JP2025068788A - Missing data complementing device, biological sensor, missing data complementing method, and program - Google Patents

Abstract

To provide a missing data complementing device capable of appropriately complementing missing of data, a biological sensor, a missing data complementing method, and a program.SOLUTION: A missing data complementing device includes: an acquisition part for acquiring time series measured data related to a plurality of pieces of biological information; and an estimation part which generates object biological information to be biological information including missing data in the measured data and time series data in reference biological information to be biological information other than the object biological information and estimates the missing data by simulation using the time series data.SELECTED DRAWING: Figure 1

Description

The present invention relates to a missing data completion device, a biosensor, a missing data completion method, and a program.

In recent years, wristwatch-type biosensors capable of acquiring various types of bioinformation have become known. For example, some biosensors simply measure, calculate, and acquire bioinformation measurement data in a time series, such as the pulse rate, pulse wave, pulse interval, body temperature, number of steps, and calories burned of the user wearing the biosensor (see, for example, Patent Document 1).

The biosensor may include, for example, a photoelectric sensor, an acceleration sensor, or other measurement sensor that measures changes in the user's biometrics, an ultraviolet (UV) sensor, etc. In the biosensor, for example, the control unit may convert the actual measurement value obtained by the measurement sensor into measurement data of biometric information, or may perform various calculations on the actual measurement value to obtain measurement data of biometric information.

JP 2016-131604 A

Conventional biosensors accumulate measurement data for, for example, about 5 to 7 days. For this reason, it is necessary to secure storage capacity, so for example, the frequency is set low for volume pulse waves based on information actually measured by an optical sensor. This reduces the accuracy of measuring the volume pulse wave, and missing values are often found in the measurement data of the heart rate calculated from the volume pulse wave. Even in normal measurements, missing values may occur in the measurement data due to body movement, etc. In particular, missing values often occur in the measurement data when exercising, etc. Missing values in the measurement data may cause inconvenience when checking using the measurement data. For example, pulse rate data is used to determine the user's autonomic nervous system, but if pulse rate data is missing, the accuracy of the determination of the autonomic nervous system decreases.

The present invention has been made in consideration of the above-mentioned circumstances, and aims to provide a missing data completion device, a biosensor, a missing data completion method, and a program that can appropriately complete missing data.

[1] In order to solve the above problem, one aspect of the present invention is a missing data completion device including an acquisition unit that acquires time-series measurement data related to multiple pieces of biological information, and an estimation unit that estimates the missing data by a simulation using target measurement data, which is measurement data of target biological information, which is biological information including missing data in the measurement data, and reference measurement data, which is measurement data of reference biological information, which is biological information other than the target biological information.

[2] In one aspect of the present invention, the target biological information is an average heart rate, and the reference biological information includes at least one of the number of steps, energy consumption, activity type, body movement type, sleep state, skin temperature, ultraviolet light level, and conversation time.

[3] In one aspect of the present invention, the estimation unit executes the simulation using multiple regression analysis using the target measurement data and the reference measurement data.

[4] In one aspect of the present invention, the estimation unit performs the multiple regression analysis using a forced input method.

[5] Also, one aspect of the present invention is a method and device for detecting a change in a living body, comprising: an actual measurement sensor for measuring a change in a living body; an acquisition unit for acquiring time-series measurement data on a plurality of pieces of biological information derived based on the biological changes; and an estimation unit for generating time-series data on target biological information, which is biological information including missing data in the measurement data, and reference biological information, which is biological information other than the target biological information, and estimating the missing data by a simulation using the time-series data;
The biosensor is equipped with a complementing unit that complements the missing data estimated by the estimation unit, and a providing unit that generates and provides provided data based on the measurement data from which the missing data has been complemented, wherein the target biometric information is an average heart rate, and the reference biometric information includes at least one of the number of steps, energy consumption, activity type, body movement type, sleep state, skin temperature, ultraviolet light level, or conversation time.

[6] In order to solve the above problem, one aspect of the present invention is a missing data completion method in which a computer acquires time-series measurement data for multiple pieces of biological information, generates time-series data for target biological information, which is biological information including missing data in the measurement data, and for reference biological information, which is biological information other than the target biological information, and estimates the missing data by a simulation using the time-series data.

