JP2018068465A - Biological information processing device and biological information processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a biological information processing device for correctly grasping a biological state by taking biological information reflecting a state of an automatic nerve into consideration in a combined manner.SOLUTION: An arithmetic processing unit 57 executes: a frequency component ratio data acquisition step for acquiring N kinds of frequency component ratio data from N kinds (N≥3) of biological information of time series; a data aggregation step for aggregating the N kinds of biological information and N kinds of frequency component ratios to acquire M kinds (M≥2 and N>M) of time series data for classification; a classification step for classifying the M kinds of time series data for classification into a plurality of clusters; and an appearance data generation step for generating appearance data of time series in each cluster.SELECTED DRAWING: Figure 42

Description

  The present invention relates to a biological information processing apparatus and a biological information processing method for processing biological information.

Conventionally, as described in Patent Document 1, for example, a feature vector is detected from a plurality of biological signals including low-frequency and high-frequency band components of a heart rate change rate representing the influence of the autonomic nervous system of the living body, A method of discriminating the state of a living body using a shunt based on learning data has been known.

Japanese Patent Laid-Open No. 2004-138

However, in the method described in Patent Document 1, since the state is determined using the shunt, even if it is possible to grasp the state of the living body, the degree of the state cannot be grasped. There was a problem that the state could not be grasped correctly.

SUMMARY An advantage of some aspects of the invention is to solve at least a part of the problems described above, and the invention can be implemented as the following forms or application examples.

Application Example 1 In the biological information processing apparatus according to this application example, the arithmetic processing unit has N types of time series (N
≧ 3) obtaining N types of frequency component ratio data which is a ratio of a low frequency component lower than a predetermined frequency and a high frequency component higher than the predetermined frequency from the biological information data; Data aggregation step of aggregating biometric information data and N types of frequency component ratio data to obtain M types (M ≧ 2 and N> M) of classification time series data;
A classification step for classifying the time-series data for classification into a plurality of clusters, and an appearance data generation step for generating time-series appearance data for each cluster are executed.

According to this application example, information reflecting the state of the autonomic nerve can be acquired by obtaining N types of frequency component ratio data from N types (N ≧ 3) of biological information data in time series. Then, by collecting N types of biological information data and N types of frequency component ratio data, acquiring time series data for classification of M types (M ≧ 2 and N> M) useful for biological state analysis Can do. Then, by classifying the M types of time-series data for classification into a plurality of clusters and generating the appearance data of each cluster, it is possible to grasp the state of the living body and the degree of the state reflecting the state of the autonomic nerve.
Therefore, it is possible to provide a biological information processing apparatus that can correctly grasp the state of the living body reflecting the state of the autonomic nerve.

[Application Example 2] In the data aggregation step described in the application example, the N types of biological information data and the N types of frequency component ratio data are integrated into X types (X ≧ M) of classification time series data. Preferably, the method includes selecting the M types from the X types of time series data for classification.

According to this application example, N types of biological information data and N types of frequency component ratio data are aggregated into X types (X ≧ M) of classification time series data, and by selecting M types,
The number of time series data for classification can be reduced.

[Application Example 3] In the data aggregation step described in the application example, the N types of biological information data and the N types of frequency component ratio data are aggregated by performing principal component analysis or factor analysis, and the classification Preferably, the selection step is based on the variance of the time series data.

According to this application example, principal component analysis or factor analysis is performed on N types of biological information data and N types of frequency component ratio data, and M types of time-series data for classification are acquired based on the variances. M types of time series data for classification useful for state analysis can be acquired.

Application Example 4 It is preferable that the data aggregation step described in the application example includes selecting and aggregating at least one of the N types of frequency component ratio data.

According to this application example, by selecting and aggregating at least one of the N types of frequency component ratio data, it is possible to acquire M types of time series data for classification useful for biological state analysis. .

Application Example 5 The classification step described in the application example includes plotting the M types of classification time-series data in an M-dimensional space and clustering the plots into the plurality of clusters. It is preferable to include.

According to this application example, it is possible to classify biological states by plotting M types of time series data for classification in an M-dimensional space, performing clustering, and classifying into a plurality of clusters.

[Application Example 6] The appearance data generation step described in the application example described above is performed for each cluster.
Preferably, the method includes calculating the appearance probability for each calculation time obtained by averaging the time series data for classification belonging to the cluster with a predetermined time width while shifting the calculation time.

According to this application example, the state of the living body at the time is calculated by averaging the time series data for classification for each cluster with a predetermined time width and calculating the time series appearance probability of the time series data for classification belonging to each cluster. Can be grasped correctly.

Application Example 7 It is preferable that the arithmetic processing unit according to the application example execute a display control step of controlling display of the time series appearance data on a dial having a time axis in the circumferential direction.

According to this application example, it is possible to intuitively grasp a change in the state of a living body with the passage of time by controlling display of time-series appearance data on a dial having a time axis in the circumferential direction.

Application Example 8 The biological information data described in the application example includes biological information of any one of heat flux, skin temperature, deep temperature, oxygen saturation, blood flow, blood pressure, activity, and pulse rate. Is preferred.

According to this application example, the biological information of any one of heat flux, skin temperature, deep temperature, oxygen saturation, blood flow, blood pressure, activity, and pulse rate is aggregated and used for each classified cluster. By generating the appearance data, it is possible to correctly grasp the state of the living body.

[Application Example 9] The biological information processing method according to this application example aggregates time-series N types (N ≧ 3) of biological information data and the N types of frequency component ratio data into M types (M ≧ 2). And N> M
A data aggregation step of acquiring time series data for classification), a classification step of classifying the M types of time series data for classification into a plurality of clusters, and appearance data generation for generating time series appearance data for each cluster And a step.

According to this application example, information reflecting the state of the autonomic nerve can be acquired by obtaining N types of frequency component ratio data from N types (N ≧ 3) of biological information data in time series. Then, by collecting N types of biological information data and N types of frequency component ratio data, acquiring time series data for classification of M types (M ≧ 2 and N> M) useful for biological state analysis Can do. Then, by classifying the M types of time-series data for classification into a plurality of clusters and generating the appearance data of each cluster, it is possible to grasp the state of the living body and the degree of the state reflecting the state of the autonomic nerve.
Therefore, it is possible to provide a living body information processing method that makes it possible to correctly grasp the state of the living body reflecting the state of the autonomic nerve.

