sufficient accuracy.

In the method of dividing biological information every 30 seconds and detecting the feature amount from each of the divided portions (segments) as disclosed in Non-Patent Documents 1 and 2, the feature amount as described above is detected accurately. Can not. That is, the more finely divided the time series data (that is, into shorter segments), the less useful information is lost, and the less accurate the stress estimation is, which is the first disadvantage. In other words, in order to perform highly accurate stress estimation based on long-period components, the longer the segment of the time series data for detecting the features, the better.

However, if the unit of segmentation of the time series data is set long, there is a second demerit that there is a high possibility that the motion state is not accurately determined. This is because there is a high possibility that two or more types of exercise states are included in one segment. Since the exercise state is determined for each segment, the longer the segment, the more likely it is that the transition of the exercise state cannot be detected when the actual exercise state is frequently switched.

That is, the shorter the length of the segment, which is the unit for analyzing the biological information and the exercise state, the larger the first demerit, and the longer the length, the larger the second demerit.

One of the objects of the present invention is to provide an apparatus and a method for performing analysis using features derived for each segment with higher accuracy.

The biometric information processing apparatus according to one aspect of the present invention is a portion of the time-series data based on the exercise state of the person to be measured during the period in which the biometric information is acquired from the time-series data of the biometric information of the person to be measured. At least one of a segment generation means for generating a segment, a feature amount derivation means for deriving a feature amount from the segment, the feature amount, and information on the state of the person to be measured based on the feature amount. It is provided with an output means for outputting.

The biometric information processing method according to one aspect of the present invention is a portion of the time-series data based on the exercise state of the person to be measured during the period in which the biometric information is acquired from the time-series data of the biometric information of the person to be measured. A segment is generated, a feature amount is derived from the segment, and at least one of the feature amount and information on the state of the person to be measured based on the feature amount is output.

The program according to one aspect of the present invention is a segment that is a part of the time-series data from the time-series data of the biometric information of the person to be measured, based on the exercise state of the person to be measured during the period in which the biometric information is acquired. Output of at least one of a segment generation process for generating Process and execute. The above program is stored, for example, on a computer-readable non-temporary storage medium.

According to the present invention, it is possible to perform analysis using the feature quantities derived for each segment with higher accuracy.

It is a block diagram which shows the structure of the biological information processing system which concerns on 1st Embodiment of this invention. It is a flowchart which shows the specific example of the flow of the segment generation processing. It is a figure which shows the concept of the derivation of a feature quantity. It is a figure which shows the concept of the integration of a feature quantity. It is a flowchart which shows the operation flow of the bio-information processing apparatus which concerns on 1st Embodiment. It is a block diagram which shows the structure of the biological information processing system which concerns on 2nd Embodiment of this invention. It is a figure which shows the concept of the allocation of a feature amount. It is a flowchart which shows the operation flow of the bio-information processing apparatus which concerns on 2nd Embodiment. It is a block diagram which shows the structure of the biological information processing system which concerns on 3rd Embodiment of this invention. It is a flowchart which shows the operation flow of the bio-information processing apparatus which concerns on 3rd Embodiment. It is a block diagram which shows the example of the hardware which comprises each part of each embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The operator "・" used in the calculation formula of the present disclosure is a multiplication operator, and the operator "/" is a division operator.

<< First Embodiment >>
First, the first embodiment of the present invention will be described.

<Structure>
FIG. 1 is a block diagram showing a configuration of a biometric information processing system 1 according to the first embodiment. The biometric information processing system 1 includes a biometric information processing device 11, a biometric information acquisition unit 5, and a motion information acquisition unit 6.

The biometric information processing device 11 includes a control unit 110, a storage unit 119, a motion state identification unit 111, a segment generation unit 112, a feature amount derivation unit 113, an integration unit 114, a stress estimation unit 115, and information output. A unit 116 is provided.

The biometric information processing device 11 is communicably connected to the biometric information acquisition unit 5 and the motion information acquisition unit 6. The biometric information processing device 11 receives the information acquired by the biometric information acquisition unit 5 and the motion information acquisition unit 6. The communication between the bio-information processing device 11 and the bio-information acquisition unit 5 and the communication between the bio-information processing device 11 and the motion information acquisition unit 6 may be performed by wire or wirelessly. May be good.

The biometric information processing device 11 is, for example, a server connected to the Internet. In this case, the biometric information processing device 11 may receive the information acquired by the biometric information acquisition unit 5 and the motion information acquisition unit 6 via the Internet. The biometric information processing device 11 may be an information processing device such as a portable information terminal or a PC (Personal Computer) possessed by the person to be measured. In that case, the biometric information processing device 11 may receive information from the biometric information acquisition unit 5 and the motion information acquisition unit 6 via, for example, a cable or communication using electromagnetic waves, sound waves, or the like. A part or all of the biometric information processing device 11, the biometric information acquisition unit 5, and the motion information acquisition unit 6 may be the same device.

=== Biometric information acquisition unit 5 ===
The biological information acquisition unit 5 acquires biological information.

Examples of biological information acquired by the biological information acquisition unit 5 include pulse wave (heartbeat), body temperature, brain wave, myoelectric potential, blood flow, blood component, breathing state, blinking state, degree of sweating, and the like. Can be mentioned. The biometric information acquired by the biometric information acquisition unit 5 may be any biometric information that shows the user's mental state, fatigue level, and the like. In the following description, the heartbeat is assumed as the acquired biometric information, but the heartbeat is just an example of the biometric information processed by the biometric information processing device 11.

The biological information acquisition unit 5 is, for example, a wearable sensing device such as a wristwatch type, a sticking type, or a wrapping type. The biological information acquisition unit 5 is brought into contact with the human body, for example, a photoelectric plethysmogram that acquires a pulse wave by PPG (Photoplethysmography) or an electrocardiograph that measures a heartbeat by measuring myocardial myocardium. It may be a sensor that acquires biometric information. Alternatively, the biological information acquisition unit 5 may be a device that estimates a pulse wave or a heartbeat in a non-contact manner using a camera or the like.

The biometric information acquisition unit 5 transmits the acquired biometric information to the biometric information processing device 11 together with the time information. The biometric information acquired by the biometric information acquisition unit 5 is stored in the storage unit 119 as time-series data by, for example, the control unit 110.

=== Motion information acquisition unit 6 ===
The motion information acquisition unit 6 acquires motion information, which is information that is an index of the strength of the exercise performed by the person to be measured. Motion information can be said to be an index showing the intensity of movement, an index showing the degree of physical load, or an index showing energy consumption. The motion information acquisition unit 6 may acquire information that can identify the state of physical activity that affects the biological information acquired by the biological information acquisition unit 5.

For example, the motion information acquisition unit 6 is an acceleration sensor attached to the human body. Alternatively, for example, the motion information acquisition unit 6 may be a combination of a camera and an image processing device that tracks the movement of a person photographed by the camera, or a depth sensor that acquires a distance to an object and a change in the distance. It may be a combination with an analyzer for analysis.

For example, the motion information acquisition unit 6 constantly (for example, 8 per second) constantly (for example, 8) accelerations (AcX, AcY, and AcZ) in the X, Y, and Z directions of a specific part of the person to be measured as motion information. Get (at the frequency of times). The X direction, the Y direction, and the Z direction are the directions defined in the motion information acquisition unit 6.

When the motion information acquisition unit 6 is an acceleration sensor attached to the subject, the motion information acquisition unit 6 may acquire the measured value itself as motion information.

When the motion information acquisition unit 6 is a set of a camera and an image processing device that tracks the movement of a person photographed by the camera, the motion information acquisition unit 6 performs image processing on the acceleration of a specific part of the person to be measured. It may be calculated by.

Even when the motion information acquisition unit 6 is a set of a depth sensor that acquires the distance to the target and an analyzer that analyzes the change in the distance, the motion information acquisition unit 6 determines the acceleration of a specific part of the person to be measured. , It may be calculated based on the fluctuation of the distance between the depth sensor and the portion.

The motion information may be information indicating only the time when the acceleration equal to or higher than a predetermined threshold value is measured. For example, the motion information acquisition unit 6 may be a device that detects acceleration equal to or higher than a predetermined threshold value, such as a pedometer.