[7] In order to solve the above problem, one aspect of the present invention is a program that causes a computer to acquire time-series measurement data for multiple pieces of biological information, generate time-series data for target biological information, which is biological information including missing data in the measurement data, and for reference biological information, which is biological information other than the target biological information, and estimate the missing data by a simulation using the time-series data.

The present invention provides a missing data completion device, a biosensor, a missing data completion method, and a program that can appropriately complete missing data.

1 is a diagram showing an example of a configuration of a biosensor 1 according to an embodiment; 4 is a flowchart illustrating an example of a process of the biosensor 1 according to the embodiment. FIG. 13 is a diagram showing HR data in verification 1 (during sleep). FIG. 13 shows HR data in verification 2 (during sleep). FIG. 13 shows HR data in verification 3 (during sleep). FIG. 13 shows HR data during verification 4 (exercise (yoga)). FIG. 13 shows HR data in verification 5 (walking).

The missing data completion device, biosensor, missing data completion method, and program according to the embodiment will be described with reference to Figures 1 to 7.

The biosensor of the embodiment is, for example, a wristwatch-type biosensor (wearable sensor) that is worn on the wrist by the user. The biosensor of the embodiment measures, for example, various data according to the user's activity, and calculates measurement data of multiple pieces of bioinformation related to various health conditions of the user based on the measured data, such as average heart rate, number of steps, and energy consumption. The biosensor provides the calculated measurement data to the user as part of the provided data. In addition to currently calculated measurement data, the biosensor may also include measurement data calculated in the past and stored in memory in the provided data.

FIG. 1 is a diagram showing an example of the configuration of a biosensor 1 according to an embodiment. The biosensor 1 according to the embodiment includes, for example, an input/output unit 10, a measurement sensor 20, a control unit 30, and a memory 40. The memory 40 stores measurement data 50 and measurement data 60. The input/output unit 10 enables a user to perform input operations and outputs information to the user. The measurement sensor 20 actually measures biochanges related to changes in a living organism. The biochanges include changes in the living organism itself, such as changes in pulse rate, movement, and skin temperature, as well as changes in the surrounding environment of the living organism, such as changes in the ultraviolet index.

The control unit 30 uses the actual measurement data measured by the measurement sensor 20 as the measurement data 50 or measurement data 60 as is, or estimates and complements missing data in the measurement data 60. The biosensor 1 of the embodiment includes a missing data generation device, and the control unit 30 functions as the missing data generation device.

The input/output unit 10 includes, for example, an input unit 11 and an output unit 12. The input unit 11 includes, for example, input devices such as a mouse and a keyboard. The input unit 11 may be configured as an interface that connects these input devices to the biosensor 1. The input unit 11 accepts input of various information to the biosensor 1. For example, when a user performs an input operation to request the provision of desired bioinformation, the input unit 11 generates request information corresponding to the input operation and outputs it to the control unit 30. The request information is, for example, the progress of the user's pulse rate and the progress of the number of steps for the past week.

The output unit 12 includes, for example, an output device such as a display or a speaker. The output unit 12 may be configured as an interface that connects these output devices to the biosensor 1. The output unit 12 outputs various information provided by the control unit 30 to the user. The input/output unit 10 may include, for example, a touch panel equipped with the input unit 11 and the output unit 12.

The actual measurement sensor 20 measures biological changes related to changes in the user. The actual measurement sensor 20 includes, for example, an optical sensor 21, an acceleration sensor 22, a UV sensor 23, a temperature sensor 24, and a microphone 25. The actual measurement sensor 20 generates various actual measurement data based on the actual measurement results by the optical sensor 21, the acceleration sensor 22, the UV sensor 23, the temperature sensor 24, and the microphone 25, and outputs the data to the control unit 30. The actual measurement sensor 20 may also include other sensors, such as a deep body thermometer that measures deep body temperature, a glucose sensor that detects glucose in interstitial fluid, and a posture sensor that detects posture.