The schematic diagram which shows the structural example of the whole system provided with the biological information processing apparatus of this embodiment. The external view which looked at the measuring apparatus from the surface side. The external view which looked at the measuring apparatus from the back side. The top view of a heat flow sensor. The schematic diagram of the AA arrow cross section shown in FIG. The figure which shows an example of each biometric information Sn (t) in the time t in a table format. Activity S ₁ (t) was graphed FIG. Graphing the pulse rate S ₂ (t) the FIG. Graphing oxygen saturation S ₃ (t) in arterial blood was FIG. Skin temperature S ₄ (t) was graphed FIG. Core Temperature S ₅ (t) was graphed FIG. Heat flux S ₆ (t) was graphed FIG. The figure which shows the low frequency area | region of a biometric information, and a high frequency area | region. Figure activity amount of the frequency component ratio R ₁ (t) and graphed. Graphed FIG frequency component ratio R ₂ (t) of the pulse rate. Graphed FIG frequency component ratio R ₃ (t) in arterial blood oxygen saturation. Graphed FIG frequency component ratio R ₄ (t) of the skin temperature. Graphed FIG frequency component ratio R ₅ a (t) of the core temperature. Graphed FIG frequency component ratio R ₆ (t) of the heat flux. Plotted FIG frequency component ratio R ₆ a (t) of the frequency component ratio R ₂ (t) and the heat flux of the pulse rate. The first principal component Y ₁ (t) was graphed Figure in this embodiment. The second principal component Y ₂ (t) was graphed Figure in this embodiment. The first principal component Y ₁ (t) and the second principal component Y ₂ (t) a prescribed coefficient vector value alpha _1n the present embodiment, shows an alpha _2n. It plotted aggregated data of the first principal component Y ₁ in the embodiment (t) and the second principal component Y ₂ (t). The figure which shows the clustering result of the aggregated data of FIG. The figure which shows the time series cluster data c (t). The figure which shows cluster attribution data B _C (t). Figure graphed appearance data P ₁ (t) of the cluster G _1. Figure graphed appearance data P ₂ (t) of the cluster G _2. Figure graphed appearance data P ₃ (t) of the cluster G _3. Figure graphed appearance data P ₄ (t) of the cluster G _4. Appearance data P ₅ (t) was graphed Figure cluster G _5. Appearance data P ₆ (t) was graphed Figure cluster G _6. Figure graphed appearance data P ₇ (t) of the cluster G _7. Figure graphed appearance data P ₈ (t) of the cluster G _8. Figure graphed appearance data P ₉ (t) of the cluster G _9. Appearance data P ₄ (t) and the appearance data P ₈ (t) and the appearance data P ₉ (t) and graphed figure superimposed. Figure appearance data P ₃ (t) and graphed. Appearance data P ₂ (t) appearing data and P ₅ (t) and graphed figure superimposed. Appearance data P ₁ (t) and the appearance data P ₆ (t) and the appearance data P ₇ (t) and graphed figure superimposed. The figure which shows the example of a display of the state analysis result in a measuring apparatus. The block diagram which shows the main function structural examples of a biological information processing apparatus. The flowchart which shows the flow of the process which a biometric information processing apparatus performs.

Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings. It should be noted that the present invention is not limited to the embodiments described below, and modes to which the present invention can be applied are not limited to the following embodiments. In the description of the drawings, the same parts are denoted by the same reference numerals.

(Embodiment)
FIG. 1 is a schematic diagram illustrating a configuration example of the entire system including the biological information processing apparatus 5 of the present embodiment. The biological information processing apparatus 5 according to the present embodiment includes the measurement apparatus 100 via a predetermined communication line N.
The user's biological information measured by the measuring apparatus 100 is acquired and used for analysis. The biological information processing apparatus 5 can be realized using electronic devices such as a smartphone, a personal computer, a tablet computer, a mobile phone, and a game device.

2 and 3 are external views showing a configuration example of the measuring apparatus 100. FIG. 2 shows the front surface (surface facing outward when worn on the user), and FIG. 3 shows the back surface (when worn on the user). The surface in contact with the user's skin). The measuring apparatus 100 is configured as, for example, a wristwatch-type electronic device that displays measured biological information together with the current time, and wraps a band 3 provided in the main body case 10 around the wrist of the user to thereby increase the surface of the living body (the skin surface of the wrist). ).

The configuration is not limited to the configuration in which the band 3 is wound around the skin surface, but may be a configuration in which an adhesive sheet or gel that is detachable from the skin surface is used and attached to the user's skin surface. Further, the measurement site to which the measurement device 100 is attached is not limited to the wrist. The measurement site is, for example, the forehead, the neck,
The upper arm, the ankle, the chest, the waist, the back of the limb, etc. may be selected as appropriate.

The measuring apparatus 100 includes a dial 11 for displaying the current time, biometric information, and the like inside the main body case 10. Further, operation switches 121 and 123 for inputting various operations such as measurement start of biological information and stop of the measurement, crown 13 for manually adjusting the current time, and the like are arranged on the outer periphery of the main body case 10. It is installed.

Above the dial 11, a pointer 111 for displaying the current time is arranged. In FIG. 2, the pointer 111 is illustrated as a two-needle type including an hour hand 112 and a minute hand 113, but a three-needle type with a second hand may be used. The pointer 111 is driven and moved by a movement (not shown) provided on the back side of the dial 11. Further, the dial 11 has a display device such as a liquid crystal panel arranged on the front surface, for example, blood pressure, arterial blood oxygen saturation (SpO) measured as biological information.
2) and the pulse rate (heart rate) are individually displayed. Note that the display of the current time on the dial 11 is not limited to analog display but may be digital display. The type of biometric information to be displayed can be selected as appropriate, and can be configured to change according to user operation input.

This measuring apparatus 100 has a plurality of sensors arranged in place. For example, measuring device 1
00 includes a heat flow sensor 21, an optical sensor 22, a motion sensor 23, and a GPS sensor 24. When a predetermined measurement start operation is performed, the measurement apparatus 100 continuously measures a plurality of types of biological information in parallel by these sensors 21 to 24, and after appropriate timing such as after the measurement end operation is performed. Then, the measurement result is transmitted to the biological information processing apparatus 5. The measurement of each biological information is performed at the same time in a predetermined sampling cycle (for example, 1 [sec]).

The heat flow sensor 21 is based on the temperature difference generated inside the heat flow sensor 21 due to heat transfer between the living body surface of the measurement site (the wrist in this embodiment) to which the measurement apparatus 100 is attached and the external environment. Measure the heat flow generated on the surface. For example, the heat flow sensor 21 has a substantially annular shape, and is disposed so that the protective layer 213 a that forms one end surface of the heat flow sensor 21 is exposed on the back surface of the main body case 10.

FIG. 4 is a plan view of the heat flow sensor 21, and FIG. 5 is a schematic diagram of a cross section taken along the line AA shown in FIG. Note that illustration of wiring and the like is omitted. 4 and 5, the heat flow sensor 21 includes a heat transfer layer 211, a protective layer 213a covering the lower side in FIG. 5 and a protective layer 213b covering the upper side through the heat diffusion layer 215. It has a bonded layer structure, and a plurality (four in FIG. 4) of detectors 217 are incorporated inside the heat transfer layer 211.

The detector 217 includes a temperature measuring body 219a arranged so as to be in contact with the protective layer 213a on the living body surface side when worn, and an external environment side (the surface side of the measuring apparatus 100) at a position facing the temperature measuring body 219a. A temperature measuring body 219b arranged so as to be in contact with the protective layer 213b, and the temperature detected by the temperature measuring body 219a is set as the skin temperature of the wrist, and the temperature detected by the temperature measuring body 219b is output as the heat transfer temperature. The heat flux (heat flow per unit area) at the position of the corresponding detector 217 can be measured from the temperature difference (up and down temperature difference) of the detected temperatures by the temperature measuring elements 219a and 219b. Further, the wrist deep temperature can be measured from the wrist temperature and the heat flux by the relational expression based on the heat conduction equation. A thermistor, a thermocouple, or the like can be used for the temperature measuring elements 219a and 219b. The configuration of the heat flow sensor 21 is not limited to a configuration using two temperature measuring elements, and a known configuration such as a thermopile can be appropriately selected and used.

The optical sensor 22 includes two light emitting portions 221 and 223 disposed in the annular inner portion of the heat flow sensor 21 so that the light emitting surface is exposed on the back surface of the main body case 10, and the main body case 1.
And a light receiving portion 225 disposed on the back surface of 0 so that the light receiving surface is exposed. The light emitting surfaces of the light emitting units 221 and 223 and the light receiving surface of the light receiving unit 225 are protected by a transparent cover glass or the like that covers the annular inner portion of the heat flow sensor 21.