The motion information acquisition unit 6 transmits the acquired motion information together with the time information to the biometric information processing device 11. The motion information acquired by the motion information acquisition unit 6 is recorded in the storage unit 119 by the control unit 110, for example, and is used for the motion state identification (described later) by the motion state identification unit 111.

When the biometric information acquisition unit 5 and the motion information acquisition unit 6 do not have the timekeeping function, the biometric information acquisition unit 5 and the motion information acquisition unit 6 send the acquired values to the biometric information processing device 11 at any time. In the biometric information processing device 11, the control unit 110 may add time information to the value.

Hereinafter, the functions of each of the components included in the biometric information processing apparatus 11 will be described. As will be described later, the control unit 110, the motion state identification unit 111, the segment generation unit 112, the feature amount derivation unit 113, the integration unit 114, the stress estimation unit 115, and the information output unit 116 execute instructions based on, for example, a program. It may be configured by a computer that includes one or more processors and memory.

In the present embodiment, the period (span) from which the biometric information to be processed by the biometric information processing apparatus 11 is acquired is defined as the period (span) from t = 0 to t = tmax.

=== Control unit 110 ===
The control unit 110 controls the operation of other components included in the biometric information processing device 11.

In addition, the control unit 110 controls the flow of data handled by the biometric information processing device 11. For example, the control unit 110 receives the information acquired by the biological information acquisition unit 5 and the motion information acquisition unit 6, and records the received information in the storage unit 119.

=== Storage unit 119 ===
The storage unit 119 temporarily or non-temporarily stores the data handled by the biometric information processing apparatus 11. The storage unit 119 is, so to speak, a working memory. The storage unit 119 may be a non-volatile storage medium. Other components included in the biometric information processing device 11 can freely read and write data to and from the storage unit 119.

=== Exercise state identification unit 111 ===
The motion state identification unit 111 identifies the motion state of the person to be measured in each of the small units of time included in the measurement period (span). In other words, the motion state identification unit 111 classifies the motion states in each of the small units of time into several types of motion states. There are three types of exercise states, for example, a "Sitting" state, a "Walking" state, and a "Running" state. The types and numbers of motor states identified are not limited to this. For example, the types of exercise states may be four types of "Sitting" (sitting) state, "Standing" (standing) state, "Walking" (walking) state, and "Running" state, and further "Running" to these four types. There may be five types with the addition of a "sleeping" state. The exercise state may be a parameter that identifies the exercise intensity of the person to be measured for each stage.

Each type of exercise state does not have to have a specific name such as "Sitting". For example, the types of exercise states may be three types: "first exercise state", "second exercise state", and "third exercise state".

The motion state identification unit 111 divides the time range of the entire data to be processed into sections (periods) for each time Ts, and identifies the motion state of the person to be measured in each section. The interval is set as, for example, interval N ₁ = [0, Ts], interval N ₂ = [Ts, 2Ts], ..., Section N _n = [(n-1) Ts, n · Ts]. NS. When the number of intervals can be generated as k, the maximum value of the above n is k. The section may be a closed section, an open section, or a half-open section.

The value of the section length Ts may be set in advance, or may be appropriately set according to the length of the data. Specific values of Ts are, for example, 1 second, 10 seconds, 30 seconds, or 1 minute. In order to discriminate the motion state more finely, the length Ts of the section should be short. However, if the exercise state is frequently switched, the segment described later becomes shorter. Since the suitable Ts value changes depending on the situation of the person to be measured and the like, it may be appropriately set by the designer or user of the biometric information processing apparatus 11.

The sections may be set so that the sections overlap each other. For example, the interval uses "W" as the degree of overlap, and the interval N ₁ = [0, Ts + W], the interval N ₂ = [Ts, 2Ts + W], ..., The interval N _n = [(n-1) Ts. , N · Ts + W].

Hereinafter, the motion state of the person to be measured in the section is also referred to as “the motion state of the section”.

A generally known method may be used as a method for identifying the motion state of the section. An example of a method for identifying the motion state of a section is shown below.

As an example, it is assumed that the exercise state identification unit 111 identifies the exercise state in a certain section Nx = [X1, X2].

The exercise state identification unit 111 may derive an index value indicating the intensity of exercise for each period (term) shorter than the length of the section, and identify the exercise state based on the index value. In the present disclosure, the index value indicating the intensity of exercise is referred to as “AM”. "AM" is an abbreviation for "Activity Magnitude".

For example, the motion state identification unit 111 equally divides the interval Nx = [X1, X2] into L periods (terms) (L is an arbitrary integer), and sets T ₀ = X1, T _L = X2, and division points. Let T ₁ , T ₂ , ..., _TL-1 . The motion state identification unit 111 derives AM for _{each period [0, T 1} ], ..., [ _TL-1 , _TL]. The period (term) may be a closed section, an open section, or a half-open section.

[Example of AM derivation method]
Let Dx = [Ta, Tb] be a certain period. An example of the method of deriving AM in the period Dx is shown below.

For example, it is assumed that the accelerations in the X direction, the Y direction, and the Z direction are acquired m times during the period Dx = [Ta, Tb] as motion information. i = 1, 2, · · ·, as m, the time at which acceleration is obtained with _{t i,} obtained at time _{t i,} X direction, Y direction, and the acceleration in the Z-direction, respectively, _ACX i, Let it be AcY _i and AcZ _i .

In this case, the exercise state identification unit 111 calculates AM in the period Dx = [Ta, Tb] based on, for example, the following mathematical formula.

RmX, RmY, and RmZ are moving averages of AcX, AcY, and AcZ for P seconds (the value of P is arbitrary) immediately before the interval Nx = [X1, X2], respectively. The value of P is, for example, 5.

Another example of the definition of AM is shown in equations (2) to (4).

The function H in equation (4) is a Heaviside function. That is, the value of H (x) is 1 when x exceeds 0 and 0 when x is 0 or less. AM derived by Equation (4) the interval Dx = [Ta, Tb] plurality of time points _t i that is included in (i = 1, 2, · · ·, m) of, a the acceleration threshold Th or Indicates the number at the time of

The threshold value Th in the equation (4) may be a value set by the designer. The threshold Th may be derived by machine learning using a decision tree or the like.

When the motion information is the time when the acceleration of the predetermined threshold value or more is measured, the motion state identification unit 111 determines the number of times that the acceleration of the predetermined threshold value or more is detected in the period Dx = [Ta, Tb]. It may be AM in [Ta, Tb].

[Example of method for identifying exercise state using AM]
Next, an example of a method of identifying the motion state of the section Nx = [X1, X2] using AM will be described.

The exercise state identification unit 111 compares, for example, AM with each of a number of boundary values one less than the type of exercise state. When there are three types of exercise states, two boundary values are used.

As an example, the movement state identification unit 111, time _D i ₌ included in the interval Nx = [X1, X2] [ T i-1, T i] (i = 1, ···, L) and AM in each of the Compare with the two boundary values "WS" and "RW". Then, the exercise state identification unit 111 records the number of AMs with AM ≦ WS, the number of AMs with WS ≦ AM ≦ RW, and the number of AMs with RW ≦ AM among each AM. Then, the motion state identification unit 111 determines the motion state in the section Nx = [X1, X2] by the following procedure.
-When the number of AMs with AM ≤ WS is the largest, the exercise state is "Sitting".
・ When the number of AMs with WS ≤ AM <RW is the largest, the exercise state is “Walking”.
・ When the number of AMs with RW <AM is the largest, the exercise state is “Running”.
However, if the maximum number of the number of AMs with AM ≤ WS, the number of AMs with WS ≤ AM ≤ RW, and the number of AMs with RW ≤ AM is 2 or more, the exercise can be determined. It may be designed to prioritize and set any one of the states.

The above procedure can also be realized by, for example, the calculations shown in the following equations (5) and (6). That is, by using the variable MaxCount defined by the equation (5) with the index value of each period included in the interval Nx = [X1, X2] as AM _{j (j = 1, ..., L), the interval} The value of the variable "State" representing the motion state of Nx = [X1, X2] may be determined by the equation (6).