The optical sensor 21 measures the user's volume pulse wave by irradiating the living body with light such as infrared light and measuring the light reflected from within the living body. The optical sensor 21 functions as a pulse sensor. The optical sensor 21 measures the user's volume pulse wave at intervals of, for example, 20 Hz (approximately 0.05 seconds) and generates data related to the user's pulse, for example, PPI (Peak to Peak Interval) data, as measured data. The optical sensor 21 may measure the user's volume pulse wave at intervals greater than or less than 20 Hz.

The acceleration sensor 22 measures the movement (acceleration) of the user wearing the biosensor 1. The acceleration sensor 22 is, for example, a three-axis sensor. The acceleration sensor 22 measures the user's acceleration in three-dimensional directions and generates data including the acceleration in the three-dimensional directions, for example, ACT (Activity) data, as measured data. The acceleration sensor 22 measures the user's acceleration in the three-dimensional directions at, for example, 20.5 Hz (approximately 0.448 seconds) and generates the ACT data. The acceleration sensor 22 may measure the user's acceleration in the three-dimensional directions at intervals greater than 20.5 Hz or less than 20.5 Hz.

The UV sensor 23 measures the ultraviolet rays in the surroundings. Based on the measured amount of ultraviolet rays, the UV sensor 23 generates data on the amount of ultraviolet rays (UV index), for example, UV measurement data, as actual measurement data. The temperature sensor 24 measures the temperature of the part of the user where the biosensor 1 is worn. Based on the temperature of the part of the user where the biosensor 1 is worn, the temperature sensor 24 generates data on the user's skin temperature, for example, skin temperature measurement data, as actual measurement data. The microphone 25 collects human voices. Based on the voice collection status, the microphone 25 generates data on whether the user is talking, for example, conversation data indicating that a conversation has taken place, as actual measurement data.

The control unit 30 includes, for example, an acquisition unit 31, a data management unit 32, an estimation unit 33, a completion unit 34, and a provision unit 35. The control unit 30 realizes these functions by, for example, a hardware processor executing a program stored in a storage device such as a memory 40.

Hardware processors include, for example, a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). Instead of storing a program in a storage device, the program may be directly built into the circuitry of the hardware processor. In this case, the hardware processor realizes its function by reading and executing the program built into the circuitry.

The hardware processor is not limited to being configured as a single circuit, but may be configured as a single hardware processor by combining multiple independent circuits to realize each function. Also, multiple components may be integrated into a single hardware processor to realize each function.

When the aggregated data generation condition is met, for example, when request information is output by the input unit 11, the control unit 30 aggregates the measurement data 60 stored in the memory 40 to generate aggregated data. The aggregated data is, for example, data obtained by aggregating each value (measurement value) stored in the memory 40 as the measurement data 60, for example, for about 30 days.

The control unit 30 generates provision data corresponding to the request information based on the generated aggregated data. The control unit 30 outputs the generated provision data to the output unit 12 that output the request information. The output unit 12 displays the output provision data on a display.

The acquisition unit 31 acquires various types of actual measurement data output by the actual measurement sensor 20. The acquisition unit 31 acquires the acquired actual measurement data as part of the measurement data 50 or the measurement data 60 and stores it in the memory 40. The measurement data 50 stored in the memory 40 includes PPI data 51 and ACT data 52. Due to capacity constraints, the memory 40 stores, for example, the PPI data 51 for the past three days and the ACT data 52 for the past 11 hours, and sequentially deletes old data as time passes.

The acquisition unit 31 acquires measurement data 60 relating to the user's biometric information in a time series based on the acquired actual measurement data. Depending on the type of measurement data 60, the acquisition unit 31 either uses the acquired actual measurement data as the measurement data 60 as is, or calculates (estimates) the measurement data 60 based on the acquired actual measurement data, thereby acquiring the time series measurement data 60 relating to the biometric information. The acquisition unit 31 acquires the measurement data 60 in one-minute increments, for example. The acquisition unit 31 stores the acquired measurement data 60 in the memory 40. The memory 40 stores the measurement data 60 for the past 30 days, and sequentially deletes older data. The time series measurement data 60 is an example of time series data.