The light emitting units 221 and 223 are LEDs or OLEDs (Orga) that emit irradiation light within a predetermined wavelength range.
nic Light Emitting Diode) and a light source such as a semiconductor laser. The wavelength range of irradiation light can be appropriately selected according to the measurement target. In the present embodiment, for example, one light emitting unit 221 emits visible light having a first wavelength in the vicinity of the wavelength region 660 [nm], and the other light emitting unit 2 is irradiated.
23 irradiates near-infrared light of the second wavelength belonging to the wavelength range 880 [nm] to 940 [nm].

The light receiving unit 225 receives transmitted light and reflected light of the irradiation light, and outputs a signal corresponding to the amount of received light.
For example, a photodiode, a CCD (Charge Coupled Device), a CMOS (Complementa
ry Metal Oxide Semiconductor).

By irradiating irradiation light from one or both of the light emitting units 221 and 223 in the optical sensor 22, and calculating the light reception result of the light receiving unit 225 (output value of the light receiving unit 225) using a known technique, Measuring biological information such as photoelectric pulse wave, volume pulse wave, pulse rate, pulse interval, blood flow rate, blood flow volume, blood perfusion rate, vascular resistance, blood pressure (diastolic blood pressure / systolic blood pressure), arterial oxygen saturation it can. The oxygen saturation level in the arterial blood is obtained by applying the irradiation light of the first wavelength and the second wavelength to the light emitting units 221,
Based on the output value of the light receiving unit 225 obtained by sequentially irradiating from 223, it can be obtained using the absorbances of oxyhemoglobin and deoxyhemoglobin at each wavelength.

The heat flow sensor 21 and the light sensor 22 protrude from the back surface of the main body case 10 so that the cover glass covering the protective layer 213a and the light sensor 22 of the heat flow sensor 21 is easily contacted with the living body surface when the measuring device 100 is mounted. It is good to arrange. By increasing the contactability, it is possible to prevent the measurement accuracy from being lowered.

The motion sensor 23 is for measuring a user's movement, and can be realized by, for example, a nine-axis sensor that detects acceleration (three axes), angular velocity (three axes), and geomagnetism (three axes). By calculating the acceleration, angular velocity, and geomagnetism output values of the motion sensor 23 using known techniques, the user's physical activity, the number of steps, the moving distance, the speed, for example, “standing”, “sitting”, “ Information such as the posture such as “prone position” and the type of exercise (behavior) such as “walking”, “running”, and “step up / down” can be measured.

The GPS sensor 24 is a sensor that receives a GPS satellite signal transmitted from a GPS satellite, which is a kind of positioning satellite, and detects the position of the user using the received GPS satellite signal. In this embodiment, the user's residential area is set from the detection result of the GPS sensor 24. Note that a method for detecting the user's position and the like using GPS is publicly known, and a detailed description thereof will be omitted.

Note that the measuring apparatus 100 may further include a sensor for measuring environmental information separately from the above. For example, a temperature sensor, a humidity sensor, or an atmospheric pressure sensor may be provided. A temperature sensor is realizable using a thermistor, a thermocouple, a platinum thermometer, etc., for example. As the humidity sensor, for example, a polymer resistance type, a polymer capacitance type, an aluminum oxide capacitance type, or the like can be appropriately selected and used. As the atmospheric pressure sensor, a MEMS capacitance type, a piezoresistive type, or the like can be appropriately selected and used.

[principle]
In the embodiment, as the N types (N ≧ 3) of biological information, the activity amount S ₁ (t) [A. U. ], Pulse rate S ₂ (t) [bpm], arterial oxygen saturation S
₃ (t) [%], skin temperature S ₄ (t) [° C.], deep temperature S ₅ (t) [° C.], and heat flux S ₆
(T) [W / m ² ], six types of biological information S _n (t) (n = 1 to 6) are exemplified. In FIG.
An example of each biological information S _n (t) at time t (t ₁ , t ₂ ,...) Is shown in a table format. 7 shows the amount of activity S ₁ (t), FIG. 8 shows the pulse rate S ₂ (t), and FIG. 9 shows the arterial oxygen saturation S ₃ (t
10 is skin temperature S ₄ (t), FIG. 11 is deep temperature S ₅ (t), and FIG. 12 is heat flux S ₆ (t).
) Are graphed with the horizontal axis as time t.

In recent years, attempts to measure biological information such as body temperature, blood pressure, and amount of physical activity and to utilize it for health management and the like are actively made. It is known that biological information varies depending on the state of the living body, such as meals, exercise, breathing, sleep, and autonomic nerve (sympathetic or parasympathetic). For example, during exercise, the pulse rate and heat flux generally increase. In addition, it is known that the frequency component of the pulse rate variation is changed by the action of the autonomic nerve. It is also said that when the parasympathetic nerve works, the pulse rate decreases and the wrist temperature rises. On the other hand, it is known that there is a mutual relationship between different biological information. For example, when the pulse rate rises, the blood flow sent to the limbs rises, so the heat flux and wrist temperature rise. Therefore, it is considered that the state of the living body can be analyzed by a technique such as threshold determination of biological information having a known correlation.

However, the relationship between the biological information and the state of the biological body is not completely formulated, and the relationship is only pointed out as an aggregation of various experimental facts. In addition, as for the mutual relationship between the biological information, there are many cases where the principle relationship is not clear, and it is difficult to formulate by clarifying the mutual relationship between all the biological information. For this reason, when analyzing the state of a certain biological body from specific biological information, the state of another biological body or fluctuations in biological information may become noise, and the analysis target state may not be correctly grasped. In particular, the function of the autonomic nerve causes fluctuations in various biological information, making it difficult to grasp the analysis results.

In general, the frequency component ratio of pulse fluctuation is known as an index indicating the function of the autonomic nerve.
In pulse fluctuation, an LF (Low Frequency) component (frequency band component of 0.05 to 0.15 [Hz]) is a band component in which an influence of blood pressure fluctuation appears, and an HF (High Frequency) component (0.15 to 0.15). The frequency band component of 0.4 [Hz] is said to be a band component that is affected by respiratory fluctuations. Since the LF component and the HF component of the pulse fluctuation fluctuate due to the effects of blood pressure fluctuation and respiratory fluctuation caused by the action of the autonomic nerve, the intensity ratio of the LF component and the HF component: LF / HF changes depending on the action of the autonomic nerve. It is known to do. Specifically, when the parasympathetic nerve is dominant among the autonomic nerves, the value of LF / HF is small, and when the sympathetic nerve is dominant, the value of LF / HF is large. In addition, since biological information other than the pulse is also affected by blood pressure fluctuation and respiratory fluctuation due to the action of the autonomic nerve, it is considered that the values of LF / HF in other biological information are also related to the autonomic nerve.

Therefore, in the biological information processing apparatus 5 according to the present embodiment, (1) N types (6 in the present embodiment).
N types of frequency component ratio data R _n (t) are acquired from each of the (type) biological information data S _n (t), and (2) the biological information data S _n (t) and the acquired frequency component ratio data R _n (t). t) is aggregated to obtain M types (M ≧ 2 and N> M) of classification time series data, (3) the obtained classification time series data is classified into clusters, and (4) each classified cluster Time-series appearance data is generated and biological state analysis is performed. The living body state analysis result is displayed on (5) the biological information processing apparatus 5, for example, and presented to the user.