なお、上記の境界値（ＲＷ、ＷＳ）は、設計者により設定された値でもよい。境界値は、決定木を用いた機械学習等によって導出されてもよい。

The above boundary values (RW, WS) may be values set by the designer. The boundary value may be derived by machine learning using a decision tree or the like.

[Another example of a method for identifying a motor state using AM]
In addition to the above, there are the following examples of methods for determining the exercise state by comparing each AM with a predetermined threshold value.

For example, the exercise state identification unit 111 imparts a “Running” state when the number of AMs with RW ≦ AM is equal to or greater than the first threshold value, and in other cases, the number of AMs with WS ≦ AM is increased. If it is equal to or higher than the second threshold value, the "Walking" state may be given, and in other cases, the "Sitting" state may be given.

Further, for example, the motion state identification unit 111 prepares the coefficients α ₁ , α ₂ , and α _{3 such that α 1} <α ₂ <α _{3 , and sets the first coefficient α 1} to the number of AMs for which AM ≦ WS. multiplying the second coefficient alpha ₂ over the number of AM is WS <AM ≦ RW, and RW <multiplying a third coefficient alpha ₃ in the number of AM is AM, to calculate the sum of these products May be good. The above sum is larger as the number of AMs larger than WS is larger, and is larger as the number of AMs larger than RW is larger. The exercise state identification unit 111 may determine the exercise state by comparing the calculated sum with a predetermined threshold value. For example, the exercise state identification unit 111 assigns a "Sitting" state when the calculated sum is equal to or less than the third threshold value, and the calculated sum is the fourth threshold value (however, the fourth threshold value> the third threshold value). If the threshold value is exceeded, a "Running" state may be given, and in other cases, a "Walking" state may be given.

Further, for example, the exercise state identification unit 111 may determine the exercise state by comparing the total sum of AM in each of the periods included in the section Nx with a predetermined threshold value. For example, when the total sum of AM is equal to or less than the fifth threshold value, the "Sitting" state is given, and when the total sum of AM exceeds the sixth threshold value (however, the sixth threshold value> the fifth threshold value), "Running" is given. A state may be given, otherwise a "Walking" state may be given.

As described above, the exercise state identification unit 111 identifies the exercise state of the person to be measured in each section.

The movement state identification unit 111 sends information related to the identified movement state to each section to the segment generation unit 112. The exercise state identification unit 111 may record information associated with the identified exercise state in each section in the storage unit 119.

=== Segment generator 112 ===
The segment generation unit 112 generates a segment from the time series data of the biological information based on the exercise state of each section identified by the exercise state identification unit 111.

The segment generation unit 112 generates a plurality of segments from the time series data so that each of the segments consists of either one section or continuous sections having the same motion state. ..

For example, the segment generation unit 112 divides the sections so that the continuous sections having the same motion state belong to the same segment and the continuous sections having different motion states belong to the different segments _Sn and _{Sn + 1.} ..

Specifically, for example, the segment generation unit 112 examines the motion state of each section in chronological order, and for a section in which the motion state does not change from the motion state of the immediately preceding section, the section is the same as the immediately preceding section. Set as a section belonging to a segment. For the section where the motion state changes from the motion state of the immediately preceding section, the segment generation unit 112 sets the section immediately before the motion state changes as the last section of the segment, and immediately after the motion state changes. Set the interval to the first interval of the new segment.

The segment generation unit 112 may generate a segment by performing processing according to the flowchart shown in FIG. 2, for example. The flowchart of FIG. 2, "i", and while increasing by 1 from 1 to k (the total number of intervals), showing a procedure for determining a segment S _n to N _i belongs. “N” is the segment number. The initial value of S _n for all n is a phi (empty set). Segment generating unit 112, first, the value of n to 1, after setting the value of i to 1 (step S21), and add a section _{N i} in the segment _{S n} (step S22), and namely, the interval _{N i} inclusion range in the scope of the segment S _n. After that, the segment generation unit 112 performs the determination process of step S24 if i = k (NO in step S23), and ends the process if i = k (YES in step S23). _{In step S24, the segment generation unit 112 increments “n” by 1 when State i,} which is a value representing the motion state of the i-th section, is not equal to State _{i + 1} (NO in step S24) (step S25). , "I" is incremented by 1 (step S26). When State _i is _{equal to State i + 1} (YES in step S24), the segment generation unit 112 increments “i” by 1 without changing “n” (step S26). By doing so, the continuous sections having the same _{kinetic state belong to the same segment Sn,} and the continuous sections having different kinetic states belong to different segments ( _Sn or _{Sn + 1} ). After step S26, the process of the segment generation unit 112 returns to the process of step S22.

When the processing of the flowchart of FIG. 2 is completed, the generation of _{n segments Sn is completed.}

If the segment becomes too long, there may be a problem that it takes an enormous amount of time to derive the feature amount from the segment. Therefore, an upper limit may be set for the length of the segment.

For example, if there is a segment whose length exceeds the upper limit (for example, 600 seconds) after all the segments are generated, the segment generation unit 112 divides the segment and the length of each segment after the division is the upper limit. It may be as follows.

Alternatively, the segment generation unit 112 may generate each segment so that the length of the segment does not exceed the upper limit in the process of generating the segment.

The segment generation unit 112 assigns a label indicating the type of the segment to each of the generated segments. Labels are used to distinguish how to treat a segment. The segment generation unit 112 assigns a label based on the motion state of the section constituting the segment. For example, the segment generation unit 112 assigns a label indicating the “Sitting” state to the segment consisting of the section in the “Sitting” state. Similarly, the segment generation unit 112 may label each segment with a label indicating either the "Sitting" state, the "Walking" state, or the "Running" state. Hereinafter, the label given to the segment is also referred to as “segment label”.

The segment generation unit 112 sends information on the generated segment (that is, information indicating a range or interval set and a given label) to the feature amount derivation unit 113. The segment generation unit 112 may record the information of the generated segment in the storage unit 119.

=== Feature Derivation Unit 113 ===
The feature amount derivation unit 113 analyzes the biometric information acquired by the biometric information acquisition unit 5 for each segment set by the segment generation unit 112, and derives the feature amount from the biometric information of each segment.

FIG. 3 is a diagram showing the concept of deriving the feature amount for each segment by the feature amount deriving unit 113. In the example shown in FIG. 3, the period from t = 0 to t = tmax from which the biometric information was acquired is divided into segments labeled with “Sitting” or the like. Although the lengths of the n segments are non-uniform, the feature amount deriving unit 113 has the same number of features (one in the example shown in FIG. 3) (C ₁ to C _{n) from the time series data of each segment.} ) Is derived.

The feature amount may be, for example, a representative value of data values (maximum value, minimum value, median value, mode value, average value, etc.) or a value obtained by performing a predetermined calculation on time series data. good. The feature amount deriving unit 113 may derive a plurality of feature amounts per segment.

For example, when the time-series data is the time-series data of the heart rate variability, the feature amount deriving unit 113 may derive the value of the HF component and the value of the LF component as the feature amount. The values of the HF component and the LF component are obtained by converting the time series data of the heart rate variability into the frequency component. Further, the value of LF / HF, which is the value obtained by dividing the value of the LF component by the value of the HF component, is also often used for stress estimation. The feature amount derivation unit 113 may derive the value of LF / HF as a feature amount.

The feature amount derivation unit 113 sends the feature amount from each segment to the integration unit 114. The feature amount deriving unit 113 may record the information in which each segment and the feature amount are related to each other in the storage unit 119.

=== Integrated unit 114 ===
The integration unit 114 integrates the feature quantities derived from each segment to derive a representative feature quantity. The typical feature amount is a feature amount that represents the feature amount obtained over the entire data to be processed. Typical features are used for stress estimation, which will be described later.

The integration unit 114 derives a representative feature amount for each label. That is, the integration unit 114 integrates the feature quantities for each segment with the same label and derives a representative feature quantity.

FIG. 4 is a diagram showing a concept of deriving a typical feature amount. The example shown in FIG. 4 is an example in which all three types of feature quantities, which are derived one by one from each segment by the feature quantity derivation unit 113, are integrated. In this example, integration derives three representative features (C_sitting, C_walking, and C_running). When a plurality of types of feature quantities are derived from one segment, typical feature quantities may be derived for each type of feature quantity.