Biometric information includes, for example, data on average heart rate (HR), number of steps (STP), energy expenditure (EE), activity type (AT), body motion (BM), sleep state (SS), skin temperature (SK), ultraviolet ray index (UV index), and conversation time (CT).

The acquisition unit 31 acquires HR data 61 relating to HR, STP data 62 relating to STP, EE data 63 relating to EE, AT data 64 relating to AT, BM data 65 relating to BM, SS data 66 relating to SS, SK data 67 relating to SK, UV data 68 relating to UV, and CT data 69 relating to CT.

The HR data 61 is data in which the HR calculated based on the PPI data 51 output by the actual measurement sensor 20 is arranged in a time series. The STP data 62 is data in which the number of steps of the user estimated based on the ACT data 52 output by the actual measurement sensor 20 is arranged in a time series.

The EE data 63 is, for example, data in which EE estimated based on the ACT data 52 output by the actual measurement sensor 20 and data such as weight stored in advance is arranged in chronological order. The AT data 64, BM data 65, and SS data 66 are, for example, data in which AT, BM, and SS calculated based on the ACT data 52 output by the actual measurement sensor 20 are arranged in chronological order.

SK data 67 is data in which actual skin temperature measurement data output by the actual measurement sensor 20 is arranged in chronological order. UV data 68 is data in which actual UV measurement data output by the actual measurement sensor 20 is arranged in chronological order. CT data 69 is data in which conversation times estimated based on conversation data output by the actual measurement sensor 20 are arranged in chronological order.

The data management unit 32 determines whether the aggregated data generation condition is met. The aggregated data generation condition is met, for example, when request information is output by the input unit 11. The aggregated data generation condition may be another condition. The aggregated data generation condition may be met, for example, every time a certain amount of time has passed, or when a certain amount of actual measurement data, measurement data, and measurement data acquired by the acquisition unit 31 has been acquired, or when a number of these conditions are satisfied.

When the aggregated data generation condition is met, the data management unit 32 determines whether the HR data 61 during the time period for generating the aggregated data includes missing data. The data management unit 32 provides the estimation unit 33 and the completion unit 34 with the result of the determination as to whether the HR data 61 includes missing data.

Of the measurement data 60, HR data 61 is data calculated based on the actual measurement value of the optical sensor 21, and is data that is relatively often missing depending on the conditions of the actual measurement by the optical sensor 21. The other measurement data 60 is data that is almost free of missing data. For this reason, HR data 61 may be measurement data that includes missing data. On the other hand, measurement data 60 of biological information other than HR data 61 can often be measured as measurement data even when HR data 61 is missing data, and the present invention uses them to supplement the missing data of HR data 61.

The average heart rate (HR), which is the subject of missing data complementation, is an example of target biological information. In addition, energy expenditure (EE), activity type (AT), body movement type (BM), sleep state (SS), skin temperature (SK), ultraviolet light level (UV), and conversation time (CT) are examples of reference biological information other than target biological information.

When the HR data 61 includes missing data, the estimation unit 33 performs a simulation using multiple regression analysis with the measurement data 60 including the missing data and the measurement data 60 not including the missing data to estimate the missing data. The measurement data 60 including the missing data includes the HR data 61 (hereinafter also referred to as the target measurement data). For the measurement data 60 not including the missing data, a simulation is performed using the STP data 62, EE data 63, AT data 64, BM data 65, SS data 66, SK data 67, UV data 68, and CT data 69 (hereinafter also referred to as the reference measurement data) to estimate the missing data.

The estimation unit 33 performs a simulation, for example, using multiple regression analysis using the target measurement data and the reference measurement data. The multiple regression analysis is performed, for example, by a forced input method in which all data, the target measurement data and the reference measurement data, are input. The estimation unit 33 may perform the multiple regression analysis by a method other than the forced input method, for example, a stepwise method, but when the amount of data (parameters) to be input is small, it is better to use the forced input method.

The estimation unit 33 performs a simulation in which the reference measurement data to be combined with the target measurement data is changed, and derives a model for calculating missing data (hereinafter, the missing data calculation model) using the reference measurement data that results in a combination with the lowest significance probability. The estimation unit 33 substitutes the reference measurement data at each time point into the derived missing data calculation model, thereby calculating an estimate of the target measurement data that has become missing data as estimated data.