(1) the frequency component ratio calculation processing as determining a frequency component ratio calculation processing frequency component ratio data, first, the predetermined sampling period (e.g., 1 [sec]) 6 kinds of biological information obtained by S _n For (t), fast Fourier transform is performed at a predetermined number of sampling points (for example, 60 [points]) to obtain a power spectrum P _n (Hz) of the biological information S _n (t). Here, the frequency component ratio is a ratio between a low frequency component lower than a predetermined frequency and a high frequency component higher than the predetermined frequency. As a specific example, the power spectrum P ₂ (Hz) of the pulse rate S ₂ (t) is illustrated in FIG. In FIG. 13, LF, which is a low frequency component lower than a predetermined frequency F ₁ [Hz], indicates the intensity of the power spectrum P ₂ (Hz) in the F ₂ [Hz] to F ₁ [Hz] band, and the predetermined frequency. F ₁ is a high high-frequency component from the [Hz] HF shows intensity of _{F 1 [Hz] ~F 3 [} Hz] band of the power spectrum P ₂ (Hz). This LF and HF
, The frequency component ratio data R _n (t) is obtained by obtaining the value of LF / HF for each predetermined number of sampling points. At this time, for the frequency of the band for obtaining the intensity ratio, for example, F ₂ = 0.05 [Hz], F ₁ [Hz] = 0.15 [Hz], F ₃ [Hz] = 0.4 [Hz]
].

14 to 19 show frequency component ratio data R _n (6) obtained from six types of biological information S _n (t).
t), and is a graph showing the horizontal axis as time t.

Here, in FIG. 14 to FIG. 19, the value after 21:00 which is the bedtime of any frequency component ratio data R _n (t) is lower than the value before 21:00. This is presumably because the parasympathetic nerve dominates during sleep and the frequency component ratio decreases as a result. From this, it is considered that the frequency component ratio data R _n (t) is information reflecting the state of the autonomic nerve.

In addition, FIG. 20 shows the frequency component ratio data R ₂ (pulse rate frequency component ratio) at each time t (t ₁ , t ₂ ,...).
The t) on the horizontal axis, is a plot frequency component ratio data R ₆ (t) to the longitudinal axis of the heat flux.
As shown in FIG. 20, the frequency component ratio data R ₂ of the pulse rate and the frequency component ratio data R _{6 of the} heat flow rate.
Since (t) has a positive correlation, it can be understood that the obtained frequency component ratio data is information reflecting the state of the autonomic nerve even with biological information other than the pulse rate.

As described above, according to the frequency component ratio calculation process, information that reflects the state of the autonomic nerve by obtaining N types of frequency component ratio data from time-series N types (N ≧ 3) of biological information data. Can be obtained.

(2) Data Aggregation Process In the data aggregation process as the data aggregation step, first, the six types of biological information _Sn (
t) and frequency component ratio data R _n (t) obtained by frequency component ratio calculation processing are normalized and used to perform principal component analysis of biological information S _n (t) and frequency component ratio data R _n (t). . The principal component analysis can be performed using a known method, and each biological information after normalization is obtained for each of X types (X ≧ M) of principal components Y ₁ (t) to Y _X (t) orthogonal to each other. A coefficient vector value α _Xn for S _n (t) and frequency component ratio data R _n (t) is obtained. When the principal component analysis is performed, M pieces are selected from the X kinds of principal components Y ₁ (t) to Y _X (t), and are used as time series data for M kinds of classification. In the following, it is assumed that M = 2, and among the first principal component Y ₁ (t) to the Xth principal component Y _X (t), the first principal component having the largest variance l ₁ (t) to l _X (t). Y ₁ (t) and the second principal component Y ₂ (t) in the next order
To obtain two types of time-series data for classification. By the processing here, the original 12-dimensional information (activity amount, pulse rate, arterial oxygen saturation, skin temperature, deep temperature, heat flux, and frequency component ratio of biological information) is converted into 2-dimensional information ( The first principal component and the second principal component) are dimensionally compressed (aggregated).

Note that normalization performed prior to principal component analysis is not essential depending on the type of biological information. For example, it is not necessary to normalize biometric information having similar numerical values and fluctuation ranges such as blood pressure and pulse rate. As a normalization method, a known method can be appropriately selected and used, and a separate method may be applied according to the type of biological information.

21 and 22 show the biological information S _n (t) shown in FIGS. 7 to 12 and FIGS.
FIG. 21 is a diagram showing the principal component analysis results of the frequency component ratio data R _n (t) shown in FIG. 9, FIG. 21 shows the first principal component Y ₁ (t), and FIG. 22 shows the second principal component Y ₂ (t). , The horizontal axis is shown as a graph with time t. Further, in FIG. 23, each piece of biological information S _n (t that defines the first principal component Y ₁ (t).
) And the coefficient vector value α _1n related to each frequency component ratio data R _n (t) and the second principal component Y ₂ (
Each biological information S _n (t) defining t) and coefficient vector value α _2n related to each frequency component ratio data R _n (t) are shown.

Here, the first principal component Y ₁ (t) is obtained by multiplying each normalized biological information S _n (t) and each normalized frequency component ratio data R _n (t) by a coefficient vector value α _1n . The second principal component Y ₂ (t) is a linear sum, and each normalized biological information S _n (t) and each normalized frequency component ratio data R
_n is a linear sum obtained by multiplying (t) by a coefficient vector value α _2n . That is, the coefficient vector value α _1n is
Corresponding biological information S _n (t) and first principal component Y ₁ (t of each frequency component ratio data R _n (t)
) Represents the degree of contribution (influence), and the coefficient vector value α _2n represents the corresponding biological information S
_n (t) and the frequency component ratio data R _n (t) in the second principal component Y ₂ (t) are represented. Therefore, from the values of the coefficient vector values α _1n and α _2n , the principal components Y ₁ (t) and Y ₂ (t
) Can be analogized.

First, in the first principal component Y ₁ (t), as shown in FIG. 23, the coefficient vector value α ₁₁ for the activity amount, the coefficient vector value α ₁₂ for the pulse rate, and the coefficient vector value α ₁₆ for the heat flux are large. A positive value of the degree. The coefficient vector value α ₁₃ for the arterial oxygen saturation is a negative value. Also, the coefficient vector value α _{14 with} respect to the skin temperature is relatively large and a negative value. Further, the coefficient vector values α _{17 to} α _1C with respect to the frequency component ratio of the biological information are all positive values. In general, the amount of activity, pulse rate, and heat flux have a positive correlation with exercise intensity, and thus are considered to have a positive correlation with exercise. On the other hand, if the metabolic rate increases due to exercise or the like, the amount of oxygen supplied from the arterial blood increases, so the arterial oxygen saturation is considered to decrease. In addition, when the sympathetic nerves become active due to exercise, the blood flow of the main arteries increases to maintain muscle activity, and the skin temperature decreases by decreasing the blood flow of the skin, which is the peripheral site. There seems to be a correlation. Further, it is considered that the frequency component ratio showing a positive response to the function of the sympathetic nerve has a positive correlation. Therefore, the positive direction of the first principal component Y ₁ (t) can be interpreted as a scale representing the degree of autonomic nervous function accompanied by movement.