Hereinafter, a typical feature amount obtained as a result of integration by label will be referred to as a “feature amount by label”. Since the label is given based on the exercise state, the feature amount for each label can be said to be the feature amount for each exercise state.

An example of a method for deriving the feature amount for each label related to a specific label (that is, a typical feature amount obtained by integrating the feature amount derived from the segment to which the specific label is given) will be described below. ..

For example, the integration unit 114 adds up each of the feature quantities derived from the segment to which a specific label is given after weighting each of the feature quantities according to the length of the segment from which the feature quantity is derived. , Derivation of label-specific features for that particular label.

For example, when the average heart rate values Ra, Rb, and Rc are derived as feature amounts from the three segments Sa, Sb, and Sc labeled to indicate the “Sitting” state, the integration unit 114 may perform the above 3 Using the lengths La, Lb, Lc of one segment as weighting factors,
(La ・ Ra + Lb ・ Rb + Lc ・ Rc) / (La + Lb + Lc)
The value derived from is a typical feature quantity.

The relationship between the weighting coefficient used for weighting and the length of the segment does not necessarily have to be a linear relationship. When a long-period component such as an HF component or an LF component is used as a feature amount, it is considered that the longer the segment length, the higher the reliability of the feature amount. In such a case, the weighting coefficient may be set so that the feature amount derived from the segment having a long length has a weight larger than the length.

For example, suppose that _{C 1} , C ₂ , ..., C _Q are derived as feature quantities from _{Q segments S 1} , S ₂ , ..., S _{Q with the same label, respectively.} Segment _{S i (i = 1, ···} , Q) when the length of the _{L i,} segment generation unit 112, the label feature quantities, the real _x 1 of any _positive, x 2 _(x 1 <x _{For 2} ), using a function f such that f (x ₁ ) / x ₁ <f (x ₂ ) / x _{2 holds,}
(Σ _{i = 1..Q} C _i · f (L _i )) / (Σ _{i = 1..Q} f (L _i ))
It may be derived by. f (x) = x ^α (α is an arbitrary real number of 1 or more) is an example of the function f.

According to such a calculation, typical features (first feature amount and second feature amount) derived from segments of different lengths (first length and second length) are represented. The ratio of the contribution ratio of the second feature quantity to the contribution ratio of the first feature quantity to the quantity is larger than the ratio of the second length to the first length.

The integration unit 114 sends a representative feature amount for each label derived by integration to the stress estimation unit 115.

=== Stress estimation unit 115 ===
The stress estimation unit 115 determines the degree of stress (which may be rephrased as the intensity or level of stress) of the person to be measured from the representative feature amount derived by the integration unit 114.

The stress estimation unit 115 has, for example, a stress estimation model generated by learning using a training data set in advance. Then, the stress estimation unit 115 applies the stress estimation model to a typical feature amount, and derives information indicating the degree of stress as an output.

The stress estimation model inputs at least one label-specific feature amount. The stress estimation model may be a model in which a set of a motion state and a feature amount for each label is input. The stress estimation model may be a model in which a set of a plurality of exercise states and a label-specific feature amount related to each of the plurality of exercise states is input. The information indicating the degree of stress may be a binary value of "no problem" or "problem", or may be numerical information such as "the degree of stress is xx%".

The training data set is, for example, a data set of label-specific features obtained from a person whose degree of stress is known. As a method of knowing the degree of stress of a person in order to prepare a training data set, a method of obtaining the degree of stress from information other than biometric information, such as a survey of the degree of stress by a questionnaire, may be adopted.

=== Information output unit 116 ===
The information output unit 116 outputs information indicating the result of processing by the biometric information processing device 11.

For example, the information output unit 116 outputs the information to the display device so that the information is displayed on the display device.

The information output unit 116 may transmit information to a device other than the biometric information processing device 11. For example, the information output unit 116 may transmit information to the biological information acquisition unit 5. When the biometric information acquisition unit 5 has a function of displaying information, the biometric information acquisition unit 5 may display the received information. According to such a configuration, the person to be measured can know the result of the processing by the biometric information processing apparatus 11.

The information output unit 116 may store the information in the storage device by outputting the information to the storage device. The storage device may be, for example, a non-temporary storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), or a temporary storage device such as a RAM (Random Access Memory). If a person such as a person to be measured accesses the information stored in the storage device, the person can know the result of processing by the biometric information processing device 11.

The information output by the information output unit 116 is, for example, information indicating the degree of stress. It can be said that the information indicating the degree of stress is information on the condition of the subject. The information output unit 116 may output the feature amount derived from each segment, the representative feature amount, the overall feature amount, and the like.

<Operation>
Next, the operation flow of the biometric information processing apparatus 11 will be described with reference to the flowchart of FIG.

First, the control unit 110 of the biometric information processing device 11 receives the biometric information and the motion information (step S110). The control unit 110 may record the biological information and the motion information in the storage unit 119.

Next, the motion state identification unit 111 identifies the motion state of each section based on the motion information acquired by the motion information acquisition unit 6 (step S111).

Next, the segment generation unit 112 generates data segments based on the motion state of each section (step S112).

Next, the feature amount deriving unit 113 derives the feature amount from each segment (step S113).

Next, the integration unit 114 integrates the feature amounts for each label and derives the feature amounts for each label (step S114).

Next, the stress estimation unit 115 determines the degree of stress of the person to be measured (step S115).

Then, the information output unit 116 outputs the result of information processing by the biometric information processing device 11 (step S116). That is, the information output unit 116 outputs, for example, information indicating the degree of stress of the person to be measured.

<Effect>
According to the biometric information processing apparatus 11 according to the first embodiment, it is possible to process the biometric information and perform accurate stress estimation while considering the influence of the exercise state on the biological information.

The reason is that the feature quantity derivation unit 113 and the integration unit 114 derive representative feature quantities for each label (by motion state), and the stress estimation unit 115 stresses based on the representative feature quantity and the stress model. This is because the estimation is performed.

Further, since the biometric information processing apparatus 11 can flexibly detect changes in the motion state and sufficiently lengthen the segment length, it is possible to perform accurate stress estimation.

The reason why the change in the motion state can be flexibly detected is that the length of the section (Ts) can be set to an arbitrary value (for example, 1 second) regardless of the length of the segment.

The reason why the length of the segment can be made sufficiently long is that the segment generation unit 112 does not generate the section as a segment, but the continuous section to which the same motion state is given belongs to the same segment. , Because it creates a segment.

As described above, the biometric information processing apparatus 11 according to the present embodiment has an effect that the analysis using the feature amount derived for each segment can be performed with higher accuracy.

In particular, if the feature amount is information that can be detected more accurately as the segment becomes longer, such as the values of the HF component and the LF component, the above-mentioned effect is large.

<< Modification example >>
[Modification 1] Consideration of time when exercise is reflected in biological information The influence of exercise state is reflected in biological information after a certain period of time has passed since the exercise state changed, such as heartbeat and sweating. There is biometric information. In one embodiment in which such biometric information is used for stress estimation, the segment generator 112 sets the segment boundary at a time point Y seconds (Y is a non-negative real number) after the section in which the motor state changes. May be good.

In other words, the segment generation unit 112 sets the range from the time point Y ₁ second after the change of the motion state to the time point Y ₂ seconds after the next change of the motion state as the range of one segment. You may. The values of each Y (Y ₁ , Y _2, etc. above) do not have to be equal. For example, the value of Y depends on the pattern of change in the exercise state (whether it is a change from the "Sitting" state to the "Walking" state, a change from the "Walking" state to the "Sitting" state, etc.). It may be a different value.

The value of Y may or may not be an integral multiple of the length of the interval.

Alternatively, the segment generation unit 112 may generate a segment so as not to include a period of Y seconds from the time when the exercise state changes. That is, the segment generation unit 112 may set the period from the time point after Y seconds from the time when the movement state changes to the time point when the movement state changes next as the range of one segment.

The segment generation unit 112 first generates a temporary segment according to the flowchart shown in FIG. 2, and then sets the period obtained by excluding the period of Y seconds from the time when the motion state changes from the temporary segment as the true segment (the true segment ( That is, it may be generated as a segment to be processed by the feature amount deriving unit 113).