The complementing unit 34 complements the estimated data calculated by the estimation unit 33 to the HR data 61, which is the target measurement data. The complementing unit 34 complements the missing data of the HR data 61 with the estimated data, thereby generating HR data 61 that has no missing data and is filled with data for each time.

The providing unit 35 generates provided data according to the requested information based on the measurement data 60 including the HR data 61 supplemented with the estimated data. For example, when the requested information is a change in the user's pulse rate over the past week, the providing unit 35 obtains the change in the pulse rate based on the HR data 61, outputs it to the output unit 12 as provided data, and provides it to the user by having the output unit 12 output it.

Next, the processing of the biosensor 1 of the embodiment will be described. Here, the processing in the control unit 30 of the biosensor 1 will be mainly described. FIG. 2 is a flowchart showing an example of the processing of the biosensor 1 of the embodiment. The control unit 30 of the biosensor 1 first determines in the acquisition unit 31 whether or not the actual measurement data output by the input unit 11 has been acquired (step S101).

If it is determined that actual measurement data has been acquired, the acquisition unit 31 generates measurement data and measurement data using the acquired actual measurement data and stores the data in the memory 40 (step S103). If it is determined that actual measurement data has not been acquired, the acquisition unit 31 skips the process of step S103 and proceeds to the process of step S105.

Then, the data management unit 32 determines whether the aggregate data generation conditions are met by, for example, outputting request information from the input unit 11 or the passage of a certain period of time (step S105). If it is determined that the aggregate data generation conditions are not met, the biosensor 1 ends the process shown in FIG. 2.

If it is determined that the aggregate data generation condition is met, the data management unit 32 reads out from the memory 40 the measurement data 60 for 30 days stored therein (step S107). Next, the data management unit 32 determines whether the HR data 61 of the read measurement data includes missing data (step S109).

When it is determined that the HR data 61 includes missing data, the estimation unit 33 performs a simulation using multiple regression analysis with a forced input method using the target measurement data (HR data 61) and the reference measurement data (STP data 62 to CT data 69) (step S111). The estimation unit 33 derives a missing data calculation model based on the results of the simulation (step S113).

When deriving the missing data calculation model, the estimation unit 33 selects one to eight different combinations of reference measurement data from the eight reference measurement data, and performs multiple regression analysis using each combination of reference measurement data and the HR data 61. The estimation unit 33 derives the missing data calculation model using the combination of reference measurement data with the lowest significance probability as a result of the multiple regression analysis and the HR data 61.

Then, the estimation unit 33 substitutes the reference measurement data at the time when the missing data occurred into the derived missing data calculation model (step S115). The estimation unit 33 generates estimated data that is estimated as missing data by substituting the reference measurement data into the missing data calculation model (step S117).

Then, the complementing unit 34 complements the estimated data generated by the estimation unit 33 as data for the time when the missing data occurred in the HR data 61, and generates HR data 61 without the missing data (step S119). Also, if it is determined in step S109 that the HR data 61 does not contain missing data, the data management unit 32 advances the process to step S121.

Then, the providing unit 35 generates provision data according to the request information using the measurement data without missing data (step S121). The providing unit 35 outputs the generated provision data to the output unit 12 and causes the output unit 12 to output the data, thereby providing the data to the user (step S123). In this way, the biosensor 1 ends the process shown in FIG. 2.

When target measurement data containing missing data occurs, the biosensor 1 of the embodiment estimates the missing data by performing a simulation using the target measurement data and reference measurement data. This makes it possible to generate highly accurate complementary data that complements the missing data, and to appropriately complement the missing data.

The biosensor 1 of the embodiment also uses multiple regression analysis when performing a simulation, which allows for higher accuracy of the complementary data. The biosensor 1 of the embodiment also performs multiple regression analysis using a forced input method, which allows for higher accuracy of the complementary data. The measurement data 60 used in the multiple regression analysis only needs to be that amount stored in the memory 40, so even with the forced input method, the computation load can be prevented from becoming too large.