Next, for the second principal component Y ₂ (t), the coefficient vector value α ₂₄ related to the skin temperature and the coefficient vector value α ₂₅ related to the deep temperature are large. In general, it is known that the temperature of the terminal parts such as hands and feet fluctuates due to the action of autonomic nerves. The second principal component is a change in biological information caused by the state of the living body excluding the first principal component. Therefore, the second principal component Y ₂ (t) is a function of autonomic nerves that do not involve exercise, such as changes in outside air temperature, meals, and mental activities (a state in which the sympathetic nerves are dominant and the parasympathetic nerves are dominant) It can be interpreted as a measure of

As described above, according to the data aggregation processing, for example, principal component analysis is performed on N types of biological information data and frequency component ratio data of N types of biological information to extract principal components, which is useful for biological state analysis. M types of time series data for classification (in this case, autonomic nerve function level with movement and autonomic nerve function level without movement) can be collected.

(3) Classification process In the classification process as the classification step, first, M types of time-series data for classification acquired by the data aggregation process are plotted in the M-dimensional space. In the present embodiment, the first principal component Y ₁ (t) and the second principal component Y ₂ (t) acquired as two types of time-series data for classification are represented by time t (t ₁ , t ₂ ,.
..) Are plotted in a two-dimensional space as aggregated data of each biological information S _n (t) measured at the corresponding time t. FIG. 24 is a diagram in which each aggregated data D1 is plotted with the horizontal axis representing the first principal component Y ₁ (t) and the vertical axis representing the second principal component Y ₂ (t). As described above, the first principal component Y ₁ (t) represents the autonomic nervous function level with movement, and the second main component Y ₂ (t) represents the autonomic nervous function level without movement. Therefore, from the distribution of each aggregated data D1, for example, it is possible to read the autonomic nerve function with exercise, the balance of the autonomic nerve without exercise, and the like.

Once the aggregated data is plotted, clustering is performed using a non-hierarchical clustering technique, and each aggregated data is classified into s clusters. As the non-hierarchical clustering method, for example, the k-means method, the k-means ++ method, the Ward method, and the like are known, and a known method can be appropriately selected and used. FIG. 25 is a diagram showing a clustering result, and shows nine clusters G _{1 to} G ₉ obtained by classifying each aggregated data D1 of FIG. 24 by the k-means ++ method with s = 9. By this processing, each aggregated data D1 is 2
For example, the data is grouped into nine groups based on the distance between the aggregated data D1 in the dimensional space.

As described above, according to the classification process, the biological state can be classified by plotting M types of time-series data for classification in an M-dimensional space, performing clustering, and classifying into a plurality of clusters.

(4) Appearance data generation process In the appearance data generation process as the appearance data generation step, first, time-series cluster data c (t) is generated based on the cluster classification result. FIG. 26 is a diagram showing time-series cluster data c (t). As shown in FIG. 26, the time-series cluster data c (t) is data in which the identification numbers c of the clusters to which each aggregated data D1 belongs are arranged in the order of their times t (t ₁ , t ₂ ,...). Is a column.

Subsequently, based on the time-series cluster data c (t), the time-series cluster membership data B _C that determines whether the cluster to which each aggregated data D1 belongs is its own cluster or not for each of the clusters G _{1 to} G _9.
(T) (c = 1, 2,..., S) is generated. FIG. 27 is a diagram showing the cluster membership data B _C (t) of each of the clusters G _{1 to} G ₉ , and also shows the time series cluster data c (t) of FIG. As shown in FIG. 27, the cluster membership data B _C (t) is data in which the value of time t related to the aggregated data belonging to the corresponding cluster is “1”, and the other values of time t are “0”. Is a column. Accordingly, in the cluster membership data B _C (t), when attention is paid to a specific time t (when the table of FIG. 27 is viewed horizontally), “1” is set to any one of them. For example, when paying attention to the time t1, since the time series cluster data c (t), a cluster to which the aggregated data belongs is cluster G ₁ identification number c = 1. Therefore, B ₁ (t ₁ ) is “1” and other B ₂ (
t ₁ ) to B ₉ (t ₁ ) are “0”.

Here, assuming that each of the clusters G _{1 to} G ₉ is related to some biological state, for example, the corresponding time depends on whether the value of B _C (t) is “1” or “0”. The state of the living body at t can be roughly known. However, in reality, the state of the living body changes continuously, and the binary value of “1” or “0” is not a value that accurately reflects the state.

Therefore, if the cluster membership data B _C (t) is generated, each cluster membership data B _C (t) is sequentially processed, and the appearance probability of the aggregated data is calculated for each calculation time to calculate each cluster G ₁ .
Appearance data P _C (t) of G ₉ is generated. Specifically, for example, each time t is set as a calculation time. Then, the cluster attribution data B _{C to be} processed with a predetermined time width based on the calculation time t
The average value of (t) is calculated according to the following equation (1). This is performed while shifting the calculation time t, and the time-series appearance data P _C (t) is obtained using the calculated average value as the appearance probability at the corresponding time t of the aggregated data belonging to the cluster. The predetermined time width may be set as appropriate, for example, 20 [min] before and after the calculation time t (that is, τ = 10 [min]).

Appearance data P ₁ cluster G ₁ (t) in FIG. 28, the appearance data P ₂ clusters G ₂ in FIG. 29 (
30), the appearance data P ₃ (t) of the cluster G _{3 in} FIG. 30, the appearance data P ₄ (t) of the cluster G _{4 in} FIG. 31, the appearance data P ₅ (t) of the cluster G ₅ in FIG. the appearance data P ₆ (t), cluster G ₉ occurrence data P ₈ (t), 36 clusters G ₈ occurrence data P ₇ (t), 35 clusters G ₇ in FIG. 34 of the cluster G ₆ The appearance data P ₉ (t) is shown as a graph with the horizontal axis as time t. According to FIGS. 28 to 36, it can be seen that the aggregated data belonging to each of the clusters G _{1 to} G ₉ appears unevenly in time.

Next, each appearance data P1 (t) to P9 (t) will be examined in detail. First, attention is paid to the appearance data P ₄ (t), the appearance data P ₈ (t), and the appearance data P ₉ (t). FIG. 37 is a graph in which the appearance data P ₄ (t), the appearance data P ₈ (t), and the appearance data P ₉ (t) are overlapped on a common time axis. According to the appearance data P ₄ (t), the appearance data P ₈ (t), and the appearance data P ₉ (t), the aggregated data belonging to the clusters G ₈ and G ₉ both appear in the morning and around noon, Appears in the evening. And in light of the daily user's behavior, the appearance time zone of the aggregated data related to the clusters G ₈ and G ₉ is the time zone when moving for commuting or lunch, that is, the time zone when exercising outdoors. Is consistent with On the other hand, cluster G ₄
The appearance time zone of the aggregated data according to is coincident with the time zone in which the user is acting indoors. From this,
Appearance data P ₈ (t) and appearance data P ₉ (t) are in a state of exercising outdoors (outdoor motion state), and appearance data P ₄ (t) are in a state of exercising indoors (indoor motion state). ). As seen in FIG. 25, the clusters G ₄ , G ₈ and G ₉ all have a large autonomic nervous function accompanied by movement and are appropriate as a movement state.

Next, attention is paid to the appearance data P ₃ (t). FIG. 38 is a graph showing the appearance data P ₃ (t) on the time axis. According to the appearance data P ₃ (t), the aggregated data belonging to the cluster G ₃ is
Appears around noon and appears from evening to night. Then, in light of the daily user behavior, the appearance time zone of the aggregated data related to the cluster G ₃ coincides with the meal time zone. From this, it can be considered that the appearance data P ₃ (t) represents a state of eating (meal state).