By setting the segment in consideration of the period from the time when the exercise state changes to the time after the lapse of a predetermined time, the biological information and the exercise state can be more accurately associated with each other. This makes it possible to improve the accuracy of stress estimation using biological information, which takes time to be affected by changes in the motor state. In particular, by setting a time point after a predetermined time from the time when the exercise state changes as the start point of the segment, it is possible to prevent deterioration of the accuracy of stress estimation due to the use of biological information immediately after the change of the exercise state.

[Modification 2] Derivation of the entire feature amount In one embodiment, the integration unit 114 integrates all the segments included in the entire range of the data to be processed and derives the entire feature amount. May be good. For example, the integration unit 114 may derive the entire feature amount by integrating the label-specific feature amounts derived for each label.

Then, the stress estimation unit 115 may estimate the degree of stress from the total feature amount by using the stress estimation model obtained by learning the relationship between the total feature amount and the stress.

When deriving the entire feature amount, the integration unit 114 may perform at least one of correction or weighting according to the label for each of the feature amounts for each label.

For example, feature quantity, average body temperature, etc., if such a strong and characteristic amount having a correlation of motion, integrating unit 114, the label feature quantities according to the "Walking" state _{V 1} reduced, the "Walking" state labels feature quantities may be performed V ₂ reduces correction according. V ₁ and V ₂ are values determined according to the state of motion represented by the label. V ₁ and V ₂ may be appropriately set based on the knowledge about the relationship between the exercise state and the change in the feature amount.

Further, when the feature amount is a feature amount whose accuracy correlates with the strength of motion, the integration unit 114 derives the entire feature amount after weighting according to the accuracy. May be good.

For example, in the exercise state (for example, "Sitting" state) given when the amount of exercise is small, it is considered that the biological information mainly depends on the mental state. In the exercise state (for example, "Running" state) given when the amount of exercise is large, it is considered that the biological information is strongly influenced by the exercise state. From this, it is considered that the more the feature amount when the exercise intensity is strong, the worse the accuracy as the value used for stress estimation.

Therefore, for example, the integration unit 114 weights the segment identified as the “Sitting” state so that the weight of the segment identified as the “Walking” state and the “Running” state is larger than the weight of the segment identified as the “Walking” state. Then, the entire feature amount may be derived. By doing so, the overall feature quantity, in which the more reliable feature quantity is weighted more, is derived.

The weighting factor used for weighting is, for example, a real number in the range 0 to 1. It is assumed that the larger the numerical value, the larger the weight. The weighting coefficient may be such that the weight of the segment labeled with the "Sitting" state is 1, and the weight of the segment labeled with the "Walking" or "Running" state is 0. In this case, only the features obtained from the segment labeled with the "Sitting" state are reflected in the overall features. In this way, more accurate stress estimation can be performed by using only the features during the period when the exercise intensity is relatively low for stress estimation.

Since the total feature quantity derived as described above is derived after at least one of correction and weighting is performed according to the motion state, the overall feature quantity is more accurate based on the total feature quantity. Stress estimation can be performed.

(Further modification example of modification 2)
As described above, when the integration unit 114 derives the overall feature amount, the segment generation unit 112 may give reliability to each segment instead of the label indicating the motion state. The reliability given to a segment indicates the high reliability that the features derived from the segment are useful as information for highly accurate stress estimation. The reliability association unit 128 associates a higher reliability with a segment having a higher reliability. The weighting factor described above is an example of reliability.

Then, the integration unit 114 weights the features derived from each segment so that the weights of the features derived from the segments with higher reliability become larger, and then derives the entire features. Just do it. As a result, the features derived from the more reliable segments contribute more to the overall features.

The segment generation unit 112 may determine the reliability of each segment based on the exercise state of the person to be measured in the segment. For example, the segment generator 112 imparts less reliability to the segment when the subject is in a stronger exercise state.

The exercise state of the person to be measured in the segment is the exercise state of the section constituting the segment. When there is more than one motion state in the section constituting the segment, the motion state of the subject in the segment is the dominant motion state among the motion states given to each of the sections constituting the segment. The dominant motion state may be, for example, the motion state having the largest number of motion states in the section constituting the segment, or includes a representative time point of the segment (for example, a center time point or an end time point). The exercise state of the section may be used.

In the embodiment in which the segment generation unit 112 imparts reliability to each segment, the segment generation unit 112 may omit the process of determining the motion state of each segment. The segment generation unit 112 may determine the reliability of each segment based on the AM of the period that affects the feature amount derived from the segment. The period that affects the feature amount derived from the segment is, for example, the period included in the range in which the range of the segment is shifted before Y seconds when the AM in a certain period affects the feature amount after Y seconds.

In the segment generation unit 112, for example, the larger the representative value (maximum value, minimum value, median value, mode value, average value, etc.) of AM during the period affecting the feature amount derived from the segment, the lower the value. Confidence may be given to the segment. In addition, the segment generation unit 112 may give a lower reliability to a segment having a lower reliability based on the distribution of AM, the number of AMs exceeding a predetermined value, and the like.

In the embodiment in which the segment generation unit 112 imparts reliability to each segment, the feature quantity deriving unit 113 does not have to derive the feature quantity from the segment to which 0 is assigned as the reliability. The storage unit 119 does not have to store the time series data in the range of the segment to which 0 is assigned as the reliability. As a result, the load on the biometric information processing device 11 can be reduced.

[Modification 3] Separation of segment In one embodiment, the segment generation unit 112 may exclude a segment having a length less than a predetermined length from the segment from which the feature amount is derived. In other words, the feature amount deriving unit 113 may not derive the feature amount from the segment having a predetermined length or more, but may derive the feature amount only from the segment having a predetermined length or more.

This guarantees that the feature amount that is detected more accurately as it is derived from the longer data is detected with a certain accuracy.

[Modification 4] Parallelization of Information Acquisition and Information Processing In one embodiment, the biometric information processing apparatus 11 may perform processing on biometric information and motion information while receiving biometric information and motion information.

In such an embodiment, for example, the motion state identification unit 111 identifies the motion state of the section at any time when the biological information and the motion information of the section of a certain length are acquired. The segment generation unit 112 determines whether or not to switch the segment in the section in which the motion state is sequentially identified. The segment generation unit 112 does not switch the segment if the motion state of the section has not changed from the motion state of the immediately preceding section. When the motion state of the section is changed from the motion state of the immediately preceding section, the segment generation unit 112 switches the segment. That is, the segment generation unit 112 sets the time point at which the motion state changes as the end point of the current segment and the start point of the new segment.

By combining such an embodiment with the configuration of the above-described third modification, the segment generation unit 112 may impart reliability to the generated segment once the segment is generated. Then, the segment generation unit 112 may delete the time series data of the segment to which "0" is given as the reliability. That is, if there is a portion of the time-series data that is determined not to be used for deriving the feature amount by the feature amount deriving unit 113, the segment generation unit 112 may delete the time-series data of that portion.

When the segment is generated, the feature amount deriving unit 113 may derive the feature amount from the generated segment. Then, the feature amount deriving unit 113 may delete the time series data of the segment from which the feature amount is derived.

By performing information processing in real time in this way, it is not necessary to retain used data or unnecessary data until the measurement of the entire range is completed, and the load on the biometric information processing device 11 can be reduced. ..

<< Second Embodiment >>
In one embodiment, stress estimation may be performed without the process of integrating the feature quantities by the integration unit 114.

FIG. 6 is a block diagram showing a configuration of the biometric information processing system 2 according to the second embodiment. In the second embodiment, the biometric information processing apparatus 12 includes a feature amount allocation unit 124 instead of the integration unit 114, and a stress estimation unit 125 instead of the stress estimation unit 115. The components other than the feature amount allocation unit 124 and the stress estimation unit 125 may be the same as the components of the first embodiment.

The feature amount allocation unit 124 allocates a feature amount to each section based on the feature amount derived by the feature amount derivation unit 113. In the first embodiment, the typical feature amount is determined for each exercise state by integration, whereas in the second embodiment, the feature amount is determined for each section by allocation. However, the feature amount to be determined is a feature amount based on the feature amount derived from the segment.