Next, we will explain the results of a verification performed on a simulation to estimate missing data. The verification was performed by collecting measurement data 60 for three patterns of user behavior: "sleeping," "exercising (yoga)," and "walking," then generating a time series of measurement data 60 in which HR data 61 was missing from each of the measurement data, and simulating the missing values.

In the verification, a missing data calculation model was derived by multiple regression analysis using data with reference measurement data for 5 minutes before and after the time the missing data occurred as parameters as measurement data for estimating the missing data. The multiple regression analysis was performed using the forced entry method, and the derived missing data calculation model was used to estimate the missing data.

(Verification 1: During sleep, missing data = 1)
First, as the measurement data during the user's sleep, the measurement data of each reference biological information of HR, STP, EE, AT, BM, SS, SK, UV, and CT was acquired at one-minute intervals for a total of 11 minutes (t1 to t11). In verification 1, the missing data was set to 1, and one piece of HR data, here the HR data at t=6, was missing. FIG. 3 is a diagram showing HR data in verification 1 (during sleep). FIG. 3(a) is a time series display of the acquired HR data, FIG. 3(b) is a time series display with the missing time=t6, and FIG. 3(c) is a time series display with the missing data complemented. The missing time is the time when the measurement data was missing. The time series display of each measurement data of STP, EE, AT, BM, SS, SK, UV, and CT is omitted.

In verification 1, the reference biological information was changed and a missing data calculation model was generated by multiple regression analysis using the forced entry method, and the correlation coefficient (=R) of each missing data calculation model was compared. As a result, the missing data calculation model generated using the three reference biological information SK, AT, and BM had the highest correlation coefficient. The missing data calculation model generated using the measurement data of the three reference biological information SK, AT, and BM is shown in the following formula (1). In addition, the summary, coefficient, and results of variance analysis of the missing data calculation model shown in formula (1) are shown in Table 1.
Deficit HR(t) = 0.72*AT(t) + 0.804*BM(t) - 0.819*SK(t) + 86.811 ... (1)

In the above formula (1), t=6 is substituted to calculate the complementary data, and the calculated complementary data is then complementary to the missing time in FIG. 3(b), generating the time-series measurement data shown in FIG. 3(c). The measurement data at t=6 shown in FIG. 3(c) is close to the measurement data at t=6 shown in FIG. 3(a).

(Verification 2: During sleep, missing data = 2)
In verification 2, measurement data during sleep was acquired in the same way as in verification 1. Measurement data was acquired for a total of 12 minutes (t1 to t12) at 1-minute intervals. In addition, two pieces of HR data were missing, in this case HR data at t=6 and 7, with two pieces of missing data. FIG. 4 is a diagram showing HR data in verification 2 (during sleep). FIG. 4(a) is a time series display of the acquired HR data, FIG. 4(b) is a time series display with missing times t=6 and 7, and FIG. 4(c) is a time series display with the missing data complemented.

In verification 2, a missing data calculation model was generated by multiple regression analysis using the forced entry method as in verification 1, and the correlation coefficients were compared. As a result, the missing data calculation model generated using the three reference biological information items SK, AT, and BM had the highest correlation coefficient. The missing data calculation model generated using the three items SK, AT, and BM is shown in (2) below. In addition, the summary, coefficients, and results of analysis of variance of the missing data calculation model shown in (2) are shown in Table 2.
Deficit HR(t) = -1.691*AT(t) - 0.529*BM(t) - 1.164*SK(t) + 103.971 ... (2)

In the above formula (2), t=6, 7 is substituted to calculate the complementary data, and the calculated complementary data is complementary to the missing time in FIG. 4(b), generating the time series measurement data shown in FIG. 4(c). The t=6, 7 measurement data shown in FIG. 4(c) is close to the t=6, 7 measurement data shown in FIG. 4(a).

(Verification 3: During sleep, missing data = 3)
In verification 3, measurement data during sleep was acquired in the same way as in verification 1. Measurement data was acquired for a total of 13 minutes (t1 to t13) at 1-minute intervals. In addition, three pieces of HR data were missing, in this case HR data at t=6, 7, and 8, with three pieces of missing data. FIG. 5 is a diagram showing HR data in verification 3 (during sleep). FIG. 5(a) is a time series display of the acquired HR data, FIG. 5(b) is a time series display with missing times=t6, 7, and 8, and FIG. 5(c) is a time series display with the missing data complemented.