Next, attention is paid to the appearance data P ₂ (t) and the appearance data P ₅ (t). FIG. 39 is a graph obtained by superimposing the appearance data P ₂ (t) and the appearance data P ₅ (t). These appearance data P
According to ₂ (t) and P ₅ (t), the aggregated data belonging to the clusters G ₂ and G ₅ both appear in the daytime activity time zone. Then, when the match against the user of the action of the day, appearance time zone of the aggregate data relating to the cluster G ₅ is consistent with the time zone of desk work, etc., appearance time zone of the aggregate data relating to the cluster G ₂ is break time zone, etc. Matches. From this, the appearance data P ₂ (
It is considered that t) represents a relaxed state where the parasympathetic nerve is dominant, and the appearance data P ₅ (t) represents a concentrated state where the sympathetic nerve is dominant (when awakening). Looking at Figure 25, clusters G ₅ is sympathetic autonomic function degree without motion dominates, cluster G ₂ is parasympathetic autonomic function degree without motion is dominant. In addition, the clusters G ₂ and G ₅ have an average degree of autonomic function with exercise and are appropriate as daytime activities.

Next, attention is paid to the appearance data P ₁ (t), the appearance data P ₆ (t), and the appearance data P ₇ (t). FIG. 40 is a graph in which the appearance data P ₁ (t), the appearance data P ₆ (t), and the appearance data P ₇ (t) are overlaid. According to the appearance data P ₆ (t), the aggregated data belonging to the cluster G ₆ appears from night to midnight. Also, the appearance data P ₁ (t) and the appearance data P ₇ (t
), The aggregated data belonging to the cluster G ₁ and the cluster G ₇ appears at midnight or early morning. And in light of the daily user's behavior, the appearance time zone of the aggregated data related to the cluster G ₆ coincides with that immediately after falling asleep, and the appearance time zone of the aggregated data related to the cluster G ₁ and the cluster G ₇ is sleeping. The time zone, in particular, the appearance time zone of the aggregated data relating to the cluster G ₇ is coincident with the deep sleep time zone at midnight. From this, it can be considered that the appearance data P ₆ (t) represents a shallow sleep state, the appearance data P ₁ (t) represents a normal sleep state, and the appearance data P ₇ (t) represents a deep sleep state. As seen in FIG. 25, the clusters G ₁ , G ₆ , and G ₇ all have a small autonomic function with movement and are appropriate as sleep states.

As described above, according to the appearance data generation process, the first principal component Y ₁ (t) representing the autonomic nervous function level with movement and the second main component Y ₂ ( By combining with t), the appearance data P _C (t) of each of the clusters G _{1 to} G ₉ can be associated with the state of the living body. Therefore, from the values of the plurality of pieces of biological information measured at each time t, the corresponding time t
It becomes possible to correctly grasp the state of the living body.

Note that the relationship between the appearance data P _C (t) of each of the clusters G _{1 to} G ₉ generated as described above and the state of the living body can be obtained by performing a known regression analysis or the like. For example, each occurrence data P
_The relative value C (t) representing the state of the living body of C (t) is obtained using another measurement method. Then, regression analysis is performed using the obtained relative value C (t) as an objective variable and appearance data P _C (t) as an explanatory variable. More specifically, a method of performing regression analysis by applying the following equation (2) to the relative value C (t) representing the state of each living body and obtaining the coefficient vector value β by the least square method or the like is given. It is done.

It is also assumed that the relative value C (t) representing the state of the living body is difficult to quantify. In that case, for example, if the state of the living body is an operation such as “walking”, “sitting”, or “standing” of the user, those operations at time t are performed by photographing the user with a video camera and analyzing the video. The presence or absence (existence = 1 / non-existence = 0) is specified. The cluster membership data B _C (t
) To generate the appearance data P _C (t) as a relative value C (t).

As described above, according to the appearance data generation process, the time series data for classification for each cluster is averaged over a predetermined time width, and the time series appearance probability of the time series data for classification belonging to each cluster is calculated. It is possible to correctly grasp the state of the living body at the time.

(5) Display control processing If the living body state analysis is performed as described above, the living body information processing apparatus 5 performs control to display the state analysis result. FIG. 41 is a diagram illustrating a display example of a state analysis result.
For example, as shown in FIG. 41, the display of the state analysis result in the display control process as a display control step executed by the arithmetic processing unit 57 of the biological information processing apparatus 5 The representative value mark MK representing the representative value of each appearance data P _C (t) is displayed at the position of the scale. In this case, the appearance data P _C (t) is sequentially processed, the display time range is divided by unit time such as 1 [min], and the average value of appearance probabilities at each time t is calculated. Not only the average value, but the mode value within the unit time may be calculated. Subsequently, the maximum value among the average values of the appearance data P _C (t) is selected for each unit time, and is set as a representative value in the corresponding unit time. Then, control for displaying the representative value mark MK based on the representative value is performed. Specifically, the display color is set to a display color corresponding to the cluster of the representative value source, and the density is adjusted based on the representative value.

In addition, the display mode of FIG. 41 is an example, and the display mode of the state analysis result is not particularly limited. For example, the value related to the time t of the appearance data P _C (t) may be displayed as a radar chart on the dial 11 having the time axis in the circumferential direction. Appearance data P _C (t
) May be freely set according to user operation input.

As described above, according to the display control process as the display control step, time-series appearance data is displayed and controlled on the dial 11 having the time axis in the circumferential direction, so that the state of the living body with time elapses. Intuitively grasp changes.

[Function configuration]
FIG. 42 is a block diagram illustrating a main functional configuration example of the biological information processing apparatus 5. As shown in FIG. 42, the biological information processing apparatus 5 includes an operation input unit 51, a display unit 53, a communication unit 55,
An arithmetic processing unit 57 and a storage unit 59 are provided.

The operation input unit 51 receives various operation inputs by the user and outputs an operation input signal corresponding to the operation input to the arithmetic processing unit 57. It can be realized by a button switch, lever switch, dial switch, touch panel, or the like.

The display unit 53 includes an LCD (Liquid Crystal Display) and an OELD (Organic Electroluminum).
luminescence display) and a display device such as an electronic paper display, and performs various displays based on a display signal from the arithmetic processing unit 57.

The communication unit 55 is a communication device for transmitting and receiving data to and from the outside (for example, the measurement device 100) under the control of the arithmetic processing unit 57. As a communication method of the communication unit 55, a wireless connection method using wireless communication, a wired connection method using a cable conforming to a predetermined communication standard, an intermediate device that is also used as a charger called a cradle, etc. Various systems such as a connection type via the network can be applied.

The arithmetic processing unit 57 performs input / output control of data with each function unit, a predetermined program and data, an operation input signal from the operation input unit 51, and a biological body acquired from the measurement apparatus 100 via the communication unit 55. Various arithmetic processes are executed based on the information S _n (t) and the like. For example, CPU (Centr
microprocessors such as al Processing Units (GPUs) and Graphics Processing Units (GPUs), Application Specific Integrated Circuits (ASICs), Field Programs (FPGAs)
mable Gate Array), IC (Integrated Circuit) memory and other electronic components.

The arithmetic processing unit 57 includes a biological information acquisition unit 571, a frequency analysis unit 571a, a data aggregation unit 572, a classification unit 574, an appearance data generation unit 575, and a display control unit 578.