Hereinafter, a method of allocating the feature amount will be described with reference to a specific example. For example, it is assumed that the feature amount deriving unit 113 derives the LF / HF value Cx as the feature amount from the segment consisting of the intervals Na, Nb, and Nc. In this case, the feature amount allocation unit 124 may determine the feature amount Cx as the feature amount assigned to the section Na, the section Nb, and the section Nc.

FIG. 7 is a conceptual diagram showing the feature amount assigned to each section when the feature amount is derived from each segment as shown in FIG. For example, the interval _N 1 to N ₄ contained in the segment _{S 1,} respectively, the feature amount _{C 1} which is derived from the segment _{S 1} is assigned.

In this way, if the derived feature amount is a feature amount whose value does not depend on the length of the segment, the feature amount allocation unit 124 constitutes the segment from which the feature amount is derived as it is. It may be assigned to the section to be used.

If the derived feature amount is a feature amount whose value is proportional to the length of the segment (cumulative value of characteristic signals, etc.), the feature amount allocation unit 124 derives the feature amount. The value divided by the number of sections constituting the segment may be assigned to the sections constituting the segment from which the feature quantity is derived.

Since the features assigned in this way are based on the features derived from the time-series data of the segment length, the accuracy is better than when derived from only one section (Na, etc.). There is expected.

The feature amount allocation unit 124 may record the assigned feature amount in, for example, the storage unit 119. As a result, the storage unit 119 stores a set of the section to which the motion state is given and the feature amount. If the feature amount allocation unit 124 allocates the feature amount to all the sections included in the entire range of the measurement data, data showing the transition of the set of the feature amount and the exercise state for each section is generated.

The stress estimation unit 125 estimates the stress using the feature amount of each section assigned by the feature amount allocation unit 124.

The stress estimation unit 125 has, for example, a stress estimation model generated by learning using a training data set in advance. Then, the stress estimation unit 125 applies the stress estimation model to the feature amount of each section determined by the feature amount allocation unit 124, and derives information indicating the degree of stress as an output. The information indicating the degree of stress may be the same as the information described in the description of the first embodiment.

In the stress estimation model, for example, a set of a motion state of a section and a feature amount is input. The stress estimation model may be a model in which the transition of the set of the feature amount and the exercise state for each section is input. Since the transition of the set of the feature amount and the motion state for each section has more information than the typical feature amount and the set of the motion state generated by the integration of the feature amount, it is expected that more accurate stress estimation can be performed. Will be done.

FIG. 8 is a flowchart showing an operation flow of the biometric information processing apparatus 12 according to the second embodiment. The processing of steps S110 to S113 and the processing of step S116 may be the same as the processing of the same reference numerals described in the first embodiment. The processing of the bio-information processing device 12 includes the processing of steps S124 and S125 instead of the processing of steps S114 and S115 as compared with the processing of the bio-information processing device 11.

In step S124, the feature amount allocating unit 124 allocates the feature amount to each section based on the feature amount derived by the process of step S113.

In step S125, the stress estimation unit 125 estimates the stress using the feature amount of each section assigned by the feature amount allocation unit 124.

According to the second embodiment, more accurate stress estimation can be performed. The reason is as already mentioned.

<< Third Embodiment >>
The biometric information processing apparatus 13 according to one embodiment will be described.

FIG. 9 is a block diagram showing the configuration of the biometric information processing device 13. The biometric information processing device 13 includes a segment generation unit 101, a feature amount derivation unit 102, and an output unit 103.

The segment generation unit 101 generates a segment that is a part of the time-series data from the time-series data of the biometric information of the person to be measured, based on the exercise state of the person to be measured during the period in which the biometric information is acquired. The segment generation unit 112 of each of the above embodiments is an example of the segment generation unit 101.

The state of motion is, for example, several types of parameters. The segment generation unit 101 may set a part or all of the period in which the exercise state is the same (that is, the period in which the same exercise state continues) as the period of the segment.

The feature amount deriving unit 102 derives the feature amount from the segment. The feature amount derivation unit 113 of each of the above embodiments is an example of the feature amount derivation unit 102.

The output unit 103 outputs at least one of the feature amount and the information regarding the state of the person to be measured based on the feature amount. The information output unit 116 of each of the above embodiments is an example of the output unit 103.

FIG. 10 is a flowchart showing the operation flow of the biometric information processing device 13.

In step S101, the segment generation unit 101 generates a segment from the time series data based on the exercise state of the subject.

In step S102, the feature amount deriving unit 102 derives the feature amount from the segment.

In step S103, the output unit 103 outputs at least one of the feature amount and the information regarding the state of the person to be measured based on the feature amount.

According to the biometric information processing apparatus 13, analysis using feature quantities derived for each segment can be performed with higher accuracy. The reason is that the segment generation unit 101 generates a segment based on the motion state of the subject, so that the motion state can be accurately identified and the accuracy of the detected feature amount can be guaranteed. .. The reason why the accuracy of the detected feature amount can be guaranteed is that there is no restriction that the length of the segment is the same as the unit for identifying the motion state.

<Hardware configuration that realizes each part of the embodiment>
In each embodiment of the present invention described above, each component of each device indicates a block of functional units.

The processing of each component may be realized, for example, by the computer system reading and executing a program stored in a computer-readable storage medium and causing the computer system to execute the processing. The "computer-readable storage medium" includes, for example, portable media such as optical disks, magnetic disks, magneto-optical disks, and non-volatile semiconductor memories, and ROMs (Read Only Memory) and hard disks built in computer systems. It is a storage device. A "computer-readable storage medium" is one that dynamically holds a program for a short period of time, such as a communication line when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In that case, it also includes the one that temporarily holds the program, such as the volatile memory inside the computer system that corresponds to the server or client. Further, the above-mentioned program may be a program for realizing a part of the above-mentioned functions, and may be a program for realizing the above-mentioned functions in combination with a program already stored in the computer system.

The "computer system" is, for example, a system including a computer 900 as shown in FIG. The computer 900 includes the following configurations.
-One or more CPUs (Central Processing Units) 901
-ROM902
・ RAM903
-Program 904A and storage information 904B loaded into RAM903
A storage device 905 that stores the program 904A and the storage information 904B.
Drive device 907 that reads and writes the storage medium 906.
-Communication interface 908 that connects to the communication network 909
-I / O interface 910 for inputting / outputting data
-Bus 911 connecting each component

For example, each component of each device in each embodiment is realized by the CPU 901 loading the program 904A that realizes the function of the component into the RAM 903 and executing the program 904A. The program 904A that realizes the functions of each component of each device is stored in, for example, in the storage device 905 or ROM 902 in advance. Then, the CPU 901 reads out the program 904A as needed. The storage device 905 is, for example, a hard disk. The program 904A may be supplied to the CPU 901 via the communication network 909, or may be stored in the storage medium 906 in advance, read by the drive device 907, and supplied to the CPU 901. The storage medium 906 is a portable medium such as an optical disk, a magnetic disk, a magneto-optical disk, and a non-volatile semiconductor memory.

There are various modifications in the method of realizing each device. For example, each device may be implemented by a possible combination of a computer 900 and a program, each of which is separate for each component. Further, a plurality of components included in each device may be realized by a possible combination of one computer 900 and a program.

In addition, some or all of the components of each device may be realized by other general-purpose or dedicated circuits, computers, or a combination thereof. These may be composed of a single chip or may be composed of a plurality of chips connected via a bus.

When a part or all of each component of each device is realized by a plurality of computers, circuits, etc., the plurality of computers, circuits, etc. may be centrally arranged or distributed. For example, a computer, a circuit, or the like may be realized as a form in which each is connected via a communication network, such as a client-and-server system or a cloud computing system.