In verification 3, a missing data calculation model was generated by multiple regression analysis using the forced entry method as in verification 1, and the correlation coefficients were compared. As a result, the missing data calculation model generated using two pieces of reference biological information, SK and BM, had the highest correlation coefficient. The missing data calculation model generated using two pieces of SK and BM is shown in the following formula (3). In addition, the summary, coefficients, and results of variance analysis of the missing data calculation model shown in formula (3) are shown in Table 3.
Deficit HR (t) = -2.863*BM(t)+38.128*SK(t)-1256.157…(3)

In the above formula (3), t = 6, 7, 8 is substituted to calculate the complementary data, and the calculated complementary data is complementary to the missing time in Figure 5 (b), generating the time series measurement data shown in Figure 5 (c). The measurement data at t = 6, 7, 8 shown in Figure 5 (c) is close to the measurement data at t = 5, 6, 7 shown in Figure 5 (a).

(Verification 4: During exercise (yoga), missing data = 1)
In verification 4, measurement data was acquired while the user was exercising (yoga). Measurement data was acquired for a total of 11 minutes (t1 to t11) at 1-minute intervals. In addition, one piece of HR data was missing, in this case the HR data at t=6, with the missing data being set as 1. FIG. 6 is a diagram showing HR data in verification 4 (during exercise (yoga)). FIG. 6(a) is a time series display of the acquired HR data, (b) is a time series display with the missing time t=6, and (c) is a time series display with the missing data complemented.

In verification 4, a missing data calculation model was generated by multiple regression analysis using the forced entry method as in verification 1, and the correlation coefficients were compared. As a result, the missing data calculation model generated using two pieces of reference biological information, SK and EE, had the highest correlation coefficient. The missing data calculation model generated using two pieces of SK and EE is shown in the following formula (4). In addition, the summary, coefficients, and results of variance analysis of the missing data calculation model shown in formula (4) are shown in Table 4.
Deficit HR(t) = -4.274*EE(t) + 5.535*SK(t) - 98.082 ... (4)

In the above formula (4), t=6 is substituted to calculate the complementary data, and the calculated complementary data is then complementary to the missing time in FIG. 6(b), generating the time-series measurement data shown in FIG. 6(c). The t=6 measurement data shown in FIG. 6(c) is close to the t=6 measurement data shown in FIG. 6(a).

(Verification 5: walking, missing data = 1)
In verification 5, measurement data was acquired when the user was active while walking. Measurement data was acquired for a total of 11 minutes (t1 to t11) at 1-minute intervals. In addition, one piece of HR data was missing, in this case the HR data at t=6, with the missing data being set as 1. FIG. 7 (walking) is a diagram showing HR data in verification 5. FIG. 7(a) is a time series display of the acquired HR data, (b) is a time series display with the missing time t=6, and (c) is a time series display with the missing data complemented.

In verification 5, a missing data calculation model was generated by multiple regression analysis using the forced entry method as in verification 1, and the correlation coefficients were compared. As a result, the missing data calculation model generated using the four reference biological information items SK, AT, STP, and EE had the highest correlation coefficient. The missing data calculation model generated using the four items SK, AT, STP, and EE is shown in the following formula (5). In addition, the summary, coefficients, and results of analysis of variance of the missing data calculation model shown in formula (5) are shown in Table 5.
Deficit HR (t) = -0.228*STP(t)+11.187*EE(t)+21.273*AT(t)-30.333*SK(t)+945.114…(5)

In the above formula (5), t=6 is substituted to calculate the complementary data, and the calculated complementary data is complementary to the missing time in FIG. 7(b), generating the time-series measurement data shown in FIG. 7(c). The t=6 measurement data shown in FIG. 7(c) is close to the t=6 measurement data shown in FIG. 7(a).