The biological information acquisition unit 571 controls data communication with the measurement device 100 and acquires biological information S _n (t) from the measurement device 100 via the communication unit 55.

The frequency analysis unit 571a performs fast Fourier transform on the biological information S _n (t) for each predetermined number of sampling points (for example, 60 points) to thereby obtain a power spectrum P _n.
(Hz) is acquired. And for every predetermined sampling point number (for example, 60 points), for the acquired power spectrum P _n (Hz), LF band: F ₂ [Hz] to F ₁ [Hz],
The frequency component ratio data R _n (t) is obtained by obtaining the intensity ratio of the power spectrum P _{n in} the HF band: F ₁ [Hz] to F ₃ [Hz].

Data aggregation unit 572, (in this embodiment, amount of activity, pulse rate, arterial blood oxygen saturation, skin temperature, six core temperature, and heat flux) N type biometric information S _n (t) and the biological information By performing principal component analysis on N types of frequency component ratio data R _n (t) obtained from S _n (t), the first principal component Y ₁ (t) to the Xth principal component Y _X (t) are obtained. Aggregated into X types of time-series data for classification. The data aggregation unit 572 includes a selection unit 573 that selects M types from X types of time series data for classification. In the present embodiment, the first principal component Y ₁ (t) and the second principal component Y ₂ (t) are selected.

The classification unit 574 plots the aggregated data of the first principal component Y ₁ (t) and the second principal component Y ₂ (t) in a two-dimensional space, and sets each aggregated data based on the distance in the two-dimensional space. Multiple (for example, 9
Classified into cluster G ₁ ~G ₉ of the pieces).

The appearance data generation unit 575 generates appearance data P _C (t) for each of the clusters G _{1 to} G ₉ .
The appearance data generation unit 575 includes a cluster membership degree calculation unit 576 and an appearance probability calculation unit 577.
With. The cluster membership calculation unit 576 generates time-series cluster membership data B _C (t) that determines for each of the clusters G _{1 to} G ₉ whether or not the cluster to which each aggregated data belongs is its own cluster. The appearance probability calculation unit 577 sequentially processes the cluster membership data B _C (t),
The appearance probability is calculated for each calculation time by averaging over a predetermined time width while shifting the calculation time.

The display control unit 578 displays the clusters G _{1 to} G ₉ generated by the appearance data generation unit 575.
The appearance data P _C (t) performs control to display on the display unit 53 as the status analysis of the living body.

The storage unit 59 is realized by a storage medium such as an IC memory, a hard disk, or an optical disk. The storage unit 59 stores in advance a program for operating the biological information processing apparatus 5 to realize various functions provided in the biological information processing apparatus 5, data used during execution of the program, and the like. Alternatively, it is temporarily stored for each processing. Note that the connection between the arithmetic processing unit 57 and the storage unit 59 is not limited to the connection by the internal bus circuit in the apparatus, but is connected to the LAN.
(Local Area Network) or a communication line such as the Internet may be used. In that case, the storage unit 59 may be realized by an external storage device different from the biological information processing device 5.

The storage unit 59 stores a data processing program 591 and analysis result data 593.

The arithmetic processing unit 57 reads and executes the data processing program 591, so that the biometric information acquisition unit 571, the frequency analysis unit 571 a, the data aggregation unit 572, the classification unit 574, the appearance data generation unit 575, the display control unit 578, and the like. Realize the function. Each of these units will be described as being realized by software by the arithmetic processing unit 57 reading and executing the data processing program 591. However, each unit is realized by hardware by configuring an electronic circuit dedicated to each unit. It is also possible to do.

The analysis result data 593 stores appearance data P _C (t) and regression analysis data C (t) of the clusters G _{1 to} G ₉ as biological state analysis results.

[Process flow]
Next, the biological information processing method according to the present embodiment will be described with reference to FIG.
FIG. 43 is a flowchart showing the flow of processing performed by the biological information processing apparatus 5. The processing described here can be realized by causing the arithmetic processing unit 57 to read out and execute the data processing program 591 from the storage unit 59 and operate each unit of the biological information processing apparatus 5.

In the biometric information processing method according to the present embodiment, first, the biometric information acquisition unit 571 is operated by the measurement apparatus 100.
And the biological information S _n (t) measured in time series by the measuring apparatus 100 is acquired (step T1).

Next, as a step of obtaining frequency component ratio data, six types of biological information such as activity amount, pulse rate, arterial oxygen saturation, skin temperature, deep temperature, and heat flux acquired by the frequency analysis unit 571a in step T1. Frequency analysis of S _n (t) and six types of frequency component ratio data R _n
(T) is obtained (step T2).

Next, as the data aggregating step, the data aggregating unit 572, for example, six types of biological information S, such as activity amount, pulse rate, arterial oxygen saturation, skin temperature, deep temperature, and heat flux, acquired in step T1. _n (t) and the frequency analysis unit 571a 6 acquired in step T2.
The principal component analysis is performed on the types of frequency component ratio data R _n (t) to obtain the first principal component Y ₁ (t) to the X principal component Y _X (t) (step T3). Then, the selection unit 573 generates M, for example, a first principal component Y ₁ (t) and a second principal component Y ₂ (t) from the obtained X types of principal components Y ₁ (t) to Y _X (t).
Is selected as M time series data for classification (step T5).

Next, as a classification step, the first principal component Y ₁ selected by the classification unit 574 in step T5.
(T) and the second principal component Y ₂ (t) are plotted in a two-dimensional space as a set (aggregated data) for each time t (step T7). Then, the plotted aggregated data is clustered to be classified into a plurality of clusters G _{1 to} G ₉ (step T9).

Next, as the appearance data generation step, the cluster attribution degree calculation unit 576 in the appearance data generation unit 575 generates time-series cluster data c (t) based on the cluster classification result (step T11). Then, based on the time-series cluster data c (t), time-series cluster membership data B _C (t) is generated for each of the clusters G _{1 to} G ₉ (step T13).
Thereafter, the appearance probability calculation unit 577 calculates the appearance probability for each cluster G _{1 to} G ₉ based on each cluster membership data B _C (t), so that each cluster G _{1 to} G ₉ appears. Data P _C (t) is generated (step T15).

If the appearance data P _C (t) is generated, the display control unit 578 performs control to display the result on the display unit 53 as a living body state analysis result (step T17).

As described above, according to the present embodiment, the N types of biological information S _n (t) and the N types of frequency component ratio data R _n (t) are considered in combination, and the clusters G _{1 to} G ₉ are combined. Appearance data P _C
The state of the living body represented by (t) can be analyzed.

As described above, according to the biological information processing apparatus 5 according to the present embodiment, the following effects can be obtained.
Information that reflects the state of the autonomic nerve can be obtained by obtaining N types of frequency component ratio data from N types (N ≧ 3) of biological information data in time series. Then, by collecting N types of biological information data and N types of frequency component ratio data, acquiring time series data for classification of M types (M ≧ 2 and N> M) useful for biological state analysis Can do.
Then, by classifying M types of time-series data for classification into a plurality of clusters and generating appearance data of each cluster, it is possible to grasp the state of the living body and the degree of the state reflecting the state of the autonomic nerve.

<Modification>
As mentioned above, although embodiment which applied this invention was described, the form which can apply this invention is not restricted to the form mentioned above. For example, the following modifications can be considered.