Some or all of the above embodiments may also be described, but not limited to:

<< Additional notes >>
[Appendix 1]
A segment generation means that generates a segment that is a part of the time-series data based on the movement state of the person to be measured during the period in which the biometric information is acquired from the time-series data of the biometric information of the person to be measured.
A feature amount deriving means for deriving a feature amount from the segment,
An output means for outputting at least one of the feature amount and information on the state of the person to be measured based on the feature amount.
A biometric information processing device.
[Appendix 2]
The segment generation means sets a part or all of the period in which the exercise state is the same as the period of the segment.
The biometric information processing device according to Appendix 1.
[Appendix 3]
The segment generation means sets a time point after a predetermined time from the time point at which the motion state transitions as the start point of the segment.
The biometric information processing device according to Appendix 1 or 2.
[Appendix 4]
The feature quantity deriving means does not derive the feature quantity from the segment having a predetermined length.
The biometric information processing device according to any one of Appendix 1 to 3.
[Appendix 5]
The segment generation means assigns a plurality of the segments with labels based on the motion state in the segment.
The biometric information processing apparatus includes an integration means for deriving a representative feature amount by integrating the feature amounts derived from a plurality of the segments having the same label.
The output means outputs at least one of the representative feature amount and information on the state of the person to be measured based on the representative feature amount.
The biometric information processing device according to any one of Appendix 1 to 4.
[Appendix 6]
The integration means is derived from a first feature amount derived from the first segment of the first length and a second segment having a second length longer than the first segment. In the case of integrating the feature amounts of the above, the ratio of the contribution rate of the second feature amount to the contribution rate of the first feature amount with respect to the representative feature amount is the second length and the first feature amount. Weights are given to the first feature amount and the second feature amount so as to be larger than the ratio to the length of 1, and the first feature amount and the second feature amount are given based on the weights. Integrate,
The biometric information processing device according to Appendix 5.
[Appendix 7]
A stress estimation means for estimating the degree of stress of the subject is further provided based on the stress estimation model generated by learning the relationship between the typical feature amount and stress.
The output means outputs the degree of stress as information regarding the state of the person to be measured.
The biometric information processing device according to Appendix 5 or 6.
[Appendix 8]
The segment generation means has high reliability for the accuracy of the feature quantity derived from the segment based on at least one of the motion state in the segment and the length of the segment in each of the plurality of the segments. Gives confidence to indicate
The biometric information processing apparatus includes an integration means for deriving the entire feature amount by integrating the feature amounts derived from each of the plurality of segments after weighting according to the reliability. ,
The output means outputs at least one of the total feature amount and information on the state of the person to be measured based on the total feature amount.
The biometric information processing device according to Appendix 1.
[Appendix 9]
The biometric information processing apparatus according to Appendix 8, wherein the segment generation means imparts the lowest reliability to the segment having a length shorter than a predetermined length.
[Appendix 10]
The segment comprises one or more sets of sections to which the exercise state is assigned.
The biometric information processing device
A feature amount allocating means for allocating a value based on the feature amount derived from the segment to each of the sections constituting the segment, and
A stress estimation means for estimating the degree of stress of the subject using a stress estimation model that inputs a variation of a set of the feature amount and the exercise state assigned to the section, and a stress estimation means.
The biometric information processing apparatus according to Appendix 1.
[Appendix 11]
The biometric information processing device according to any one of Appendix 1 to 10.
The biometric information acquisition means for acquiring the biometric information and
The motion information acquisition means for acquiring the motion information representing the motion of the person to be measured includes.
The biometric information processing device includes a motion state identification means for identifying the motion state based on information representing the motion.
Biometric information processing system.
[Appendix 12]
From the time-series data of the biometric information of the person to be measured, a segment that is a part of the time-series data is generated based on the exercise state of the person to be measured during the period in which the biometric information is acquired.
Derivation of features from the segment
Outputs at least one of the feature amount and information on the state of the person to be measured based on the feature amount.
Biometric information processing method.
[Appendix 13]
In the generation of the segment, a part or all of the period in which the exercise state is the same is set as the period of the segment.
The biometric information processing method according to Appendix 12.
[Appendix 14]
In the generation of the segment, a time point after a predetermined time from the time point at which the motion state transitions is set as the start point of the segment.
The biometric information processing method according to Appendix 12 or 13.
[Appendix 15]
In deriving the feature amount, the feature amount is not derived from the segment having a length less than a predetermined length.
The biometric information processing method according to any one of Appendix 12 to 14.
[Appendix 16]
In the generation of the segment, a plurality of the segments are labeled based on the motion state in the segment.
By integrating the feature quantities derived from the plurality of segments having the same label, a representative feature quantity can be derived.
Outputs at least one of the representative feature amount and information on the state of the person to be measured based on the representative feature amount.
The biometric information processing method according to any one of Appendix 12 to 15.
[Appendix 17]
The first feature quantity derived from the first segment of the first length and the second feature quantity derived from the second segment having a second length longer than the first segment are In the case of integration, the ratio of the contribution rate of the second feature amount to the contribution rate of the first feature amount with respect to the representative feature amount is the ratio of the second length and the first length. The first feature amount and the second feature amount are weighted so as to be larger than the ratio, and the first feature amount and the second feature amount are integrated based on the weights.
The biometric information processing method according to Appendix 16.
[Appendix 18]
Based on the stress estimation model generated by learning the relationship between the typical features and stress, the degree of stress of the subject is estimated.
The degree of stress is output as information regarding the state of the person to be measured.
The biometric information processing method according to Appendix 16 or 17.
[Appendix 19]
After the segment is generated, each of the plurality of segments has a high degree of reliability for the accuracy of the feature quantity derived from the segment based on at least one of the motion state in the segment and the length of the segment. Gives confidence to indicate
The features derived from each of the plurality of segments are weighted according to the reliability and then integrated to derive the overall features.
Outputs at least one of the total feature amount and information on the state of the person to be measured based on the total feature amount.
The biometric information processing method according to Appendix 12.
[Appendix 20]
The biometric information processing method according to Appendix 19, which imparts the lowest reliability to the segment having a length less than a predetermined length.
[Appendix 21]
The segment comprises one or more sets of sections to which the exercise state is assigned.
A value based on the feature amount derived from the segment is assigned to each of the sections constituting the segment.
The degree of stress of the person to be measured is estimated by using a stress estimation model in which the variation of the set of the feature amount and the exercise state assigned to the section is input.
The biometric information processing method according to Appendix 12.
[Appendix 22]
From the time-series data of the biometric information of the person to be measured, a segment generation process for generating a segment which is a part of the time-series data based on the movement state of the person to be measured during the period when the biometric information is acquired, and
The feature amount derivation process for deriving the feature amount from the segment and
An output process that outputs at least one of the feature amount and information on the state of the person to be measured based on the feature amount, and
A computer-readable, non-temporary storage medium that stores programs that cause a computer to run.
[Appendix 23]
In the segment generation process, a part or all of the period in which the exercise state is the same is set as the period of the segment.
A computer-readable non-temporary storage medium according to Appendix 22.
[Appendix 24]
In the segment generation process, a time point after a predetermined time from the time point at which the motion state transitions is set as the start point of the segment.
A computer-readable, non-temporary storage medium according to Appendix 22 or 23.
[Appendix 25]
The feature amount derivation process does not derive the feature amount from the segment having a length less than a predetermined length.
The computer-readable non-temporary storage medium according to any one of Appendix 22 to 24.
[Appendix 26]
In the segment generation process, a plurality of the segments are labeled based on the motion state in the segment.
The program causes the computer to execute an integration process for deriving a representative feature amount by integrating the feature amounts derived from the plurality of segments having the same label.
The output process outputs at least one of the representative feature amount and information on the state of the person to be measured based on the representative feature amount.
A computer-readable non-temporary storage medium according to any one of Appendix 22 to 25.
[Appendix 27]
The integration process is derived from a first feature quantity derived from the first segment of the first length and a second segment having a second length longer than the first segment. In the case of integrating the feature amounts of the above, the ratio of the contribution rate of the second feature amount to the contribution rate of the first feature amount with respect to the representative feature amount is the second length and the first feature amount. Weights are given to the first feature amount and the second feature amount so as to be larger than the ratio to the length of 1, and the first feature amount and the second feature amount are given based on the weights. Integrate,
A computer-readable, non-temporary storage medium according to Appendix 26.
[Appendix 28]
The program causes the computer to further perform a stress estimation process for estimating the degree of stress of the subject based on a stress estimation model generated by learning the relationship between the representative feature amount and stress.
The output process outputs the degree of stress as information regarding the state of the person to be measured.
A computer-readable, non-temporary storage medium according to Appendix 26 or 27.
[Appendix 29]
The segment generation process is highly reliable for the accuracy of the feature quantity derived from the segment, based on at least one of the motion state in the segment and the length of the segment in each of the plurality of segments. Gives confidence to indicate
The program executes an integration process on the computer to derive the entire feature amount by integrating the feature amounts derived from each of the plurality of segments after weighting according to the reliability. Let me
The output process outputs at least one of the total feature amount and information on the state of the person to be measured based on the total feature amount.
A computer-readable non-temporary storage medium according to Appendix 22.
[Appendix 30]
The computer-readable non-temporary storage medium according to Appendix 29, wherein the segment generation process imparts the lowest reliability to the segment of less than a predetermined length.
[Appendix 31]
The segment comprises one or more sets of sections to which the exercise state is assigned.
The program
A feature amount allocation process that allocates a value based on the feature amount derived from the segment to each of the sections constituting the segment, and
A stress estimation process for estimating the degree of stress of the subject using a stress estimation model that inputs a variation of a set of the feature amount and the exercise state assigned to the section, and a stress estimation process.
A computer-readable, non-temporary storage medium according to Appendix 22, which causes the computer to execute the above.

Although the invention of the present application has been described above with reference to the embodiment, the invention of the present application is not limited to the above embodiment. Various changes that can be understood by those skilled in the art can be made within the scope of the present invention in terms of the structure and details of the present invention.

1, 2 Bio-information processing system 5 Bio-information acquisition unit 6 Motion information acquisition unit 11, 12, 13 Bio-information processing device 101 Segment generation unit 102 Feature quantity derivation unit 103 Output unit 110 Control unit 111 Motion state identification unit 112 Segment generation unit 113 Feature amount derivation unit 114 Integration unit 115 Stress estimation unit 116 Information output unit 119 Storage unit 124 Feature quantity allocation unit 125 Stress estimation unit 900 Computer 901 CPU
902 ROM
903 RAM
904A Program 904B Storage information 905 Storage device 906 Storage medium 907 Drive device 908 Communication interface 909 Communication network 910 Input / output interface 911 Bus

Claims (10)

A segment generation means that generates a segment that is a part of the time-series data based on the movement state of the person to be measured during the period in which the biometric information is acquired from the time-series data of the biometric information of the person to be measured.
A feature amount deriving means for deriving a feature amount from the segment,
An output means for outputting at least one of the feature amount and information on the state of the person to be measured based on the feature amount.
Equipped with a,
The segment generation means assigns a plurality of the segments with labels based on the motion state in the segment.
An integration means for deriving a representative feature amount by integrating the feature amounts derived from a plurality of the features having the same label is provided.
The output means outputs at least one of the representative feature amount and the information regarding the state of the person to be measured based on the representative feature amount.
The integration means is derived from a first feature amount derived from the first segment of the first length and a second segment having a second length longer than the first segment. In the case of integrating the feature amounts of the above, the ratio of the contribution rate of the second feature amount to the contribution rate of the first feature amount with respect to the representative feature amount is the second length and the first feature amount. Weights are given to the first feature amount and the second feature amount so as to be larger than the ratio to the length of 1, and the first feature amount and the second feature amount are given based on the weights. Integrate,
Biometric information processing device. The segment generation means sets a part or all of the period in which the exercise state is the same as the period of the segment.
The biometric information processing device according to claim 1. The segment generation means sets a time point after a predetermined time from the time point at which the motion state transitions as the start point of the segment.
The biometric information processing device according to claim 1 or 2. The feature quantity deriving means does not derive the feature quantity from the segment having a predetermined length.
The biometric information processing device according to any one of claims 1 to 3. A segment generation means that generates a segment that is a part of the time-series data based on the movement state of the person to be measured during the period in which the biometric information is acquired from the time-series data of the biometric information of the person to be measured.
A feature amount deriving means for deriving a feature amount from the segment,
An output means for outputting at least one of the feature amount and information on the state of the person to be measured based on the feature amount.
Equipped with a,
The segment generation means has high reliability for the accuracy of the feature quantity derived from the segment based on at least one of the motion state in the segment and the length of the segment in each of the plurality of the segments. Gives confidence to indicate
An integration means for deriving the entire feature amount by integrating the feature amounts derived from each of the plurality of segments after weighting according to the reliability is provided.
The output means outputs at least one of the total feature amount and information on the state of the person to be measured based on the total feature amount.
Biometric information processing device. The biometric information processing apparatus according to claim 5 , wherein the segment generation means imparts the lowest reliability to the segment having a length shorter than a predetermined length. From the time-series data of the biometric information of the person to be measured, a segment that is a part of the time-series data is generated based on the exercise state of the person to be measured during the period in which the biometric information is acquired.
Derivation of features from the segment
Output at least one of the feature amount and information on the state of the person to be measured based on the feature amount.
A plurality of the segments are labeled based on the motion state in the segment.
By integrating the feature quantities derived from the plurality of segments having the same label, a representative feature quantity can be derived.
At least one of the representative feature amount and the information on the state of the person to be measured based on the representative feature amount is output.
The first feature quantity derived from the first segment of the first length and the second feature quantity derived from the second segment having a second length longer than the first segment are In the case of integration, the ratio of the contribution rate of the second feature amount to the contribution rate of the first feature amount with respect to the representative feature amount is the ratio of the second length and the first length. The first feature amount and the second feature amount are weighted so as to be larger than the ratio, and the first feature amount and the second feature amount are integrated based on the weights.
Biometric information processing method. From the time-series data of the biometric information of the person to be measured, a segment that is a part of the time-series data is generated based on the exercise state of the person to be measured during the period in which the biometric information is acquired.
Derivation of features from the segment
Output at least one of the feature amount and information on the state of the person to be measured based on the feature amount.
Each of the plurality of said segments is given a reliability indicating a high degree of reliability with respect to the accuracy of the feature quantity derived from the segment based on at least one of the motion state in the segment and the length of the segment. death,
The features derived from each of the plurality of segments are weighted according to the reliability and then integrated to derive the overall features.
Outputs at least one of the total feature amount and information on the state of the person to be measured based on the total feature amount.
Biometric information processing method. From the time-series data of the biometric information of the person to be measured, a segment generation process for generating a segment which is a part of the time-series data based on the exercise state of the person to be measured during the period when the biometric information is acquired, and
The feature amount derivation process for deriving the feature amount from the segment and
An output process that outputs at least one of the feature amount and information on the state of the person to be measured based on the feature amount, and
Let the computer run
In the segment generation process, a plurality of the segments are given labels based on the motion state in the segment.
By integrating the feature quantities derived from the plurality of segments having the same label, a computer is made to execute an integration process for deriving a representative feature quantity.
In the output process, at least one of the representative feature amount and the information regarding the state of the person to be measured based on the representative feature amount is output.
In the integration process, the first feature quantity derived from the first segment of the first length and the second feature derived from the second segment having a second length longer than the first segment are derived. When integrating the feature amounts of the above, the ratio of the contribution rate of the second feature amount to the contribution rate of the first feature amount with respect to the representative feature amount is the second length and the first feature amount. Weights are given to the first feature amount and the second feature amount so as to be larger than the ratio to the length of 1, and the first feature amount and the second feature amount are given based on the weights. Integrate,
A program that lets a computer do things. From the time-series data of the biometric information of the person to be measured, a segment generation process for generating a segment which is a part of the time-series data based on the movement state of the person to be measured during the period when the biometric information is acquired, and
The feature amount derivation process for deriving the feature amount from the segment and
An output process that outputs at least one of the feature amount and information on the state of the person to be measured based on the feature amount, and
Let the computer run
In the segment generation process, each of the plurality of segments has high reliability for the accuracy of the feature quantity derived from the segment based on at least one of the motion state in the segment and the length of the segment. Gives confidence to indicate
The computer is made to execute an integration process for deriving the entire feature amount by integrating the feature amounts derived from each of the plurality of segments after weighting according to the reliability.
In the output process, at least one of the total feature amount and the information regarding the state of the person to be measured based on the total feature amount is output.
A program that lets a computer do things .

Images