In the above embodiment, the biosensor 1 generates a missing data calculation model based on the measurement data, but the missing data calculation model may be generated including the actual measurement data. Also, in the above embodiment, the providing unit 35 generates the provided data including the measurement data 60, but the providing unit 35 may generate the provided data including the measurement data 50 such as the PPI data 51 and the ACT data 52 in addition to the measurement data 60. The user may be able to perform input operations on the input unit 11 by including the provision of the measurement data in the request information.

In the above embodiment, the missing data generating device is provided as the control unit 30 in the biosensor 1, but the missing data generating device may be provided in an external device outside the biosensor 1. In this case, the missing data generating device may be provided in a server or the like as an external device that can be connected to the biosensor 1 by wire or wirelessly via a network.

In addition, in the above embodiment, measurement data is generated at one-minute intervals, but measurement data may be generated at intervals less than one minute, for example, at 30-second intervals, or may be generated at intervals greater than one minute, for example, at five-minute intervals.

The above describes an embodiment of the present invention with reference to the drawings. However, the missing data completion device, biosensor, missing data completion method, and program are not limited to the above-described embodiment, and various modifications, substitutions, combinations, and/or design changes may be made without departing from the spirit of the present invention.

The effects of the above-described embodiments of the present invention are merely examples. Therefore, in addition to the effects described above, the embodiments of the present invention may also have other effects that a person skilled in the art may recognize from the description of the above-described embodiments.

1 Biosensor, 10... Input/Output Unit, 11... Input Unit, 12... Output Unit, 20... Actual Measurement Sensor, 21... Light Sensor, 22... Acceleration Sensor, 23... UV Sensor, 24... Temperature Sensor, 25... Microphone, 30... Control Unit, 31... Acquisition Unit, 32... Data Management Unit, 33... Estimation Unit, 34... Complement Unit, 35... Provision Unit, 40... Memory, 50... Measurement Data, 51... PPI Data, 52... ACT Data, 60... Measurement Data, 61... HR Data, 62... STP Data, 63... EE Data, 64... AT Data, 65... BM Data, 66... SS Data, 67... SK Data, 68... UV Data, 69... CT Data

Claims (7)

an acquisition unit that acquires time-series measurement data relating to a plurality of pieces of biological information;
and an estimation unit that estimates the missing data by a simulation using target measurement data, which is measurement data of target biological information, which is biological information including missing data in the measurement data, and reference measurement data, which is measurement data of reference biological information, which is biological information other than the target biological information.
A device for filling in missing data. The target biological information is an average heart rate,
The reference biological information includes at least one of the number of steps, energy consumption, activity type, body movement type, sleep state, skin temperature, ultraviolet light level, and conversation time.
2. The missing data repair device according to claim 1. the estimation unit executes the simulation by utilizing a multiple regression analysis using the target measurement data and the reference measurement data.
3. The missing data repair device according to claim 2. The estimation unit performs the multiple regression analysis by a forced input method.
4. The missing data repair device according to claim 3. A measurement sensor that measures a biological change related to a biological change;
an acquisition unit that acquires time-series measurement data relating to a plurality of pieces of biological information derived based on the biological changes;
an estimation unit that generates time series data of target biological information, which is biological information including missing data in the measurement data, and reference biological information, which is biological information other than the target biological information, and estimates the missing data by a simulation using the time series data;
a completion unit that completes the missing data estimated by the estimation unit;
a providing unit that generates and provides provision data based on the measurement data from which the missing data has been complemented;
The target biological information is an average heart rate,
The reference biological information includes at least one of the number of steps, energy consumption, activity type, body movement type, sleep state, skin temperature, ultraviolet light level, and conversation time.
Biometric sensors. The computer
Acquire time-series measurement data relating to multiple pieces of biological information;
generating time series data of target biological information, which is biological information including missing data in the measurement data, and reference biological information, which is biological information other than the target biological information, and estimating the missing data by a simulation using the time series data;
Missing data imputation methods. On the computer,
Acquire time-series measurement data relating to multiple pieces of biological information;
generating time series data of target biological information, which is biological information including missing data in the measurement data, and reference biological information, which is biological information other than the target biological information, and estimating the missing data by a simulation using the time series data;
program.

Images