For example, biological information to be aggregated is not limited to the six types of activity amount, pulse rate, arterial blood oxygen saturation, skin temperature, deep temperature, and heat flux exemplified in the present embodiment, but N types (N ≧ 3). ) Biological information may be selected as appropriate and used as sampling data. In addition to biological information, environmental information such as temperature, humidity, and atmospheric pressure may be included in the sampling data.

Here, as the biological information to be aggregated, the amount of activity, which is the main vital information of the living body,
Selecting any of pulse rate, arterial oxygen saturation, skin temperature, deep temperature, heat flux, blood flow, and blood pressure is preferable because it reflects the state of the living body more accurately in the aggregated time series data It is. In addition, in this embodiment, said biometric information can be acquired using a well-known technique with the heat flow sensor 21, the optical sensor 22, and the motion sensor 23 built in the measuring apparatus 100.

As described above, according to the present modification, biometric information of any one of heat flux, skin temperature, deep temperature, oxygen saturation, blood flow, blood pressure, activity, and pulse rate is aggregated and used in M types. By generating appearance data for each classified cluster, it is possible to correctly grasp the state of the living body.

In the data aggregation process, not all of the N types of frequency component ratio data R _n (t) may be used, and any one or more types may be selected and used.

As described above, according to this modification, by selecting and aggregating at least one of N types of frequency component ratio data, M types of time series data for classification useful for biological state analysis are obtained. Can be acquired.

In addition, the number of principal components selected from X types of principal components (classification time series data) Y ₁ (t) to Y _X (t) is not limited to two types, ie, the first principal component and the second principal component, but three or more types. (However, M ≧ 2 and N>
M) may be selected. At that time, among the X type of principal component _{Y 1 (t) ~Y X (} t), it may select the M kinds descending order of variance _{l 1 (t) ~l X (} t).

In the classification process, if the types of time-series data for classification increase, it becomes possible to classify the state of the living body by reflecting more biological information. On the other hand, the arithmetic processing unit 57 performs an operation required for the classification process. The load increases. For this reason, it is preferable that the time series data for classification used in the classification process is appropriately selected depending on the purpose of the biological information processing.

Further, the number M of principal components to be selected is not limited to a configuration in which a predetermined number (two in the above embodiment) is set in advance. For example, the selection of M types may be performed by selecting a principal component satisfying the selection condition by setting the cumulative contribution rate Pm expressed by the following equation (3) to be equal to or greater than the threshold value Tr. The threshold value Tr may be set in advance, for example, “0.8”.

As described above, according to this modification, N types of biological information data and N types of frequency component ratio data are aggregated into X types (X ≧ M) of time-series data for classification, and M types are selected. As a result, the number of time-series data for classification can be reduced.

Further, the method for aggregating N types of biological information data into time-series data for classification is not limited to principal component analysis. For example, N types of biological information data may be subjected to factor analysis, and M types may be selected from the obtained X types of common factors Z ₁ (t) to Z _X (t). At that time, among the X types of common factors Z ₁ (t) to Z _X (t), M types may be selected in descending order of the variances l ₁ (t) to l _X (t). .

As described above, according to the present modification, N types of biological information data and N types of frequency component ratio data are subjected to principal component analysis or factor analysis, and M types of time series data for classification are acquired based on the variances. Thus, it is possible to acquire M types of time series data for classification useful for analyzing the state of a living body.

Moreover, in this embodiment, the biometric information processing apparatus 5 acquired the user's biometric information from the separate measuring apparatus 100, analyzed, and illustrated the structure which displays an analysis result. On the other hand, the function of the biological information processing apparatus 5 may be incorporated in the measurement apparatus 100, and the biological information processing apparatus 5 and the measurement apparatus 100 may be configured as an integrated apparatus. In addition, the biological information processing apparatus 5 of the present embodiment may be configured as a server via a communication network such as the Internet, and the user may access and use the server via the communication network.

As described above, according to this modification, N types (N ≧ 3) of biological information data in time series are used to calculate N
Information that reflects the state of the autonomic nerve can be acquired by obtaining the frequency component ratio data of the kind. Then, by collecting N types of biological information data and N types of frequency component ratio data, acquiring time series data for classification of M types (M ≧ 2 and N> M) useful for biological state analysis Can do. Then, by classifying M types of time-series data for classification into a plurality of clusters and generating appearance data of each cluster, it is possible to grasp the state of the living body and the degree of the state reflecting the state of the autonomic nerve.

DESCRIPTION OF SYMBOLS 5 ... Biological information processing apparatus, 51 ... Operation input part, 53 ... Display part, 55 ... Communication part, 57 ... Arithmetic processing part, 571 ... Biological information acquisition part, 571a ... Frequency analysis part, 572 ... Data aggregation part, 5
73 ... Selection unit, 574 ... Classification unit, 575 ... Appearance data generation unit, 576 ... Cluster attribution calculation unit, 777 ... Appearance probability calculation unit, 578 ... Display control unit, 59 ... Storage unit, 591 ... Data processing program, 593 ... analysis result data, 100 ... measuring device, 21 ... heat flow sensor, 22
... light sensor, 23 ... motion sensor, 24 ... GPS sensor, N ... communication line.

Claims (9)

The arithmetic processing unit
N types of frequency component ratio data, which is a ratio between a low frequency component lower than a predetermined frequency and a high frequency component higher than the predetermined frequency, from N types (N ≧ 3) of biological information data in time series Seeking steps,
A data aggregating step for aggregating the N types of biological information data and the N types of frequency component ratio data to obtain M types (M ≧ 2 and N> M) of time series data for classification;
A classification step of classifying the M types of time-series data for classification into a plurality of clusters;
An appearance data generation step for generating time-series appearance data for each cluster;
The biological information processing apparatus characterized by performing this. The data aggregation step includes
The N types of biological information data and the N types of frequency component ratio data are converted into X types (X ≧ M
The biological information processing apparatus according to claim 1, further comprising: selecting the M types from the X types of time series data for classification. The data aggregation step performs aggregation by performing principal component analysis or factor analysis of the N types of biological information data and the N types of frequency component ratio data, and based on the variance of the time series data for classification The biological information processing apparatus according to claim 1, wherein the biological information processing apparatus is a step of performing selection. The biological data according to any one of claims 1 to 3, wherein the data aggregation step includes selecting and aggregating at least one of the N types of frequency component ratio data. Information processing device. The classifying step includes plotting the M types of time series data for classification in an M-dimensional space, and clustering the plot into the plurality of clusters. Item 5. The biological information processing apparatus according to any one of items 4. The appearance data generation step includes, for each cluster, calculating an appearance probability for each calculation time obtained by averaging the time series data for classification belonging to the cluster with a predetermined time width while shifting the calculation time. The biological information processing apparatus according to any one of claims 1 to 5. The arithmetic processing unit is
The biological information according to any one of claims 1 to 6, wherein a display control step is performed for controlling display of the time series appearance data on a dial having a time axis in a circumferential direction. Processing equipment. 8. The biological information data includes biological information of any one of heat flux, skin temperature, deep temperature, oxygen saturation, blood flow, blood pressure, activity, and pulse rate. The biological information processing apparatus according to any one of the above. Data aggregation for collecting time-series data for classification of M types (M ≧ 2 and N> M) by collecting time-series N types (N ≧ 3) of biological information data and the N types of frequency component ratio data Steps,
A classification step of classifying the M types of time-series data for classification into a plurality of clusters;
An appearance data generation step for generating time-series appearance data for each cluster;
A biological information processing method comprising: