JP2012179202A - Biological state estimation device and computer program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique with which a state of a person is accurately recognized.SOLUTION: The time series waveforms of frequencies are obtained from time series waveforms of biological signals sampled from an upper body of a person, and further time series waveforms of frequency gradients are obtained. The gradient time series waveforms are frequency-analyzed, and in the relation of logarithmic power spectral density and a logarithmic frequency, a branching point of a regression line is captured, and a state is determined using the appearing position of the branching point and the shape of the regression line, etc. Since the position of the branching point and the shape of the regression line, etc., can be patterned according to the state of the person, the state is easily and accurately determined.

Description

  The present invention relates to a technique for estimating a state of a living body from a relative change from a predetermined state based on a biological signal, in particular, a biological signal obtained from a human trunk, taking into account other biological signals as necessary.

  In recent years, monitoring the biological state of a driver during driving has attracted attention as an accident prevention measure or the like. In Patent Documents 1 to 3, the present applicant discloses a method of determining a sleep onset symptom by arranging a pressure sensor in a seat cushion portion, collecting and analyzing a heel pulse wave.

  Specifically, the maximum value and the minimum value of the time series waveform of the pulse wave are obtained by the smoothing differential method using Savitzky and Golay, respectively. Then, the maximum value and the minimum value are divided every 5 seconds, and the average value of each is obtained. The square of the difference between the average values of the obtained local maximum and local minimum is used as a power value, and this power value is plotted every 5 seconds to create a time series waveform of the power value. In order to read the global change of the power value from this time series waveform, the gradient of the power value is obtained by the least square method for a certain time window Tw (180 seconds). Next, the next time window Tw is similarly calculated at the overlap time Tl (162 seconds), and the result is plotted. This calculation (movement calculation) is sequentially repeated to obtain a time series waveform of the gradient of the power value. On the other hand, the maximum Lyapunov exponent is obtained by chaos analysis of the time series waveform of the pulse wave, the maximum value is obtained by smoothing differentiation, and the time series waveform of the gradient of the maximum Lyapunov exponent is obtained by moving calculation.

  The time series waveform of the power value slope and the time series waveform of the maximum Lyapunov exponent slope are in opposite phase, and furthermore, the time series waveform of the power value slope has a low frequency and large amplitude waveform. The waveform is determined as a characteristic signal indicating a sleep onset sign, and the point at which the amplitude subsequently decreases is determined as the sleep onset point.

  Further, as Patent Document 4, an air bag (air pack) in which a three-dimensional solid knitted fabric is inserted is provided, and the air pack is disposed at a portion corresponding to a person's back to measure air pressure fluctuations of the air pack. A system for detecting a human biological signal from the obtained time-series data of air pressure fluctuation and analyzing the state of the human biological body is disclosed. Non-Patent Documents 1 and 2 also report attempts to detect a human biological signal by arranging an air pack sensor along the lumbar gluteal muscle. This air pressure fluctuation of the air pack is due to shaking of the descending aorta accompanying the movement of the heart, and it is possible to capture a state change closer to the movement of the heart than when using the buttocks pulse wave of Patent Documents 1 and 2. .

JP 2004-344612 A JP 2004-344613 A WO2005 / 092193A1 publication JP 2007-90032 A

“Original, Development of a method for measuring the symptoms of falling asleep using fingertip plethysmogram information” Yasunori Fujita (8 others), Ergonomics Vol 41, No. 1 4 (’05) "Application of biological fluctuation signals measured by non-invasive sensors to fatigue and sleep prediction", Naoki Ochiai (6 others), 39th Annual Meeting of the Japan Ergonomics Society, Chugoku-Shikoku Branch, 2006 Issued on May 25, Publisher: Japan Ergonomics Society Chugoku-Shikoku Branch Office "Prototype of vehicle seat with non-invasive biological signal sensing function", Shinichiro Maeda (4 others), 39th Japan Ergonomics Society China-Shikoku Branch Conference, Proceedings, November 25, 2006 Place: Japan Ergonomics Society Chugoku / Shikoku Branch Office

  As described above, in the techniques of Patent Documents 1 to 4 and Non-Patent Documents 1 to 3, the time-series waveform of the power value gradient and the time-series waveform of the gradient of the maximum Lyapunov exponent are in opposite phases, and the power value gradient When the low-frequency and large-amplitude waveforms occur in the time-series waveform, it is regarded as a sleep onset symptom phenomenon.

  The present applicant has also proposed the following technique as Japanese Patent Application No. 2009-237802. In other words, it is a technique that obtains a time series waveform of a frequency from a time series waveform of a biological signal obtained by a biological signal measuring means, and uses a frequency gradient time series waveform and a frequency variation time series waveform obtained from the time series waveform of this frequency. , Positive / negative of frequency slope time series waveform, positive / negative of integral waveform of frequency slope time series waveform, appearance of reverse phase when frequency slope time series waveform and frequency fluctuation time series waveform are output overlaid (appearance of reverse phase falling asleep) This is a technique for determining a person's condition by combining the signs).

  Although the present applicant has proposed a technique for grasping a person's state using a biological signal as described above, a proposal for a technique for grasping a person's state more accurately is always desired. Further, if there are a plurality of methods for grasping the state of the person, it becomes possible to grasp the state of the person more accurately by using them together. The present invention has been made in view of the above, and an object of the present invention is to provide a technique for analyzing a biological signal using a new analysis method and grasping the state of a person.

  Here, in atrial fibrillation which is one of heart diseases, the place where the characteristics of fluctuations of the heart and circulatory system are switched is said to be 0.0033 Hz, and fluctuations for adjusting fluctuations of 0.0033 Hz are It is said that it exists below 0.0033 Hz. A frequency-gradient time-series waveform for calculating the low-frequency fluctuation inherent in the biological signal is obtained and subjected to frequency analysis. As a result, 0.0017 Hz, which is lower than 0.0033 Hz, and 0.0035 Hz in the vicinity of 0.0033 Hz are mainly used. It was confirmed that the fluctuation occurred. In addition to these two, it was confirmed that fluctuation centered at 0.0053 Hz occurred.

  In view of the above, the inventor of the present invention based on a 0.0035 Hz signal (hereinafter referred to as “fatigue acceptance signal”) as a signal indicating the degree of progress of fatigue in a normal activity state. The signal (hereinafter referred to as “activity adjustment signal”) is a signal in which the degree of influence by the control of the brain and the autonomic nervous system appears during the activity, and a 0.0017 Hz signal (hereinafter referred to as function) having a frequency lower than 0.0033 Hz. The adjustment signal was focused on as a signal for controlling body modulation and functional deterioration.

  On the other hand, the branch point of the regression line that can be grasped from the relationship between the logarithmic power spectral density and logarithmic frequency of the electrocardiogram RR interval variation for 24 hours is in a long period band of 3 to 5 minutes. Usually, the RR interval fluctuation component of the electrocardiogram is in the vicinity of 1 Hz. Therefore, RR interval fluctuation data is required for about 24 hours, and it takes time and labor to measure, so it is not suitable for daily physical condition management. The same can be said for the analysis of the frequency components of the time-series waveform obtained by the zero cross detection method of the biological signal that captures the shaking state of the atrium and the aorta in the vicinity of 0.5 Hz. Therefore, the time-series waveform obtained by the zero-cross detection method of the biological signal is subjected to a gradient time-series analysis to create a time-series waveform composed of components having a long period of 3 to 5 minutes, and analyzed with short-time measurement data. A method was devised. The result of frequency analysis of this time series waveform is considered to show the same tendency as the result of frequency analysis of the time series waveform for 24 hours obtained by the zero cross detection method of the biological signal. Even in about 0.5 to 1 hour, it was estimated that the branch point of the regression line could be captured in the relationship between the logarithmic power spectral density of the periodic function and the logarithmic frequency, and the present invention was completed.

  In addition, as a biological signal, not a finger plethysmogram, but vibration generated on the body surface of the back due to shaking of the atrium, ventricle and aorta (in this specification, the body surface pulse wave detected through the back is particularly We focused on capturing “heart rocking wave” or “Aortic Pulse Wave (APW)”. This is because the wall of the aorta is rich in elasticity among the arteries, and can receive the high pressure of blood pumped directly from the heart, and the valve for preventing backflow is just from the left ventricle of the heart. There is an aortic valve that is. For this reason, by analyzing the biological signal that captures the shaking state of the atrium, ventricle and aorta, it is possible to capture the regulatory ring of the negative feedback mechanism of the brain and autonomic nervous system for maintaining the stasis of the living body, It captures the degree of activity of the heart from the pressure waves generated by the torsional movement of the heart controlled by it, and grasps not only the activity of the autonomic nerve but also the activation state of the brain function, and more accurately This is because the state can be estimated.

That is, the biological state estimation device of the present invention is a biological state estimation device that estimates a biological state using a biological signal collected by a biological signal measuring means,
A frequency calculating means for obtaining a time series waveform of a frequency from a time series waveform of a biological signal at a predetermined measurement time obtained by the biological signal measuring means;
In the time-series waveform of the frequency of the biological signal obtained by the frequency calculation means, movement calculation is performed for obtaining a slope of the frequency for each predetermined time window set with a predetermined overlap time, and obtained for each time window. A frequency gradient time series analysis calculating means for outputting a time series change of the frequency gradient as a frequency gradient time series waveform;
Frequency analysis means for performing frequency analysis on the frequency slope time series waveform obtained from the frequency slope time series analysis calculation means and outputting an analysis waveform indicating the relationship between logarithmic power spectral density and logarithmic frequency;
For the analysis waveform output by the frequency analysis means, a regression line calculation means for obtaining a regression line for each predetermined frequency range;
A branch point specifying means for specifying, as a branch point, a point at which the slope of the regression lines in adjacent frequency ranges changes by a predetermined angle or more from a plurality of regression lines obtained by the regression line calculation means;
The position of the branch point specified by the branch point specifying means, the shape of each regression line obtained by the regression line calculation means, and the magnitude of the amplitude of the analysis waveform of the logarithmic power spectral density obtained by the frequency analysis means And determining means for determining the state of the living body using one or more of the elements.

The determination means includes the analysis of the position of the branch point specified by the branch point specifying means, the shape of each regression line obtained by the regression line calculation means, and the logarithmic power spectral density obtained by the frequency analysis means. It is preferable that at least one element of the amplitude of the waveform is configured to compare the arousal state of these elements with the determination time point and determine the state based on the relative change.
The determination means has no branch point and the slope of each regression line approximates 1 / f ⁿ , or one branch point consisting of the intersections of any of the regression lines, When the position is on the long cycle side with a frequency of 0.01 Hz or less and the slope of the regression line on the short cycle side approximates 1 / f ⁿ with this branch point as the boundary, it is determined that the reference is healthy and awake It is preferable to adopt a configuration to do so.
The determination means has two or more branch points consisting of intersections of any one of the regression lines, and the position of one branch point is on the long period side with a frequency of 0.01 Hz or less, and the one branch point It is preferable to adopt a configuration in which a drowsiness state is determined when there is a difference of a predetermined value or more in logarithmic power spectral density between the first branch point and the other branch point.
Depending on whether the slope of the regression line on the long-period side approximates 1 / f ⁿ , is close to a horizontal state, or rises to the right, in a state of accepting, enduring, It is preferable to adopt a configuration in which it is determined that the player is fighting back.
The state of accepting sleepiness in turn depending on whether the slope of the regression line closer to the short period than the other branch point is close to 1 / f ⁿ , close to a horizontal state, or rising to the right, It is preferable to adopt a configuration in which it is determined that the state is enduring or is countering.

The determination means further reduces the logarithmic power spectral density value corresponding to the frequency of the function adjustment signal in the regression line on the long period side from the branch point of 0.01 Hz or less, compared to when awakening, Means for determining a sleep onset symptom when the value of the logarithmic power spectral density corresponding to the frequency of the activity adjustment signal rises and is higher than the value of the logarithmic power spectral density corresponding to the frequency of the function adjustment signal A configuration is preferable.
The determination means has four or more branch points composed of intersections of any of the regression lines, two of which are located on the long period side with a frequency of 0.01 Hz or less, and the other two The branch point is at a frequency of 0.006 Hz or more and is located on the short cycle side from the branch point on the long cycle side,
There is a difference greater than or equal to a predetermined value in terms of logarithmic power spectral density between the two branch points on the long period side,
It is preferable to adopt a configuration in which a disease state is determined when there is a difference of a predetermined value or more in logarithmic power spectral density between two branch points on the short cycle side.
The slope of the regression line connecting the other branch point on the long cycle side and one branch point on the short cycle side and the regression line on the short cycle side further than the other branch point on the short cycle side is 1 / f ⁿ Depending on whether it is approximated, close to a horizontal state, or rises to the right, it may be determined that the stress associated with the disease is, in effect, accepted, endured, or counterattacked. preferable.
Furthermore, it is preferable that the determination unit determines the active state based on the amplitude of the analysis waveform on the short cycle side of 0.006 Hz or more.

The determination means determines that cancer is suspected when at least one of the regression lines set on the basis of the short-period branch point located at a frequency of 0.006 Hz or more is rising. It is preferable to adopt a configuration in which it is determined that there is.
The determination means is a regression line set based on the short-period branch point located at a frequency of 0.006 Hz or higher in the analysis process of the biological signal measured in the awake state using the biological signal measuring means. Of these, it is preferable to determine that there is a suspicion of cancer when at least two regression lines rise to the right.

The biological signal measuring means is brought into contact with the back portion, and using the biological signal obtained from the back portion, the frequency calculating means, the frequency inclination time series analysis calculating means, the frequency analyzing means, the regression line calculating means, the branch point specifying means, and It is preferable that the determination unit estimates the state of the living body.
The biological signal measuring means is further brought into contact with at least one part of the abdomen and chest, and the biological waveform obtained from at least one part of the abdomen and chest is compared with the analytical waveform obtained from the back part to determine the state of the biological body It is preferable to adopt a configuration for estimating.
The determination means further compares an analysis waveform obtained by frequency analysis of electrocardiogram data obtained from an electrocardiograph with an analysis waveform obtained from the back, abdomen or chest, and between these analysis waveforms, a logarithm It is preferable to have a means for determining whether or not there is a predetermined difference or more in the value of power spectral density.
The determination means is a case where the analysis waveform of the waveform of the electrocardiogram is high, the analysis waveform based on a biological signal collected from the analysis waveform obtained from the back, abdomen or chest is low, and there is a predetermined difference or more between the two. Thus, it is preferable that the sleep state be determined when the difference is within a predetermined range.
The determination means further compares the analysis waveform obtained by frequency analysis of fingertip volume pulse wave data, compares the analysis waveform based on the fingertip volume pulse wave, and a biological signal obtained from the back, abdomen or chest It is preferable to include a means for determining whether or not there is a difference greater than or equal to a predetermined value in the logarithmic power spectral density value with respect to the analysis waveform.
The determination means further compares the analysis waveform obtained by frequency analysis of fingertip volume pulse wave data, and compares the stress caused by the disease or the amplitude of the analysis waveform based on the fingertip volume pulse wave. With regard to sleepiness, it is preferable to have a configuration that determines the degree in the accepting state, the enduring state, or the counterattacking state.
The determination means has a high analysis waveform of the waveform of the electrocardiogram, a low analysis waveform based on a biological signal collected from an analysis waveform obtained from the back, abdomen, or chest, and there is a predetermined difference or more between them. It is preferable to adopt a configuration in which it is determined that there is a tendency for cardiac hypertrophy.
The determination means further compares the analysis waveform obtained by frequency analysis of fingertip volume pulse wave data, compares the analysis waveform based on the fingertip volume pulse wave, and the analysis waveform of the biological signal obtained from the back The slope of the regression line on the short cycle side of the analysis waveform of the fingertip plethysmogram approximates 1 / f ⁿ , whereas the slope of the analysis waveform of the biological signal obtained from the back is horizontal It is preferable to adopt a configuration in which it is determined that there is a sign of atrial fibrillation when close to

The computer program of the present invention is a computer program set in a biological state estimation device that estimates a biological state using a biological signal collected by a biological signal measuring means,
A frequency calculation step for obtaining a time series waveform of a frequency from a time series waveform of a biological signal at a predetermined measurement time obtained by the biological signal measuring means;
In the time-series waveform of the frequency of the biological signal obtained by the frequency calculating step, the movement calculation for obtaining the slope of the frequency is performed for each predetermined time window set with a predetermined overlap time, and is obtained for each time window. A frequency slope time series analysis step of outputting the time series change of the slope of the frequency as a frequency slope time series waveform;
A frequency analysis step for performing a frequency analysis on the frequency gradient time series waveform obtained from the frequency gradient time series analysis calculation step and outputting an analysis waveform indicating a relationship between the logarithmic power spectrum density and the logarithmic frequency;
For the analysis waveform output by the frequency analysis step, a regression line calculation step for obtaining a regression line for each predetermined frequency range;
A branch point specifying step for specifying a point where the slope of the regression lines in the adjacent frequency range changes more than a predetermined angle as a branch point from a plurality of regression lines obtained by the regression line calculation step;
The position of the branch point specified by the branch point specifying step, the shape of each regression line determined by the regression line calculation step, and the amplitude of the analysis waveform of the logarithmic power spectral density determined by the frequency analysis means And a determination step of determining a state of the living body using one or more elements among them.

The determination step includes the analysis of the position of the branch point specified by the branch point specifying step, the shape of each regression line obtained by the regression line calculation step, and the logarithmic power spectral density obtained by the frequency analysis step. It is preferable to determine the state of one or more elements of the amplitude of the waveform by comparing the wakefulness state of these elements with the determination time point, and comparing the relative time.
In the determination step, when there is no branch point and the slope of each regression line is approximated to 1 / f ⁿ , or there is one branch point formed by the intersection of any one of the regression lines, When the position is on the long cycle side with a frequency of 0.01 Hz or less and the slope of the regression line on the short cycle side approximates 1 / f ⁿ with this branch point as the boundary, it is determined that the reference is healthy and awake It is preferable to do.
In the determination step, there are two or more branch points composed of intersections of any one of the regression lines, and the position of one branch point is on the long period side with a frequency of 0.01 Hz or less, and the one branch point It is preferable to determine a drowsiness state when there is a difference of a predetermined value or more in logarithmic power spectral density between the first branch point and the other branch point.
Depending on whether the slope of the regression line on the long-period side approximates 1 / f ⁿ , is close to a horizontal state, or rises to the right, in a state of accepting, enduring, It is preferable to determine that a counterattack is occurring.
The state of accepting sleepiness in turn depending on whether the slope of the regression line closer to the short period than the other branch point is close to 1 / f ⁿ , close to a horizontal state, or rising to the right, It is preferable to determine that the state is enduring or counterattacking.
In the determination step, the value of the logarithmic power spectral density corresponding to the frequency of the function adjustment signal is further reduced in the regression line on the longer period side than the branch point of 0.01 Hz or less, compared to when awakening, A step of determining a sleep-onset phenomenon when the value of the logarithmic power spectral density corresponding to the frequency of the activity adjustment signal is higher than the value of the logarithmic power spectral density corresponding to the frequency of the function adjustment signal. It is preferable.
In the determination step, there are four or more branch points composed of intersections of any of the regression lines, two of which are located on the long period side with a frequency of 0.01 Hz or less, and the other two The branch point is at a frequency of 0.006 Hz or more and is located on the short cycle side from the branch point on the long cycle side,
There is a difference greater than or equal to a predetermined value in terms of logarithmic power spectral density between the two branch points on the long period side,
It is preferable that the disease state is determined when there is a predetermined difference or more in logarithmic power spectral density between the two branch points on the short cycle side.
The slope of the regression line connecting the other branch point on the long cycle side and one branch point on the short cycle side and the regression line on the short cycle side further than the other branch point on the short cycle side is 1 / f ⁿ It is preferable to determine whether the stress associated with the disease is in an accepting state, a tolerating state, or a countering state in order depending on whether it is approximate, close to a horizontal state, or rises to the right.
Furthermore, it is preferable that the determination step determines the active state based on the amplitude of the analysis waveform on the short cycle side of 0.006 Hz or more.

  The present invention has means for obtaining a time series waveform of a frequency from a time series waveform of a biological signal, and further obtaining a time series waveform of a frequency gradient to perform frequency analysis. By performing such a gradient time series analysis, a time series waveform consisting of components with a long period of 3 to 5 minutes is created, and the result of frequency analysis of the time series waveform for 24 hours while being short-time measurement data Data showing the same tendency as can be obtained. In other words, it is possible to grasp the state of shaking of the atrium, ventricle and aorta in the vicinity of 0.5 Hz, which will not appear unless they are measured for a long time of about 24 hours. Effective above. Capturing the atrial, ventricular, and aortic swaying is to capture the reaction force of the living body's homeostatic function and the stress to changes in the body environment caused by the lesion, and this reverses the movement of the heart. The state of control can be estimated. As a result, according to the present invention, it is possible to capture not only the current state estimation but also a sudden change signal of physical condition that may occur in the future. In addition, the present invention performs frequency analysis of the gradient time series waveform, captures the branch point of the regression line in the relationship between the logarithmic power spectral density and the logarithmic frequency, and uses the position where the branch point appears, the shape of the regression line, etc. However, since the position of the branch point, the shape of the regression line, and the like can be classified according to the state of the person, the state determination can be performed easily and accurately.

FIG. 1 is a diagram showing a biological signal measuring means used in one embodiment of the present invention. FIG. 2 is a diagram showing another aspect of the biological signal measuring means according to the embodiment. FIG. 3 is a diagram showing a process of incorporating the biological signal geodetic means into the sheet. FIG. 4 is a diagram illustrating a configuration of the biological state estimation apparatus according to an embodiment of the present invention. FIG. 5 shows the result of frequency analysis of data obtained by connecting the original waveforms of heart swing waves in the sleepless experiment of 22 subjects without sleepiness (active state), and heart swing in the sleepy state. It is the figure which showed the frequency analysis result of the data which connected the original waveform of the wave. FIG. 6 is a diagram for explaining a method of processing a biological signal detected from the biological signal measuring unit using a peak detection method and a zero cross method. FIG. 7A is a diagram showing a comparison of detection points when the peak detection method and the zero cross method are applied to the APW, and FIG. 7B is a diagram illustrating the APW obtained by applying the peak detection method and the zero cross method. It is a time series waveform. FIG. 8 is a diagram comparing the detection sensitivities of the time series waveforms using the peak detection method and the zero cross method by frequency analysis. FIG. 9 is a diagram for explaining how to obtain a frequency gradient time-series waveform. FIGS. 10A and 10B are data showing analysis waveforms and regression lines in a state where the subjects a and b do not have sleepiness (wakefulness state). FIGS. 11A and 11B are data showing analysis waveforms and regression lines in a state in which subjects c and d do not have sleepiness (wakefulness state). FIGS. 12A and 12B are data showing analysis waveforms and regression lines in a state where the subjects e and f do not have sleepiness (wakefulness state). FIGS. 13A and 13B are data showing analysis waveforms and regression lines in a state where the subjects g and h do not have sleepiness (wakefulness state). FIGS. 14A and 14B are data showing analysis waveforms and regression lines in a state where the subjects i and j do not have sleepiness (wakefulness state). FIG. 15 is data showing an analysis waveform and a regression line in a state where the subject k does not have sleepiness (wakefulness state). FIG. 16 is data showing an analysis waveform and a regression line in a state in which the subject A is drowsy. FIGS. 17A and 17B are data showing analysis waveforms and regression lines in a state where the sleepiness of subjects B and C occurred. FIGS. 18A and 18B are data showing analysis waveforms and regression lines in a state where the sleepiness of subjects D and E is generated. FIGS. 19A and 19B are data showing analysis waveforms and regression lines in a state where the sleepiness of subjects F and G occurred. FIGS. 20A and 20B are data showing analysis waveforms and regression lines in a state where the sleepiness of subjects H and I occurred. FIGS. 21A and 21B are data showing analysis waveforms and regression lines in a state where the sleepiness of subjects J and K occurred. 22 (a) and 22 (b) are data showing analysis waveforms and regression lines in a state where the sleepiness of subjects L and M occurred. 23 (a) and 23 (b) are data showing analysis waveforms and regression lines in a state where the sleepiness of subjects N and O occurred. 24 (a) and 24 (b) show the time series waveform obtained by linking the data of all subjects in the sleepless state (wakeful state) and the sleepy state, respectively, and processing by the zero cross method, and performing frequency analysis. It is the graph which calculated | required the analysis waveform and the regression line. FIG. 25 shows time-series waveforms of healthy male subject S in the 60s in Test Example 2, and in order from the top, the frequency gradient time-series waveform and peak detection by the zero cross method of the heart oscillating wave Frequency gradient time-series waveform by the method, frequency gradient weighted average absolute value time-series waveform of the heart swing wave by the zero-cross method and peak detection method, waveform showing the state judgment result obtained from each time-series waveform, frequency by the zero-cross method It is the figure which showed the frequency analysis rate analysis waveform in the direction of fluctuation, and the frequency analysis rate analysis waveform in the direction of frequency fluctuation by the peak detection method. FIG. 26 is a diagram showing an analysis waveform and a regression line of a healthy male subject S in his 60s in Test Example 2 according to the present invention. FIG. 27 is a diagram (from 9:06 to 1.5 hours) showing an analysis waveform and a regression line according to the present invention of a healthy male subject F in his 50s in Test Example 2. FIG. 28 is a diagram (from 11:12 to 1.5 hours) showing an analysis waveform and a regression line according to the present invention of a healthy male subject F in his 50s in Test Example 2. FIG. 29 is a diagram showing an analysis waveform and a regression line according to the present invention of a healthy male subject F in his 50s in Test Example 2 (from 13:54 to 1 hour). FIGS. 30A and 30B are diagrams showing an analysis waveform and a regression line of the logarithmic power spectral density of the male subject FG in Test Example 3 when he / she is overworked (sitting position) and sleeping (prone position). FIG. 31 is a diagram showing an analysis waveform of a logarithmic power spectral density and a regression line when the male subject FG in Test Example 3 recovers (when getting up). FIG. 32 is a diagram showing an analysis waveform and a regression line of logarithmic power spectral density at the time of onset of food poisoning (at the time of onset of dehydration) of male subject FG in Test Example 3. FIG. 33A is a diagram showing an analysis waveform and a regression line of the logarithmic power spectral density at the time of administration of the male subject FG in Test Example 3, and FIG. 33B is a logarithmic power spectral density on the first day of discontinuation of administration. It is the figure which showed the analysis waveform and regression line. 34 (a) and 34 (b) are graphs showing logarithmic power spectral density analysis waveforms and regression lines for the second day and third day of discontinuation of administration of male subject FG in Test Example 3. FIG. FIGS. 35A to 35F are data of healthy male AO in his 20s in Test Example 4. FIG. 36 (a) to 36 (f) are data of healthy male U in his twenties in Test Example 4. 37A to 37F are data of a healthy male HO in the 89s in Test Example 4. 38A to 38F are data when the sleep time of a healthy male HO in the 89s in Test Example 4 is 4 and a half hours. FIGS. 39A to 39F are data of a 86-year-old male subject (Mr. Yoshito Fujita) in Test Example 4. FIG. FIG. 40 is a diagram showing the test results of the subject's blood and the like in FIG. FIG. 41 is a diagram showing the CT imaging results of the subject in FIG. 42 (a) to 42 (f) are diagrams showing measurement results of a 62-year-old male subject YA. 43 (a) to 43 (f) are diagrams showing data about four months ago for the subject YA in FIG. 44 (a) to 44 (f) are data of a 73-year-old subjectively healthy male subject HO. FIG. 45 (a) is a diagram showing a representative example of time-series data of fingertip plethysmogram in which a sleep onset symptom has occurred, and FIG. 45 (b) shows the sleep onset symptom in the analysis waveform of the present invention. FIG. 45 (c) is a diagram showing a means for determining a sleep onset sign phenomenon based on the characteristic three signal distribution rates. FIG. 46 is a diagram schematically showing a case of each subject in Test Example 4 collectively. FIG. 47 is a diagram for explaining a determination method using an analysis waveform and a regression line according to the present invention. FIG. 48 is a diagram showing the relationship between the waveforms of the distribution ratios of the three frequency components of the heart oscillating wave and the state of each subject for the subject example of the sleep induction experiment. FIG. 49 is a diagram showing a gradient time-series waveform pattern by the zero cross method in the awake state. FIG. 50 is a diagram showing the correlation between the results of FIGS. 48 and 49 and the human state. FIG. 51 is a diagram showing the correlation between the results of FIGS. 48 and 49 and the human condition. 52 (a) and 52 (b) are diagrams showing data obtained by analyzing a cardiac sway wave (APW) from the back by the biological signal measuring means 1 in a static sitting state for a patient with atrial fibrillation. 52 (c) is a diagram showing data obtained by analyzing the fingertip volume pulse wave obtained from the fingertip volume pulse wave meter. 53 shows an analysis result (corresponding to FIG. 52 (a)) obtained by analyzing 71-minute measurement data of APW of the subject (Mr. W) in FIG. 52, and APW of another subject (Mr. I) having atrial fibrillation. It is the figure which showed the regression line obtained from the analysis result of measurement data for 10 minutes. FIG. 54 is a diagram showing a regression line for healthy subjects and a regression line for patients with atrial fibrillation. FIG. 55 is a diagram showing analysis results obtained by laying the subject U on the ambulance vibration isolator, where (a) to (c) show data for 30 minutes, and (d) show data for 40 minutes. It is. FIG. 56 is a diagram showing analysis results obtained by laying the subject KO on the ambulance vibration isolator, where (a) to (c) show 30-minute data, and (d) shows 40-minute data. It is. FIGS. 57A to 57C are diagrams showing the results of a sleep experiment in the 60-minute sitting posture of the subject U. FIG. 58 is a diagram summarizing determination methods for state estimation by the biological state estimation apparatus of the present invention.

  Hereinafter, the present invention will be described in more detail based on the embodiments of the present invention shown in the drawings. 1 and 2 are biological signals to be analyzed by the biological state estimating apparatus 60 according to the present embodiment, in this case, cardiac swing waves (atrial and ventricular motion and aorta detected from the back of a human upper body). FIG. 3 is a diagram showing a biological signal measuring means 1 for collecting a biological signal that is a body surface pulse wave accompanying the oscillation of the body and directly reflects the motion of the heart. It is the figure which showed the process of incorporating in the seat 100 for motor vehicles. First, the biological signal measuring means 1 will be described. The biological signal measuring means 1 includes a three-dimensional solid knitted fabric 10, a three-dimensional solid knitted fabric support member 15, a film 16, plate-like foams 21 and 22, and a vibration sensor 30.

  For example, as disclosed in JP-A-2002-331603, the three-dimensional solid knitted fabric 10 includes a pair of ground knitted fabrics spaced apart from each other and a pair of ground knitted fabrics that reciprocate between the pair of ground knitted fabrics. The knitted fabric has a three-dimensional three-dimensional structure having a large number of connecting yarns for joining the two.

  One ground knitted fabric is formed by, for example, a flat knitted fabric structure (fine stitches) that is continuous in both the wale direction and the course direction from a yarn obtained by twisting a single fiber. For example, a knitted structure having a honeycomb-shaped (hexagonal) mesh is formed from a yarn obtained by twisting short fibers. Of course, this knitted fabric structure is arbitrary, and it is also possible to adopt a knitted fabric structure other than a fine structure or a honeycomb shape, and a combination thereof is also arbitrary, such as adopting a fine structure for both. The connecting yarn is knitted between two ground knitted fabrics so that one ground knitted fabric and the other ground knitted fabric maintain a predetermined distance. In this embodiment, since the solid vibration of the three-dimensional solid knitted fabric, in particular, the string vibration of the connecting yarn is detected, the connecting yarn is preferably composed of a monofilament. However, the resonance depends on the type of biological signal to be collected. In order to adjust the frequency, the connecting yarn can also be composed of multifilaments.

  Further, the three-dimensional solid knitted fabric 10 has a load-deflection characteristic in the thickness direction that is placed on a measurement plate and pressed with a pressure plate having a diameter of 30 mm or 98 mm in diameter within a range up to a load of 100 N. It is preferable to provide a spring constant approximating the load-deflection characteristics of the muscles of the buttocks. Specifically, the spring constant when pressed with a pressure plate with a diameter of 30 mm is in the range of 0.1 to 5 N / mm, or the spring constant when pressed with a pressure plate with a diameter of 98 mm is 1 to 10 N / mm. It is preferable to use one that is mm. By approximating the load-deflection characteristics of the muscles of the human buttocks, the three-dimensional knitted fabric balances with the muscles, and when biological signals such as heartbeat, breathing, atrial and aortic oscillations are propagated, the three-dimensional The three-dimensional knitted fabric generates vibration similar to that of human muscles, and can propagate a biological signal without greatly attenuating it.

  As such a three-dimensional solid knitted fabric, for example, the following can be used. Each three-dimensional solid knitted fabric can be used by stacking a plurality of pieces as necessary.

(1) Product number: 49076D (manufactured by Sumie Textile Co., Ltd.)
Material:
Front side ground knitted fabric: twisted yarn of 300 dtex / 288 f polyethylene terephthalate fiber false twisted yarn and 700 dtex / 192 f polyethylene terephthalate fiber false twisted yarn Back side ground knitted fabric: 450 dtex / 108 f polyethylene Combination of terephthalate fiber false twisted yarn and 350 decitex / 1f polytrimethylene terephthalate monofilament Linked yarn ... 350 decitex / 1f polytrimethylene terephthalate monofilament

(2) Product number: 49011D (manufactured by Sumie Textile Co., Ltd.)
Material:
Ground knitted fabric (warp) ... 600 dtex / 192f polyethylene terephthalate fiber false twisted yarn Ground knitted fabric (weft) ... 300 dtex / 72f polyethylene terephthalate fiber false twisted yarn Linked yarn ... ... 800 decitex / 1f polyethylene terephthalate monofilament

(3) Product number: 49013D (manufactured by Sumie Textile Co., Ltd.)
Material:
Front side ground knitted fabric: two twisted yarns of 450 dtex / 108f polyethylene terephthalate fiber false twisted yarn Back side ground knitted fabric ... 450 twists of polyethylene terephthalate fiber false twisted yarn of 108 dtex / 108f Connecting thread: 350 dtex / 1f polytrimethylene terephthalate monofilament

(4) Product number: 69030D (manufactured by Sumie Textile Co., Ltd.)
Material:
Front side ground knitted fabric: two twisted yarns of 450 dtex / 144 f polyethylene terephthalate fiber false twisted yarn Back side ground knitted fabric: 450 dtex / 144 f polyethylene terephthalate fiber false twisted yarn and 350 dtex / 1 f Combined with polytrimethylene terephthalate monofilaments of linking yarns ... 350 dtex / 1f polytrimethylene terephthalate monofilaments

(5) Product number manufactured by Asahi Kasei Fibers Co., Ltd .: T24053AY5-1S

  The plate-like foams 21 and 22 are preferably composed of bead foams. As the bead foam, for example, a foam molded body by a bead method of a resin containing at least one of polystyrene, polypropylene, and polyethylene can be used. The plate-like foams 21 and 22 made of bead foam propagate biosignals with minute amplitudes as membrane vibrations due to the characteristics of spherical resin films formed by foaming that constitute individual fine beads. . This membrane vibration is transmitted to the three-dimensional solid knitted fabric as a string vibration, the membrane vibration and the string vibration are superimposed, and the biological signal is a vibration described later as a mechanical vibration amplified by the superposition of the membrane vibration and the string vibration. Detected by sensor 30. Therefore, detection of a biological signal becomes easy.

  When the plate-like foams 21 and 22 are formed of bead foams, the foaming ratio is preferably in the range of 25 to 50 times and the thickness is preferably less than the average diameter of the beads. For example, when the average diameter of 30 times expanded beads is about 4 to 6 mm, the thickness of the plate-like foams 21 and 22 is sliced to about 3 to 5 mm. As a result, soft elasticity is imparted to the plate-like foams 21 and 22, and solid vibration resonating with vibration having a small amplitude is easily generated. The plate-like foams 21 and 22 may be arranged on both sides of the three-dimensional solid knitted fabric 10 as in the present embodiment, but are arranged on either one side, preferably only on the seat back side. It can also be configured.

  Here, the three-dimensional solid knitted fabric 10 is a strip having a width of 40 to 100 mm and a length of 100 to 300 mm. With this size, preliminary compression (a state in which tension is generated in the connecting yarn) is likely to occur in the three-dimensional solid knitted fabric 10, and an equilibrium state is easily created between the person and the three-dimensional solid knitted fabric 10. In this embodiment, in order to reduce a sense of incongruity when a person comes into contact with the back, two sheets are arranged on the object with a portion corresponding to the spinal column in between. In order to easily arrange the three-dimensional solid knitted fabric 10 at a predetermined position, the three-dimensional solid knitted fabric 10 is preferably supported by a three-dimensional solid knitted fabric support member 15 as shown in FIG. The three-dimensional three-dimensional knitted fabric support member 15 is formed in a plate shape, and two vertically arranged through holes 15a and 15a are formed at symmetrical positions with a portion corresponding to the spine. The three-dimensional three-dimensional knitted fabric support member 15 is preferably composed of a bead foam formed in a plate shape, like the plate foams 21 and 22. When the three-dimensional three-dimensional knitted fabric support member 15 is composed of a bead foam, a preferable foaming ratio and thickness range are the same as those of the plate-like foams 21 and 22. However, in order to cause the membrane vibration more remarkably by the biological signal, the thickness of the plate-like foams 21 and 22 laminated on the upper and lower sides of the three-dimensional solid knitted fabrics 10 and 10 is set to It is preferable that the thickness is smaller than the thickness.

  In the state where the two three-dimensional solid knitted fabrics 10 and 10 are inserted and arranged in the placement through holes 15a and 15a formed in the three-dimensional solid knitted fabric support member 15, the film 16 on the front side and the back side of the three-dimensional solid knitted fabric 10 and 10 16 are stacked. In the present embodiment, the peripheral portions of the films 16 and 16 are attached to the peripheral portions of the placement through holes 15a and 15a and laminated. Note that the formation positions of the placement through holes 15a, 15a (that is, the placement positions of the three-dimensional three-dimensional knitted fabrics 10, 10) are swaying caused by movement accompanying the atrium and aorta (particularly, the “descending aorta”) A position corresponding to a region where the movement of the aortic valve can be detected is preferable. As a result, the three-dimensional solid knitted fabrics 10 and 10 are sandwiched between the upper and lower surfaces by the plate-shaped foams 21 and 22 and the peripheral portion is surrounded by the three-dimensional solid knitted fabric support member 15. The three-dimensional solid knitted fabric support member 15 functions as a resonance box (resonance box). The wall of the aorta is rich in elasticity among the arteries, and can receive the high pressure of blood pumped directly from the heart. There is a certain aortic valve. For this reason, when the position of the three-dimensional solid knitted fabric is set to the above position, the movement of the negative feedback mechanism between the brain and the autonomic nervous system for maintaining the continuity of the living body can be well understood.

  Further, it is preferable to use a material in which the three-dimensional solid knitted fabrics 10 and 10 are thicker than the three-dimensional solid knitted fabric support member 15. That is, when the three-dimensional solid knitted fabrics 10 and 10 are arranged in the placement through holes 15a and 15a, the front and back surfaces of the three-dimensional solid knitted fabrics 10 and 10 protrude from the placement through holes 15a and 15a. Thickness. Thereby, since the three-dimensional solid knitted fabrics 10 and 10 are pressed in the thickness direction when the peripheral portions of the films 16 and 16 are attached to the peripheral portions of the placement through holes 15a and 15a, the reaction force of the films 16 and 16 Tension is generated and solid vibration (membrane vibration) is likely to occur in the films 16 and 16. On the other hand, pre-compression occurs also in the three-dimensional solid knitted fabrics 10 and 10, and the connecting yarn that holds the thickness form of the three-dimensional solid knitted fabric is also subjected to tension due to reaction force, and string vibration is likely to occur. In addition, although it is preferable to provide the films 16 and 16 on both sides of the front side and the back side of the three-dimensional solid knitted fabrics 10 and 10, it is also possible to have a configuration provided on at least one of them.

  Since the connecting yarn of the three-dimensional solid knitted fabric 10, 10 is stretched between a pair of ground knitted fabrics, it becomes a long string wound in a coil shape, and a film that functions as a resonance box (resonance box) at upper and lower nodes. 16 and 16 and plate-like foams 21 and 22 are disposed. Since a biological signal typified by heart rate variability has a low frequency, it is amplified by a resonance system having such a long string and a large number of nodes. That is, the string vibration of the connecting yarn causes the film vibration of the films 16 and 16 and the film vibration of the beads of the plate-like foams 21 and 22 through a large number of nodes, which act in an overlapping manner and are amplified. In addition, the space | interval between the nodes of the connection thread | yarn of a three-dimensional solid knitted fabric, ie, the arrangement | positioning density of a connection thread | yarn, is so preferable.

  Further, the films 16 and 16 are adhered to the plate-like foams 21 and 22 in advance and integrated, and the plate-like foams 21 and 22 are simply laminated on the three-dimensional three-dimensional knitted fabric support member 15. It is also possible to adopt a configuration in which 16 can be arranged on the front side and the back side of the three-dimensional solid knitted fabrics 10 and 10. However, in order to apply pre-compression to the three-dimensional solid knitted fabrics 10, 10, it is preferable to fix the films 16, 16 to the surface of the three-dimensional solid knitted fabric support member 15 as described above. In addition, as shown in FIG. 1, a film is not disposed corresponding to each three-dimensional solid knitted fabric 10, but two two-dimensional solid knitted fabrics 10, 10 are covered as shown in FIG. You may make it use the film 16 of the magnitude | size which can be used.

  For example, a plastic film made of polyurethane elastomer (for example, product number “DUS605-CDR” manufactured by Seadam Co., Ltd.) is preferably used as the films 16 and 16 in order to capture heart rate fluctuation. However, since the films 16 and 16 generate membrane vibration due to resonance if the natural frequencies match, this is not restrictive, and it depends on the object to be collected (heartbeat, breathing, shaking of the atrium, ventricle, and aorta). It is preferable to use one having a natural frequency. For example, as shown in a test example described later, a biaxially formed material made of a material having low stretchability, for example, a nonwoven fabric made of thermoplastic polyester (for example, polyethylene naphthalate (PEN) fiber (1100 dtex) manufactured by Teijin Ltd.) It is also possible to use a woven fabric (length: 20 / inch, width: 20 / inch). Further, for example, an elastic fiber nonwoven fabric having an elongation of 200% or more and a recovery rate of 100% or more at 80% or more (for example, trade name “Espancione” manufactured by KB Seiren Co., Ltd.) It is also possible to use.

  The vibration sensor 30 is fixedly disposed on one of the three-dimensional solid knitted fabrics 10 before the above-described films 16 and 16 are laminated. The three-dimensional three-dimensional knitted fabric 10 is composed of a pair of ground knitted fabrics and connecting yarns, and the string vibration of each connecting yarn is applied to the films 16 and 16 and the plate-like foams 21 and 22 via the nodes with the ground knitted fabric. In order to be transmitted, the vibration sensor 30 preferably fixes the sensing unit 30a to the surface of the three-dimensional solid knitted fabric 10 (the surface of the ground knitted fabric). As the vibration sensor 30, it is preferable to use a microphone sensor, in particular, a condenser microphone sensor. In this embodiment, since it is not necessary to consider the sealing property of the portion where the microphone sensor is disposed (that is, the placement through hole 15a where the three-dimensional solid knitted fabric 10 is disposed), the lead wire of the microphone sensor is easily wired. be able to. In the present embodiment, as described above, the vibration of the body surface via the human muscle and skeleton accompanying the biological signal propagates not only to the three-dimensional solid knitted fabric 10 but also to the plate-like foams 21 and 22 and the film 16. These are vibrated (string vibration, membrane vibration) and superimposed and amplified. Therefore, the vibration sensor 30 is not limited to the three-dimensional solid knitted fabric 10, and the sensing unit 30 a can be fixed to the plate-like foams 21 and 22 and the film 16 constituting the vibration transmission path. In this embodiment, since the three-dimensional solid knitted fabric 10, the three-dimensional solid knitted fabric support member 15, the plate-like foams 21, 22 and the film 16 mechanically amplify the biological signal, these constitute a mechanical amplification device. To do.

  For example, as shown in FIG. 3, the above-described biological signal measuring unit 1 is disposed inside the skin 120 covered by the seat back frame 110 of the automobile seat 100. In order to facilitate the arrangement work, the three-dimensional solid knitted fabric 10, the three-dimensional solid knitted fabric support member 15, the film 16, the plate-like foams 21, 22, the vibration sensor 30 and the like constituting the biological signal measuring means 1 are previously united. It is preferable to make it.

   The biological signal measuring means 1 described above is a mechanical amplification device including a three-dimensional solid knitted fabric 10 and plate-like foams 21 and 22 laminated around the three-dimensional solid knitted fabric 10, preferably the three-dimensional solid knitted fabric 10. And the plate-like foams 21 and 22 have a mechanical amplification device in which a film 16 is disposed, and a vibration sensor is attached to the mechanical amplification device. Microvibrations on the body surface due to human biological signals such as heartbeat, breathing, shaking of the atrium, ventricle, and aorta are propagated to the plate-like foams 21, 22, the film 16, and the three-dimensional solid knitted fabric 10. The foams 21 and 22 and the film 16 cause membrane vibration, and the three-dimensional solid knitted fabric causes yarn string vibration.

  Further, the three-dimensional solid knitted fabric 10 includes a connecting yarn disposed between a pair of ground knitted fabrics, and has a load-deflection characteristic approximate to a load-deflection characteristic of human muscles. Therefore, the load-deflection characteristic of the mechanical amplifying device including the three-dimensional solid knitted fabric 10 is approximated to that of the muscle, and it is arranged adjacent to the muscle, so that the muscle and the three-dimensional solid knitted fabric are arranged. The internal and external pressure differences are equal, and biological signals such as heartbeat, respiration, atrial and aortic oscillations can be accurately transmitted, and thereby string vibration is generated in the yarn (particularly the connecting yarn) constituting the three-dimensional solid knitted fabric 10. Can be generated. Further, the plate-like foams 21 and 22, preferably bead foams, laminated on the three-dimensional solid knitted fabric 10 tend to cause membrane vibration on each bead due to the soft elasticity and small density of the beads. Since the film 16 has a peripheral edge fixed and elastically supported by a three-dimensional solid knitted fabric that approximates the load-deflection characteristics of human muscles, a predetermined tension is generated, and thus membrane vibration is likely to occur. That is, according to the biological signal measuring means 1, in a mechanical amplifying device having a load-deflection characteristic that approximates a load-deflection characteristic of a muscle by a biological signal such as heartbeat, respiration, atrial, ventricular and aortic swinging. Membrane vibrations are generated in the plate-like foams 21 and 22 and the film 16, and string vibrations are generated in the three-dimensional solid knitted fabric 10 having a load-deflection characteristic approximate to the load-deflection characteristic of human muscles. The string vibration of the three-dimensional solid knitted fabric 10 again affects the film vibration of the film 16 and the like, and these vibrations are superimposed and act. As a result, the vibration input from the body surface along with the biological signal is directly detected by the vibration sensor 30 as a solid vibration amplified by superposition of the string vibration and the membrane vibration.

  As the biological signal measuring means 1 used in the present invention, it is possible to use a conventional configuration for detecting the air pressure fluctuation in the sealed bag, but since the volume and the pressure are in inverse proportion, the sealed It is difficult to detect pressure fluctuations unless the bag volume is reduced. On the other hand, according to the biological signal measuring means 1 described above, it is propagated to the mechanical amplification devices (three-dimensional solid knitted fabric 10, plate-like foams 21, 22, and film 16) as described above, not the air pressure fluctuation. The volume (volume) is rarely limited from the viewpoint of detection sensitivity, and the amplitude associated with heartbeat, breathing, atrial and ventricular and aortic oscillations, etc. Small vibrations can be detected with high sensitivity. For this reason, it can respond to people with various physiques. Therefore, the biosignal measuring means 1 can detect a biosignal with high sensitivity even in an environment where people having various physiques, such as a vehicle seat, and various external vibrations are input. Further, since it is not necessary to make a sealed structure, the manufacturing process is simplified, the manufacturing cost can be reduced, and it is suitable for mass production.

  In addition, although the above-mentioned biological signal measuring means 1 is incorporated in the inside of the skin 120 of the seat 100, it may be incorporated in a seat cushion that is attached to the surface of the skin 120 later. However, when attaching as a retrofit, it is preferable to provide a hard surface between the sheet and the three-dimensional solid knitted fabric so that the three-dimensional solid knitted fabric is likely to be pre-compressed due to the weight. It is preferable to insert a three-dimensional knitted fabric or a plate having a thickness of about 1 to 2 mm made of synthetic resin such as polypropylene. For example, in the case of a sheet having a soft compression characteristic, a three-dimensional solid knitted fabric is not pre-compressed, and thus a biological signal is absorbed without being reflected, but by providing a hard surface as described above, Variations in the compression characteristics on the sheet side are absorbed, and a biological signal having a large amplitude is easily obtained.

  Next, the configuration of the biological state estimation device 60 will be described with reference to FIG. The biological state estimation device 60 includes a frequency calculation unit 610, a frequency gradient time series analysis calculation unit 620, a frequency analysis unit 630, a regression line calculation unit 640, a branch point identification unit 650, and a determination unit 660. The biological state estimation device 60 is configured by a computer, the frequency calculation unit 610 executes the frequency calculation step, the frequency gradient time series analysis calculation unit 620 executes the frequency gradient time series analysis calculation step, and the frequency analysis unit 630 performs the frequency calculation. An analysis step is executed, a regression line calculation step is executed by the regression line calculation means 640, a branch point specification step is executed by the branch point specification means 650, and a determination step is executed by the determination means 660. The computer program can be provided by being stored in a recording medium such as a flexible disk, a hard disk, a CD-ROM, an MO (magneto-optical disk), a DVD-ROM, or a memory card, or transmitted through a communication line. Is possible.

  The frequency calculation means 610 is time-series data (APW) of the output signal obtained from the vibration sensor 30 of the biological signal measurement means 1, preferably a filtering process (for example, a filtering process for removing a frequency component generated by body movement). A time-series waveform of the frequency in the time-series data in the predetermined frequency region is obtained.

  In the frequency calculation means 610, the time series waveform of the frequency is obtained by using a point (hereinafter referred to as “zero cross point”) where the output signal is obtained from the vibration sensor of the biological signal measuring means 1 and switched from positive to negative. A method of obtaining (hereinafter referred to as “zero-cross method”) and a method of obtaining a time-series waveform using a maximum value (peak) by smoothing and differentiating a time-series waveform of an output signal obtained from the vibration sensor of the biological signal measuring means 1 There are two methods (hereinafter referred to as “peak detection method”).

  Here, FIG. 5 shows a result of frequency analysis by connecting APW data of 22 subjects in a sleep induction experiment in a state where there are 22 subjects without sleepiness and with sleepiness. Regardless of drowsiness or not, there is a peak at 0.6Hz and 1.2Hz. However, there was no drowsiness, and the magnitude of the power spectrum near 0.6 Hz occurred. Therefore, 0.6Hz and 1.2Hz are components generated by heart motion and heart rate / pulse, and we thought that the heart motion and heart rate / pulse were controlled by these frequencies. In particular, since the 0.6Hz component changes depending on the presence or absence of sleepiness, it was considered to be a frequency strongly related to changes in the state of humans such as the presence or absence of sleepiness, that is, changes in autonomic nerves, particularly sympathetic nerves. Drowsiness occurs as fatigue progresses and is reduced when the compensatory effects of sympathetic nerves occur. In addition, when sleepiness occurs, the activity level decreases, the control of the homeostatic function is simplified, and the rhythm and amplitude of cardiac contraction / dilation may be relaxed. Therefore, the 0.5th-order component decreased when drowsiness occurred, and the heart movement became simple. On the other hand, in the active state, since the heart pumps out the necessary blood volume, a large movement occurs in the heart, and the 0.5th-order component increases due to the rotational motion with the addition of the torsion of the ventricle. Therefore, there was a possibility that the occurrence of drowsiness could be captured and predicted by capturing the fluctuation of 0.6 Hz. The APW that shows heart motion and heart rate / pulse and pulse wave components is a band centered on 1/3 to 1/2 frequency components of the heart rate / pulse rate (0.5th order) and a heart rate / pulse rate band (1 It is composed of two major components. The first-order component can be directly regarded as APW, but the 0.5th-order component cannot be directly regarded as APW. Therefore, a method for obtaining the 0.5th order component is required. As a method of capturing this 0.5th order component, the present inventor considered the above-described zero cross detection method and peak detection method (corresponding to the pulse rate).

  The zero-cross detection method is suitable for capturing changes in the APW fundamental frequency, and the peak detection method is the heart rate, that is, the APW composite wave, for example, the frequency fluctuation of the high frequency component counted as the heart rate and the fluctuation of the fundamental frequency. It is suitable for capturing. In the fluctuation analysis, the slope time series analysis is applied to the time series waveform obtained by the zero cross detection method and the peak detection method. Fluctuation, that is, the directivity of change is captured by analysis using the frequency components of the heartbeat, pulse wave, and APW. FIG. 6 shows the calculation procedure of these zero-cross detection method and peak detection method. FIGS. 7 (a) and 7 (b) show a calculation procedure in which the peak detection method and the zero cross detection method are applied to APW, and FIG. 8 shows the result of frequency analysis. The zero-cross detection method can detect fluctuations below the 0.5th order component, and the peak detection method can detect both the 0.5th and 1st order components. Therefore, if you want to see the fluctuation of the 0.5th-order component, use the zero-crossing detection method, and use the peak detection method together with the judgment including the heartbeat fluctuation of the primary component. That is, the zero-cross detection method is used for continuous observation, and the peak detection method is used in combination to increase the determination accuracy, and the threshold value is increased. In both detection methods, the band below the 0.5th order component is emphasized with respect to the 1st order component.

  In the zero cross method, first, when the zero cross point is obtained, it is divided every 5 seconds, for example, and the reciprocal of the time interval between the zero cross points of the time series waveform included in the 5 seconds is obtained as the individual frequency f. The average value of the individual frequencies f in the second is adopted as the value of the frequency F in the five seconds (step [1] in FIG. 9). Then, by plotting the frequency F obtained every 5 seconds, a time-series waveform of the frequency is obtained (step [2] in FIG. 9). In the peak detection method, for example, a maximum value is obtained by a smoothing differential method using Savitzky and Golay. Next, for example, the local maximum value is divided every 5 seconds, and the reciprocal of the time interval between the local maximum values of the time series waveform included in the 5 seconds (the peak on the peak side of the waveform) is obtained as the individual frequency f. The average value of f is adopted as the value of the frequency F for 5 seconds (step [1] in FIG. 9). Then, by plotting the frequency F obtained every 5 seconds, a time-series waveform of the frequency is obtained (step [2] in FIG. 9).

  The frequency gradient time series analysis calculation means 620 is obtained from the time series waveform (APW) of the frequency of the output signal of the vibration sensor of the biological signal measurement means 1 obtained by the frequency calculation means 610 using the zero cross method or the peak detection method. A time window having a predetermined time width is set, the slope of the frequency of the output signal of the vibration sensor is obtained by the least square method for each time window, and the time-series waveform is output. Specifically, first, the frequency gradient in a certain time window Tw1 is obtained by the least square method and plotted (steps [3] and [5] in FIG. 9). Next, the next time window Tw2 is set at the overlap time Tl (step [6] in FIG. 9), and the frequency gradient in this time window Tw2 is similarly obtained by the least square method and plotted. This calculation (movement calculation) is sequentially repeated, and the time-series change in the frequency gradient of the airpack signal is output as a frequency gradient time-series waveform (step [8] in FIG. 9). The time width of the time window Tw is preferably set to 180 seconds, and the overlap time Tl is preferably set to 162 seconds. As shown in Patent Document 3 (WO2005 / 092193A1) by the present applicant, this is a characteristic signal waveform from a sleep experiment in which the time width of the time window Tw and the overlap time Tl are variously changed. Is selected as the value that appears most sensitively.

  Further, as described above, the fluctuation characteristics of atrial fibrillation are switched at 0.0033 Hz, and fluctuations for adjusting the fluctuation of 0.0033 Hz are said to exist in the vicinity of 0.0033 Hz or less. Therefore, when the fluctuation state centered at 0.0017 Hz, which is between 0 Hz and 0.0033 Hz, is considered, an outline of the fluctuation state occurring at 0.0033 Hz or less appears even if variations occur. The time corresponding to a quarter period of the 0.0017 Hz waveform is 147 seconds. Assuming that the waveform is smoothed in the 90% lap state, adding 10% of the time before and after, gives 147 / 0.8 = about 180 seconds. From this point, it can be said that 180 seconds is preferable. Here, the state of change in 180 seconds and 3 minutes, the tendency of change, the differential coefficient, and the slope are captured as time series waveforms. Then, the slope, which is the average value for 180 seconds, is overlapped by 90% for the time of 180 seconds for smoothing, and plotted every 18 seconds to create a time series waveform, which is analyzed using the time series waveform of this slope. I do. This tilt time-series waveform captures fluctuations in homeostasis, and is used to globally evaluate the degree of control of heart rate variability as a time-series waveform.

  The frequency analysis unit 630 is a unit that performs frequency analysis on the frequency gradient time-series waveform obtained from the frequency gradient time-series analysis calculation unit 620 and obtains a power spectrum.

  The regression line calculation means 640 is a means for obtaining a regression line for each predetermined frequency range for the analysis waveform output from the frequency analysis means 630. The branch point identification unit 650 is a unit that identifies, as a branch point, a point at which the slope of the regression lines in adjacent frequency ranges changes by a predetermined angle or more from a plurality of regression lines obtained by the regression line calculation unit 640.

  For each predetermined frequency range in the regression line calculation means 640, for example, a regression line is obtained every 0.001 Hz or every 0.002 Hz. Or you may partition | divide like the range below 0.0033Hz, the range of 0.0033Hz-0.006Hz, the range above 0.006H. Also, for example, when obtaining every 0.002 Hz, the next regression line is obtained by overlapping the regression line obtained first at an arbitrary rate (for example, 90%) in the frequency band. Is also possible.

  When the regression line is obtained in this way, if there is a point where the slope of the regression line changes greatly, the branch point specifying means 650 specifies that point as a branch point. A large change in inclination is caused by a point where the outside angle between one adjacent regression line and the other regression line is a predetermined angle (for example, 225 degrees) or more, and between the following two straight lines between positive and negative. It can be set as a switching point. Since it is considered that a large change in the state is affected, the power spectrum rapidly changes between (0.0035−0.0017) /2=0.0009 Hz, which is regarded as a branching phenomenon in which fluctuation does not occur. It is also preferable to use a point or a point that switches between positive and negative as a branch point because a branch phenomenon that is a destructive mode can be easily specified as a branch point.

  The determination unit 660 uses at least one of the analysis waveform output from the frequency analysis unit 630, the shape of the regression line obtained by the regression line calculation unit 640, and the position of the branch point obtained by the branch point specifying unit 650. Determine the state of the person. Note that which element is in what state and what state is stored in advance in the storage unit of the computer as reference data, and the determination means 660 determines the obtained analysis waveform, the shape of the regression line, The position of the branch point is collated with reference data to determine the state of the person.

The determination unit 660 determines, for example, as follows from a test example described later.
(1) When there is no branch point and the slope of each regression line approximates 1 / f ⁿ ... Healthy and highly awake state

(2) There is one branch point consisting of the intersections of any of the regression lines, and the position is on the long cycle side with a frequency of 0.01 Hz or less, and the slope of the regression line on the short cycle side is the boundary of this branch point. When approximating 1 / f ⁿ・ ・ ・ Healthy, moderate or low arousal state

(3) There are two or more branch points formed by the intersections of any of the regression lines, and the position of one branch point is on the long period side with a frequency of 0.01 Hz or less, the one branch point and the other branch point When the logarithmic power spectral density value is more than a predetermined difference from the bifurcation point ... Drowsiness state or state with sleep onset

(4) The regression line of the analysis waveform using the central oscillation wave (APW) shows the same tendency as the above (1) and (2) up to the predetermined position on the short cycle side, and the short line from the predetermined position. While the regression line on the cycle side is close to the horizontal state, the regression line of the analysis waveform using the fingertip plethysmogram shows the same tendency as in (1) and (2) above, and the regression on the short cycle side When there is a divergence in the slope of the straight line ... Suspected arrhythmia

(5) There are four or more branch points consisting of intersections of any of the regression lines, two of which are located on the long cycle side with a frequency of 0.01 Hz or less, and the other two branch points are The frequency is 0.006 Hz or more and is located on the short period side with respect to the long-period branch point, and there is a predetermined difference or more in logarithmic power spectral density between the two long-period branch points. When a logarithmic power spectral density value is more than a predetermined difference between two branch points on the short-cycle side: disease state Hereinafter, based on a test example, the determination method in the determination unit 660 will be described in more detail.

(Test Example 1)
(Judgment test regarding the presence or absence of sleepiness)
The biological signal measuring means 1 shown in FIG. 1 is laminated on the back side of the product name “Twin Lumber” made by Delta Touring Co., Ltd., seat cushion, attached to an automobile seat installed indoors, and a subject is seated Then, a biological signal (hereinafter referred to as “heart swing wave” but sometimes abbreviated as “APW”) by swinging the atrium, ventricle and aorta in a sitting posture was collected. The plate-like foams 21 and 22 and the three-dimensional three-dimensional knitted fabric support member 15 constituting the biological signal measuring means 1 were bead foams having an average bead diameter of about 5 mm and slice-cut to a thickness of 3 mm. The three-dimensional solid knitted fabric 10 was manufactured by Sumie Textile Co., Ltd., product number: 49011D, and had a thickness of 10 mm. As the film 16, a product number “DUS605-CDR” manufactured by Seadam Co., Ltd. was used. The test subject is a healthy man in his 20s and 50s.

  The frequency fluctuation means 610 applies the zero cross (0x) method to the heart fluctuation wave obtained from the biological signal measurement means 1 to obtain a time series waveform of the frequency. Processing was performed by the time series analysis calculation means 620 to obtain a time series waveform of the frequency gradient. Next, the frequency analysis means 630 frequency-analyzes the frequency gradient time series waveform to obtain a power spectrum, takes the frequency (logarithmic value) on the horizontal axis, and the power spectrum density (logarithmic value) on the vertical axis. The displayed graphs are shown in FIGS. A thick line superimposed on the analysis waveform indicates a regression line obtained by the regression line calculation means 640. The regression line 1 is obtained in 0.0033 Hz or less, the regression line 2 is obtained in the range of 0.0033 Hz to 0.006 Hz, and the regression line 3 is obtained in each frequency region of 0.006 Hz or more. However, when the angle formed by the regression line 2 and the regression line 3 is 10 degrees or less, both are represented by a single line and are displayed as the regression line 2 in the figure (FIGS. 10 to 15). Moreover, although the intersection of each regression line (including the extension line) is a branch point, in this test example, it set so that it might plot, when those intersection angles exceed 10 degree | times. In this test example and the following test example, it is set to 10 degrees, but this is only an example, and it can be changed according to the accuracy of the required analysis result, or changed for each subject. You can also

10 to 15 are data in a state where each subject a to k does not have drowsiness (wakefulness state). From these figures, in any case, there is one branch point (between the regression lines 1 and 2), the position of this branch point exists at 0.001 to 0.005 Hz, and the branch point The regression line 2 on the shorter cycle side approximates 1 / f ⁿ in the direction. In this specification, the approximation to the horizontal state means that the slope with respect to the frequency is in the range of −0.5 to 0, and the approximation to 1 / f means that the slope with respect to the frequency is −0.5 or less and −1.5. Also, the approximation to 1 / f ² is a range that is −1.5 or less and greater than −2.5.

16 to 23 are data in a state in which each subject A to O is drowsy. From these figures, it can be seen that there are two branch points between the regression lines 1 and 2 and between the regression lines 2 and 3. And both two branch points exist in the range of 0.003-0.009 Hz, and exist in the short period side rather than the state which does not have drowsiness of FIGS. Further, there is a predetermined difference or more in log power spectral density between the two branch points. That is, there is a difference between the two regression lines 1 and 3 across the regression line 2, and the regression line 2 has an inclination close to the vertical, when compared with the arousal state of FIGS. It is a big feature. That is, it can be said that the regression line 2 is a line indicating a function deterioration accompanying the progress of fatigue while the state of the subject moves from the state of the regression line 1 to the state of the regression line 3. In addition, the regression line 3 on the short cycle side from the second branch point has a slope that is almost 1 / f in these examples. This indicates that each subject ak feels drowsy and is relatively relaxed.
From the above, by using the number of branch points of the analysis waveform, the difference between the branch points, and the slope of the regression line, it is possible to determine whether the state is awake or drowsy. As described above, in all of the test examples, healthy subjects are subjects, and there is a clear difference between subjects with illness. Details of these will be described later.

  24 (a) and 24 (b) show the time series waveform obtained by linking the data of all subjects in the sleepless state (wakeful state) and the sleepy state, respectively, and processing by the zero cross method, and performing frequency analysis. It is the graph which calculated | required the analysis waveform and the regression line. In this graph, the data obtained by connecting the data and measuring the data for 24 hours and the data for the data measured for 1 hour are displayed together.

  The fluctuations obtained by processing the data for 24 hours with the zero cross detection method and further analyzing the frequency are as follows. When there is no sleepiness in (a), the branch point is below 0.002 Hz, and the sleepiness in (b) When there is, the branch point moves to around 0.005Hz. When there is no drowsiness, it approaches 1 / f, and when there is drowsiness, more than one branch point occurs in the band of 0.006Hz or more, and two or more branch points show a branching phenomenon and are judged to be abnormal. it can. Furthermore, the inclination of the fluctuation approaches from 1 / f to almost horizontal. In addition, the fluctuations obtained by processing the data for one hour measurement with the zero cross detection method, further analyzing the slope time series, and analyzing the frequency are the same as the fluctuation curves obtained from the result of frequency analysis of the data for 24 hours. Show the trend. Therefore, the fluctuation obtained by converting the data for about 1 hour measurement into a slope time series waveform and performing frequency analysis is the presence or absence of drowsiness equivalent to the fluctuation obtained by performing frequency analysis for data for about 24 hours measurement. The possibility of being able to judge is suggested.

(Test Example 2)
(Real car running test)
As in Test Example 1, the biological signal measuring means 1 shown in FIG. 1 is laminated on the back side of the product name “Twin Lumber”, a seat cushion, manufactured by Delta Touring Co., Ltd. A vehicle running test was conducted. A healthy male subject S in his 60s drove a course simulating a highway for about 60 minutes.

  FIG. 25 is a diagram showing the results. When analyzed with a time series waveform according to the conventional index developed by the present applicant, drowsiness occurs in about 16 to 18 minutes. It can be determined that the belt is awake. According to the actual comment from the subject S, he felt light sleepiness in the first half of the experiment, but did not feel sleepiness in other time zones.

  On the other hand, FIG. 26 shows an analysis waveform according to the present invention. From this figure, there are two branch points of the regression line, and a predetermined difference occurs between the regression line up to about 0.006 Hz and the regression line after about 0.007 Hz which is the second branch point. Yes. From this, it can be determined that the function decline occurred between the two branch points, and drowsiness was caused, and the analysis result according to the present invention is effective.

  27 to 29 are diagrams showing analysis waveforms when a healthy male subject F in his 50s continuously drives a course simulating a highway. 27 shows an analysis result for 1.5 hours from 9:06, FIG. 28 shows an analysis result for 1.5 hours from 11:12, and FIG. 29 shows an analysis result for 1 hour from 13:54. Indicates.

  In both figures, there are two branch points with a boundary of around 0.006 Hz, and there is a difference between the regression line before 0.006 Hz and the regression line after that, and subject F feels sleepiness due to functional decline You can see that According to the comment from the subject F, he generally felt a relatively strong drowsiness, and in particular, he felt a strong drowsiness in the second half of driving.

Therefore, when comparing the case of the subject Mr. S in FIG. 26 who had mild sleepiness with the case of the subject F in FIG. 27 to FIG. 29 who felt stronger sleepiness, first, the long period side from around 0.006 Hz. In the case of FIG. 26, the slope of the regression line of FIG. 26 is almost the same as the horizontal state, but in the case of FIGS. From this, the slope of the regression line on the long cycle side before the first bifurcation point rises to the right as the sleepiness is felt, and as the slope falls from the horizontal to the right shoulder (as the slope approaches 1 / f ⁿ ), the sleepiness becomes higher. It can be said that it is weak. In addition, in FIG. 28 and FIG. 29, the angle of the upward slope of the regression line on the short cycle side is larger than that in the case of FIG. 27, rather than the second branch point around 0.006 Hz. I understand. In the data of the subject Mr. F, in the case of FIG. 27 with mild sleepiness, it is close to the lower right shoulder (1 / f). In both cases, the consciousness that you must keep awake because you are driving (ie consciousness of counterattack (attack) against drowsiness) works, but the slope of the regression line on the short cycle side from the second branch point is right shoulder It can be seen that the higher the rise, the stronger the counterattack consciousness corresponding to the drowsiness.

  That is, when there are two branch points and a predetermined difference exists between the regression line on the long cycle side from one branch point and the regression line on the short cycle side from the other branch point, the long cycle It becomes a determination index that sleepiness is stronger as the regression line on the side rises to the right, and a determination index that the counterattack (aggressiveness) against sleepiness is stronger as the regression line on the short cycle side rises to the right.

(Test Example 3)
(Examination by condition of the same subject)
For the male test subject FG in his 50s, logarithmic power spectral density analysis waveforms were obtained in the same manner as above for each state of overwork (sitting position), sleeping (post position), and waking up (sitting position). 34.

  In the sleep state of FIG. 30B, there are two branch points near 0.004 Hz and 0.008 Hz. The recovery point (when waking up) in FIG. The regression line on the shorter cycle side shows that the recovery time (when waking up) is closer to the horizontal, and the physical condition is still insufficient when waking up. On the other hand, it can be seen that the amplitude (0.006 Hz or more) on the short period side is larger and the degree of activity is higher when recovering (when waking up) than when sleeping. In the overworked state of FIG. 30 (a), there are a plurality of branch points, and the slope of the regression line on the short cycle side from the branch point is almost horizontal or slightly upward, compared to sleep. It can be read that it shows a strong counterattack (aggressiveness) against fatigue.

  32 to 34 are data when the subject FG in the 50s same as above caused food poisoning due to oysters, in order of food poisoning onset, at the time of medication, on the first day of medication discontinuation, on the second day of medication discontinuation The analysis waveform of the logarithmic power spectrum density of each state on the third day of discontinuation of administration is shown.

  At the onset of FIG. 32, there are two branch points, and the change in amplitude on the short cycle side is larger than that at the time of medication in FIG. In the case of FIG. 33A at the time of medication, the regression line is closer to 1 / f as a whole than in FIG. 32, but a regression line close to horizontal is observed on the short cycle side. This suggests that medications have suppressed symptoms but still survive the disease. On the other hand, after stopping the medication in FIG. 33 (b) and FIG. 34, the number of branch points becomes two or more, and the amplitude on the short cycle side is significantly larger than that in the medication in FIG. 33 (a). The regression line on the short cycle side is also close to the horizontal line or rises to the right. Therefore, it can be seen that when the medication is stopped, the counterattack (aggressiveness) against the disease becomes stronger.

Note that the amplitude disturbance is not only compared by the amplitude amplitude, but also by calculating the area surrounded by the waveform indicating the amplitude and depending on how large the area is compared to normal. It can also be determined.
Further, for example, at the time of recovery in FIG. 31, the regression line on the short cycle side is almost horizontal, but in FIG. 30 (b) before that, the regression line is descending to the right (close to 1 / f). Yes. From the accepting state (relaxed state) showing the negative fluctuation in which the fluctuation gradually attenuates when viewed in time series (in the order of FIG. 30B and FIG. 31), Although it has changed to the endured state, it does not change suddenly between the two states, but there is a state that is stiff as an intermediate state between them.
On the other hand, when moving from FIG. 33 (b) to FIG. 34 (a), the regression line that rises to the right in FIG. 33 (b) changes to a state close to horizontal in FIG. 34 (a). In this case, the state changes from a counterattack (attack) that shows aggressive fluctuations in which the fluctuation gradually spreads, to a state that endures, but as an intermediate state in this case, a state that endures the counterattack (attack) Exists.
Therefore, the intermediate state between each state can be estimated by obtaining the temporal change of the analysis waveform and the regression line according to the present invention.

(Test Example 4)
(Comparative test with electrocardiogram and fingertip volume pulse wave)
In addition to the means for collecting the heart sway wave (APW) in the same manner as described above by applying the biological signal measuring means 1 shown in FIG. 1 to the back, the living body having the same structure as above in the chest and abdomen on the front side of the subject. Signal measurement means (however, the magnitude is smaller than that of the above-mentioned biological signal measurement means 1), and biological signal data by the biological signal measurement means from a total of three locations (hereinafter collectively referred to as “body surface pulse wave”) Got. At the same time, the electrocardiogram was measured with an electrocardiograph at the time of the test, and the fingertip volume pulse wave was also measured using a fingertip volume pulse wave meter (manufactured by AMCO, finger clip probe SR-5C). The results are shown in FIGS. 35 to 39 for each subject. In each figure, (a) is an analysis waveform of a biological signal obtained from the chest, (b) is an analysis waveform of a biological signal obtained from the abdomen, and (c) is a biological signal obtained from the back (heart swinging). (D) shows the analysis waveform of the electrocardiogram data obtained from the electrocardiograph, (e) shows the analysis waveform of (a) to (d) together, (f) shows the finger The analysis waveforms of the plethysmogram data are shown respectively.

  FIG. 35 is data of a healthy male AO in his 20s. The subject AO during the experiment started outing 5 minutes after the start of the experiment and then slept for a long time, but it was shallow enough to wake up every time he made a noise.

  From FIG. 35, it can be determined that there is a predetermined difference between the branch points between the two regression lines, and drowsiness due to functional decline is occurring (including during sleep). Each data shows almost the same tendency, but there is a slight difference between the electrocardiogram data and the chest, abdomen and back data.

  FIG. 36 is also data of a healthy male U in his 20s. The state of the subject U during the experiment was awake until 10 minutes after the start of the experiment, and after that, after 20 minutes, she became light sleep.

  From FIG. 36, it can be determined that there is a predetermined difference due to functional deterioration between the branch points between the two regression lines, and drowsiness is generated (including during sleep). Each data shows almost the same tendency, but there is a slight difference between the electrocardiogram data and the chest, abdomen and back data.

  Referring to FIGS. 35 (e) and 36 (e), in the case of a healthy person in his twenties, if there is a slight divergence between the data on the chest, abdomen and back with respect to the electrocardiogram data, Since there is an individual difference in the width of the divergence, which is considered to indicate that the user is sleeping, a predetermined threshold value can be set by storing data in advance.

  FIG. 37 is data of a healthy male HO of 89 years old. Although this subject HO is also drowsy like the above, each data generally shows the same tendency. On the other hand, FIG. 38 is a test result in the state where the same subject HO felt a headache in the morning on the measurement day in the sleep time of 4 hours and a half on the day before the measurement, but no longer felt the headache at the time of measurement. In this figure as well, all tend to be similar overall, but in (e), the electrocardiogram data is slightly different from the other data. This is thought to be due to the burden on the heart due to lack of sleep due to being elderly.

  That is, as shown in FIG. 35 to FIG. 37, by comparing the body surface pulse wave data, electrocardiogram data, and fingertip volume pulse wave data with different measurement sites, Although it can be said to be healthy, if the data are inconsistent with each other and the degree of divergence is large, it is considered that an abnormality has occurred in the body.

Therefore, a test was conducted on persons with diseases.
FIG. 39 shows an 86-year-old male subject (Mr. Yoshito Fujita) who suffered from colorectal cancer, and performed lymph node excision and partial excision of the large intestine approximately one month before the measurement. At the time of measurement, he can walk in the house and has recovered to such an extent that he can sit and eat. The result of the blood test performed on the same day of the measurement is shown in FIG. 40. As long as the result of the blood test is seen, there is no data that is significantly different from the normal value range. It can be judged that the course is good and there is no special abnormality.

However, the following can be said from the measurement results shown in FIG. First, it can be seen from FIG. 39 (f) that the difference between the electrocardiogram data and other data is extremely large. This means that there is a great burden on the heart. From (c), it can be seen that there are a plurality of branch points and the amplitude on the short cycle side is remarkably large, indicating a tendency of illness. Moreover, both of the two regression lines of 0.006 Hz or more are rising to the right, indicating a strong counterattack (aggressiveness) against functional deterioration due to external stress. The chest of (a) and the abdomen of (b) also show reduced function and strong counterattack (aggressiveness) as in (c), but the abdomen is larger than the chest in the short period side amplitude, High activity. In addition, although the peripheral system that can be read from the fingertip volume pulse wave data in (e) has four or more branch points and shows signs of illness, the slope of the regression line on the short cycle side of 0.006 Hz or higher is 1 / f. Nearly, the counterattack (aggressiveness) of the peripheral system is slightly lower than that of the heart sway artery wave of (c).
On the other hand, two weeks after the day of blood test and measurement of FIG. 39, CT imaging of the trunk of the subject (Ryoto Fujita) was performed. This CT image is shown in FIG. From FIG. 41, a large amount of ascites and pleural effusion were observed, suggesting suspicion of peritoneal dissemination and pleural dissemination. Further, a tendency of cardiac hypertrophy is seen from the CT image, and it is considered that a large divergence between the electrocardiogram data and other data in FIG. 39 (f) is caused by cardiac hypertrophy.

  In the analysis result of the body surface pulse wave using the present invention, as shown in FIG. 39, pathological signs of the chest and abdomen are detected. On the other hand, when the analysis waveform of the chest of (a) and the abdomen of (b) is compared with the analysis waveform of the back of (c), the analysis waveform of the chest and abdomen is relatively approximate, The analysis waveforms are different, and the amplitude is particularly large on the short period side. This is due to stress in the chest and abdomen due to the increase of pleural and ascites, pleural dissemination, peritoneal dissemination, aortic aneurysm, etc. It is considered a thing. Therefore, it can be said that the stressed part can be identified by comparing the analysis waveform of the back part (heart swing wave) with other body surface pulse waves. That is, it can be said that the analysis result by the biological state estimation apparatus of the present invention can obtain information that cannot be obtained only by blood tests. CT imaging was performed because the physical condition of the subject (Ryoto Fujita) became unsatisfactory. In that sense, changes in the physical condition in the near future, the analysis result of the body surface pulse wave using the present invention is It can be said that it captures.

In addition, the biological state analyzing apparatus of the present invention is an apparatus for analyzing a biological signal (body surface pulse wave including heart sway artery wave) obtained non-invasively using the above-described biological signal detecting means 1. For this reason, the test subject's burden is very small, and even an elderly and sick patient can measure and analyze easily without imposing a burden. Furthermore, the degree of burden on the heart can be determined by comparing with the electrocardiogram data. This test subject was actually diagnosed as peritoneal dissemination and pleural dissemination after 8 days, and after removing 1 liter of ascites and obtaining temporary recovery, he was diagnosed with a life expectancy of 1 month. One day later, the arterial oxygen partial pressure suddenly dropped, so 3 liters of pure oxygen was sucked in, but the condition suddenly changed. A characteristic of hypoxemia is that the partial pressure of arterial oxygen does not increase due to such absorption of pure oxygen. That is, this seems to have caused oxygenation disorder. This disorder of oxygenation is thought to originate from a diffusion disorder and is presumed to have been caused by pressure from pleural effusion. In general, oxygen saturation (SpO ₂ ) is 99% in adults, and when this becomes 85% or less, oxygen does not go to the brain and death occurs. In the case of this test subject, the condition changed suddenly, and the oxygen saturation (SpO ₂ ) that was 88% immediately after removing ascites suddenly dropped, and even when 3 liters of oxygen was sucked, it reached only 91%.

  FIG. 42 shows the measurement results of a 62-year-old male subject YA who has lung cancer. There are four or more branch points in the chest, abdomen, and back (heart swing wave), and the recurrence (aggressiveness) in which the regression line on the short cycle side rises to the right is shown. That is, functional deterioration occurs in the chest, abdomen and heart, and sympathetic nerves are increased. Further, it can be seen from (f) that the deviation from the analysis result with the electrocardiogram data is relatively large, and the burden on the heart is large. Furthermore, when comparing the amplitude on the short cycle side, the data of the central system (chest, abdomen and heart) of (a) to (c) is also large, but the amplitude of the peripheral system data of (e) is larger, There are also concerns about abnormalities (diseases) in the peripheral system.

FIG. 43 shows data on the supine position of the same male subject YA, and is a measurement result about four months before FIG. Also in this data, it is understood from (a) to (c) that an abnormality of the central system is detected. It can also be seen that the degree of abnormality in the peripheral system of (e) is high. On the other hand, when the comparison with the electrocardiogram data of (f) is seen, the deviation is smaller than that in FIG. Therefore, it can be seen that an abnormality (disease) that burdens the heart has progressed in about 4 months.
Here, all of the subjects in FIGS. 39 to 43 are cancer patients. In the case of subject YA, FIG. 42 shows data after the diagnosis of cancer, while FIG. 43 shows data before the diagnosis of cancer. However, in both cases, there is a phenomenon in which the regression line on the short cycle side rises to the right. In particular, the phenomenon of the heart swinging wave rising to the right is remarkable. Therefore, when a regression line that rises to the right in this way appears at least in the heart oscillating wave, it is measured in an awake state without sleepiness, especially so as not to be confused with signs of sleepiness. When at least one, preferably two or more regression lines, appear, it is preferable to determine that “suspected of cancer”. The biological state estimation apparatus of the present embodiment is not intended to make a strict diagnosis, but is intended to roughly screen a person who is suspected of having a disease, here, a suspected cancer, by simple measurement. In this sense, it is preferable to use a determination means that can extract that the patient is suffering from some kind of illness, especially that there is a possibility of cancer, if such a phenomenon of rising to the right appears.

  FIG. 44 is data of a 73-year-old subjectively healthy male subject HO. In the back part (heart swing wave (heart)) of (c), there is one branch point, and the slope of the regression line after the branch point is close to 1 / f, but the chest part of (a) and (b In the abdomen, the number of branch points is 4 or more, and the regression line on the short cycle side has a slope close to the horizontal state or rises to the right, increasing the counterattack (aggressiveness) from the living body to the disease. It can be determined that an abnormality (disease) is seen in the abdomen. Looking at (f), it can be determined that there is a discrepancy with the electrocardiogram data and the burden on the heart is large. Further, the peripheral system of (e) also has four or more branch points, and the regression line on the short cycle side is close to the horizontal state, so it can be determined that there is a possibility of illness.

(Summary)
FIG. 45 (a) shows a typical example of fingertip volume pulse wave data in which a sleep onset sign phenomenon has occurred. This uses fingertip plethysmogram as a biometric signal. Wavelet analysis of the original waveform of the fingertip plethysmogram and the time-series waveform of the frequency slope are obtained from the original waveform of the fingertip plethysmogram. Data. The time point when the LF / HF burst waveform is generated in the wavelet analysis shown in FIG.

  FIG. 45B shows a regression line of the analysis waveform obtained by frequency analysis of the above-described frequency gradient time series waveform before and after this sleep onset signal. In this figure, at the time of awakening, the regression line has a 1 / f slope, but it can be seen that a branch point occurs in the state of sleepiness after the onset of sleep signal. In other words, in the region of less than 0.006 Hz, the slope of the lower right shoulder at the time of awakening changes to the slope of the upper right shoulder after the appearance of the sleep onset sign signal. Among them, FIG. 45C shows the transition of 0.0017 Hz, 0.0035 Hz, and 0.0053 Hz, which are the characteristic signals described above. From this figure, when drowsiness occurs, the value of the power spectrum density of the 0.0053 Hz signal (activity adjustment signal) becomes relatively high, in contrast to the power spectrum of the 0.0017 Hz signal (function adjustment signal). The density value is relatively small, and the 0.0035 Hz signal (fatigue acceptance signal) is slightly higher. Therefore, this representative case is particularly effective in determining whether sleepiness is occurring by comparing the change in the activity adjustment signal with that at awakening.

  FIG. 46 is a diagram schematically showing a case of each subject in Test Example 4 collectively. If the analysis waveform of the electrocardiogram data and the analysis waveform of the fingertip volume pulse wave are all approximate to the analysis waveform of the body surface pulse wave of the chest, abdomen, and back (heart swing wave), judge. When the analysis waveform of the electrocardiogram data is deviated from the health case, it indicates that the heart / vascular system is burdened with aging. In addition, if the waveform of the back (heart swing wave) deviates from that of the chest and abdomen, and the amplitude of the waveform of the fingertip volume pulse wave is large, Indicates that there is a heavy burden. Furthermore, in addition to the divergence of the ECG data, the analysis waveform of the body surface pulse wave of the abdomen, the body surface pulse wave of the back (heart swing wave), and the fingertip volume pulse wave are approximated, while the analysis waveform of the chest is If there is a divergence, it indicates the possibility that an abnormality has occurred in a disease near the abdomen, for example, the abdominal aorta. The degree of divergence of the electrocardiogram data and the like varies depending on the individual. For example, when the amount of divergence between the electrocardiogram data and the heart sway wave at normal time is used as a reference, for example, when the distance is 5 times or more, And so on.

FIG. 47 shows all of the above-mentioned items collectively. The analysis waveform having a slope of 1 / f shown in the center diagram is used as a reference, and when this is a healthy and awake state, 0.006 Hz. close to or regression line 1 / f ⁿ also short cycle side from, or close to the horizontal state, or soaring, or each, predetermined state (drowsiness, illness, etc.) occurring in the living body has received a It is possible to determine whether the state is endured or whether the state is counterattacked or attacked. In particular, as described above, the state of sleepiness can be detected more reliably by paying attention to the transition of signals of 0.0017 Hz, 0.0035 Hz, and 0.0053 Hz.

In addition, when there are two or more branch points and there is a predetermined difference or more between the two regression lines, this indicates that the function has deteriorated due to fatigue (sleepiness) or illness. Moreover, than 0.0053Hz short cycle side, or preferably the slope of the regression line in the short period side from 0.006Hz is close to 1 / ^{f n,} or close to the horizontal state, or soaring, the inclination angle It means that the closer to the level, the higher the right shoulder, the stronger the resistance to fatigue (drowsiness) and functional deterioration due to illness. That is, when it is close to 1 / f ⁿ , it means that the state is relatively relaxed, and when it is close to the horizontal state, it means that the state is endured. If so, it shows a stronger resistance to the state, meaning that the living body shows a counterattack or attack.

  In addition, the greater the amplitude of the analysis waveform on the shorter cycle side than 0.053 Hz, preferably the shorter cycle side than 0.006 Hz, the higher the activity level, that is, the disease indicates that it is fighting the disease. Is.

  As described above, by using the biological state estimating apparatus of the present invention, it is possible to analyze and grasp the state of a person more finely than before.

  In addition, a sleep experiment was performed to prove that the above-mentioned 0.0017 Hz (function adjustment signal), 0.0035 Hz (fatigue acceptance signal), and 0.0053 Hz (activity adjustment signal) exhibit characteristics related to the sleep onset symptom phenomenon. The results shown in FIGS. 48 to 51 were obtained. The experiment is an isolated indoor space with no traffic, 30 minutes after the start of the experiment, tolerate sleepiness and maintain wakefulness, and then leave it to the will after 30 minutes. Those who can continue to do their best can do their best, and the frequency of the heart rocking wave (APW) was analyzed, and the waveforms of the distribution ratios of the three frequency components and the state of each subject were analyzed. The distribution rate is obtained by extracting frequency components corresponding to the above-described function adjustment signal, fatigue acceptance signal, and activity adjustment signal from the time-series change of the power spectrum, and taking the sum of the power spectrum values of these three frequency components as 100. The ratio of each of the three frequency components is calculated and obtained in time series. The three frequency components were calculated by the zero cross method (0x detection method) and the peak detection method (peak detection method), respectively.

  First, in the distribution ratio waveform of at least one of the zero-cross method (0x detection method) and the peak detection method (peak detection method), the time zone in which 0.0017 Hz and 0.0035 Hz are in opposite phases, and the test subject in that time zone The state was compared. The results are shown in FIG. From FIG. 48, in the time zone in which 0.0017 Hz and 0.0035 Hz are in opposite phases, the number of subjects who are out of bed or fall asleep is relatively large.

  Further, when the distribution rate is higher than 0.0035 Hz (APW (1)) in the time zone of the opposite phase, when the distribution rate is higher than 0.0017 Hz (APW (2)). When a time zone with a high distribution rate of 0.0017 Hz and a time zone with a high distribution rate of 0.0035 Hz are continuously generated (APW (3)), the opposite phases of 0.0017 Hz and 0.0035 Hz are generated. In the case of an increase of 0.0053 Hz (APW (4): 0.0017 Hz is around 40%, 0.0035 Hz is 10% or less, and 0.0053 Hz is 50% or more) I investigated.

  As a result, as shown in FIG. 48, APW (1) was 55.6%, APW (2) was 66.7%, APW (3) was 50%, and APW (4) was 72.2%. It was. Therefore, the waveform change pattern of APW (1) is set to “weak sleepiness”, the waveform change pattern of APW (2) is set to “strong sleepiness”, and the waveform change pattern of APW (3) is set to “resist sleepiness”. The correlation data in which the waveform change pattern of the APW (4) is in the state of “pregnant sleep” is stored in the storage unit. If the waveform change pattern is specified, the determination unit 660 specifies which of the waveform change patterns corresponds to and determines the state of the person.

  Note that, as shown in FIG. 48, the three frequency components can be determined for sleepiness and sleep onset, regardless of whether the zero cross method (0x detection method) or the peak detection method (peak detection method) is used. However, since there is an individual difference in sensitivity, it may be set which one is used for each individual. It is also possible to always use both, and when both appear in opposite phases, it is possible to determine drowsiness or a predictive sleep phenomenon. However, when the biological signal measuring means of the present invention is used in a car or the like and used for snoozing detection, even if an opposite phase appears in either one, if it is determined that it is drowsiness or a sign of falling asleep, It is considered effective.

  On the other hand, FIG. 49 is a diagram showing a gradient time-series waveform pattern by the zero cross method in the awake state, and shows a comparison between the waveform and psychological judgment (whether it is awake or feels light sleepiness).

In FIG. 50, in the psychological determination (APW (+)) in FIG. 48, “Awake” is determined as “Awake”, “Outout / Drowsiness”, and “Sleep” are determined as “Sleep”. In the psychological judgment (APW (−)) in FIG. 49, the correlation between “Awake” and “Sleep” determined as “Awake” and “Slight sleepiness” is shown. is there. As a result, X ² (chi-square value) = 30.98, which is larger than 3.841 which is a value of P = 5% in the case of 1 degree of freedom, and therefore between APW (−) and APW (+). Can be said to have a significant difference.

FIG. 51 shows the number of cases of falling asleep in APW (1), (2), (3), and (4) in the psychological judgment (APW (+)) in FIG. FIG. 49 summarizes the cases of falling asleep and cases of not falling asleep in the psychological judgment (APW (−)) in FIG. 49. As a result, X ² (chi-square value) = 26.97, which is larger than 5.991 which is a value of P = 5% in the case of 2 degrees of freedom, so APW (−) and APW (1) (2 It can be said that there is a significant difference between (3) and APW (4).
From the above, a method for determining an analysis waveform and a regression line by frequency analysis of the present invention and determining a state including drowsiness and a sleep onset symptom based on the slope thereof, wavelet analysis and a sleep onset symptom by a time-series waveform of the slope 45 and FIGS. 48 to 51 are all correlated with each other as a method for determining a state including drowsiness and a sleep onset symptom phenomenon based on the distribution rates relating to the above three characteristic signals. It can be said that there is.

(Examples of patients with atrial fibrillation)
FIG. 52 shows data (FIGS. 52 (a) and (b)) obtained by analyzing a cardiac sway wave (APW) from the back by the biological signal measuring means 1 in a static sitting state for a patient with atrial fibrillation, The data (FIG. 52 (c)) which analyzed the fingertip volume pulse wave obtained from the fingertip volume pulse wave meter are shown. FIG. 52 (a) shows a waveform obtained on the short period side near 0.01 Hz or more with one regression line drawn. FIG. 52 (b) shows a more microscopic analysis and draws two regression lines. It is a thing.

The APW of (a) and (b) is measurement data for 71 minutes, and the fingertip volume pulse wave of (c) is measurement data for 62 minutes. The experiment was conducted during the meeting in which the subjects were attending. During the meeting, the subjects did not speak but maintained a high arousal state. In (a), there is a second branch point in the vicinity of 0.01 Hz, and it can be seen that the regression line on the short cycle side shows a slope close to the horizontal state. On the other hand, in (c), the regression line on the short period side after the branch point at 0.004 Hz shows a slope close to 1 / f. From this, when judging only from the analysis result of (c) of the fingertip volume pulse wave, it can only be judged that the subject is in a relaxed state. However, the analysis result of APW (a) shows that the inclination of the regression line on the short cycle side is almost horizontal, and the subject can withstand that state. Heart diseases such as atrial fibrillation do not appear in heart rate variability or finger plethysmogram, and at first glance are in a healthy state and not seen as a disease. This is due to modulation of fluctuations in the heart itself, and is a heart disease caused by so-called occult arrhythmia. That is, it is expressed in the heart and is difficult to grasp as an arrhythmia in the fingertip volume pulse wave. The reason why it is difficult to capture the characteristics of the fingertip plethysmogram is that the heart is working hard to keep the periphery normal. However, if the heart part oscillating wave (APW) is a biological signal from the back using the above-described biological signal measuring means 1, the feature of the heart motion itself can be clearly understood. Therefore, when the inclination of the analysis waveform of the fingertip plethysmogram is close to 1 / f and the short period side of the APW analysis waveform shows a tendency to be close to the horizontal state, the tendency of atrial fibrillation is indicated. Can be determined.
However, as shown in (b), when a waveform on the short period side of 0.01 Hz or more is analyzed in more detail and a regression line is drawn, a branch point is recognized between 0.001 Hz and 0.002 Hz. Two regression lines can be drawn. That is, by performing a more detailed analysis, atrial fibrillation can also be regarded as one of the disease states in which there are four or more branch points (the regression line is two or more stages on the short cycle side).

53 shows an analysis result (corresponding to FIG. 52 (a)) obtained by analyzing 71-minute measurement data of APW of the subject (Mr. W) in FIG. 52, and APW of another subject (Mr. I) having atrial fibrillation. The regression line obtained from the analysis result of 10-minute measurement data is shown. It also shows the regression line of the power spectrum obtained from the RR interval variation of the electrocardiogram for 24 hours of the conventionally known patients with atrial fibrillation (reference: Junichiro Hayano et al., Spectral characteristics). of ventricular response to atrial fibrillation, Am J Physiol Heart Circ Physiol, 273: 2811-2816, 1997).
FIG. 54 shows the regression line of the healthy subject and the regression line of the patient with atrial fibrillation obtained from the RR interval variation of the electrocardiogram shown in the above-mentioned literature, and the frequency gradient time series waveform of the present invention. The regression line of the healthy person obtained by analysis and the above-mentioned regression line of Mr. W and Mr. I are shown together.
From FIGS. 53 and 54, if the subject is healthy, the analysis waveform of the electrocardiogram data and the analysis waveform obtained from the frequency gradient time-series waveform both change with a slope of approximately 1 / f. In both cases, the inclination is almost horizontal on the short cycle side from 0.0033 Hz. This shows a state withstanding external stress. From these facts, the method of obtaining the regression line by analyzing the frequency gradient time-series waveform of the present invention can obtain the same determination result as the method using the analysis waveform of the conventional electrocardiogram data, and detects atrial fibrillation. It can be said that it is appropriate as a means.

(Target evaluation test)
The test subject was laid on an anti-vibration stand for an ambulance and traveled for 30 to 40 minutes to obtain the above-described analysis waveform and regression line. The test subjects were a healthy male subject U in his twenties and a healthy male subject KO in his thirties, and the results are shown in FIGS. 55 and 56. The anti-vibration stand for ambulances is a type (NA-III) for vibration isolation with two axes, top and bottom and front and rear, a type with NA-III top and bottom and front and rear vibration isolation functions locked (NA-III lock), The three types (NA-IV) of the type that operates in this manner and is isolated were prepared and mounted on a one-box type vehicle for running.

  From FIG. 55, the subject Mr. U has one branch point and the slope of the regression line on the short cycle side approximates 1 / f with respect to NA-III and NA-IV. However, the slope of NA-IV is closer to 1 / f, and NA-III is slightly closer to the horizontal state. Therefore, although both of the subjects U feel comfortable about these ambulance anti-vibration stands, it can be said that NA-IV feels more comfortable. On the other hand, in the NA-III lock, the subject Mr. U shows that there is a large difference between the regression line on the long cycle side and the regression line on the short cycle side, resulting in functional deterioration. In addition, the slope of the regression line on the short cycle side approximates a horizontal state, and it can be seen that the consciousness to endure patiently with unpleasant situations is working strongly.

  In the case of the subject KO, from FIG. 56, in any case, there are two or more branch points, and there is a difference between the regression line on the long cycle side and the regression line on the short cycle side, so drowsiness occurs. Yes. However, among these, for NA-IV, the slope of the regression line on the short cycle side approximates to 1 / f, and the amplitude on the short cycle side is more stable than 0.006 Hz. Therefore, it can be determined that the subject feels NA-IV most comfortable. As for NA-III, the slope of the regression line on the short cycle side approximates to 1 / f and is relaxed, but resists stress due to disturbance in the amplitude of the analysis waveform on the short cycle side than 0.006 Hz. You can see that On the other hand, in the NA-III lock, there are two regression lines on the short cycle side from 0.006 Hz that rise to the right and those that approximate the horizontal state, and counterattack (attack) against unpleasant situations It can be seen that the consciousness to act or endure is working strongly.

  From these, the subject U feels NA-IV most comfortable, and then feels NA-III comfortable. Subject KO also feels NA-IV most comfortable and NA-III comfortable. Accordingly, it can be said that the present invention can evaluate the difference in feeling or preference that a person feels comfortable.

  FIG. 57 shows the results of a sleep experiment in the 60-minute sitting posture of the subject U. (A) is the result when the patient is in a free state for 60 minutes (the state in which the subject is free to leave or sleep), (b) drinks the energy drink after feeling tired, and then This is a result in a case where the user is in a free state for 60 minutes, and (c) is a result in a case where the first half awake state is maintained and the second half is in a free state.

  In (a), drowsiness due to fatigue occurs, four branch points occur, and the two regression lines on the short cycle side from 0.006 Hz are both nearly horizontal, indicating that they are trying to endure drowsiness. . On the other hand, in (b), there is one branch point and the slope of the regression line on the short cycle side is close to 1 / f. This shows that the subject Mr. U has good efficacy of the energy drink and quickly recovered from fatigue. In (c), it turns out that there were two branch points and it was a drowsiness state. From this, it can be seen that the effect of taking a medicine such as an energy drink can be confirmed by using the present invention. This point is the same tendency as the data at the time of medication and when the medication was discontinued in FIGS. 33 (a) and 33 (b), and it can be said that the present invention is useful for confirming the effect of taking the drug. . However, in FIG. 57 (c), when the short period side of 0.006 Hz or higher is analyzed in more detail, two regression lines can be drawn in the range of 0.006 to 0.01 Hz and higher range. . In such an analysis, Mr. U who looks healthy at first glance is determined to have the same signs as the atrial fibrillation described above. In fact, Mr. U said he had subjective symptoms of atrial fibrillation.

  FIG. 58 is a diagram summarizing the above description for easier understanding. As shown in this figure, the biological state estimating apparatus of the present invention first determines whether health and illness are roughly divided. And in a healthy state, each state of awakening, sleepiness, sleep onset symptom, and sleep is determined. Among illnesses, screening for suspicion of cancer, atrial fibrillation, and cardiac hypertrophy. In FIG. 58, each typical regression line is displayed at the bottom of each state. As can be seen from this figure, the state estimation technique of the present invention shows a different regression line in each state. Therefore, the state estimation can be performed more accurately.

  The present invention is not limited to the case where biological signal measuring means is disposed on a vehicle seat such as an automobile to estimate the state of occupant sleepiness, etc., but the biological signal measurement is performed on a chair, office chair, etc. It can also be applied to state estimation by arranging means. In addition, biosignal measurement means is placed on bedding such as a bed in a hospital or a nursing facility, and the heart swing wave at the back is captured and analyzed by the above-described biosignal measurement device to apply to human state estimation You can also Thereby, it is possible to easily grasp the health status of a sleeping person (especially a sick person or a person who needs care) from the screen displayed on the monitor of the display means. In addition to heart rocking waves, the biological signal measuring means is also brought into contact with the chest and abdomen, and body surface pulse waves from the chest and abdomen are collected and analyzed together with the heart rocking waves in the back. Can be used to determine the above-mentioned diseases. In addition, more accurate determination can be made by comparing the electrocardiogram data and fingertip volume pulse wave data together.

  In addition, the present invention is not limited to humans, and can be used for judging physical condition and disease using a biological state estimation device by contacting a biological signal measuring means with the body surface of an animal such as a thermostat.

DESCRIPTION OF SYMBOLS 1 Biosignal measuring means 10 3D solid knitted fabric 15 3D solid knitted fabric support member 15a Arrangement through-hole 16 Film 21, 22 Plate-like foam 30 Vibration sensor 100 Sheet 110 Seat back frame 120 Skin 60 60 Living body state estimation apparatus 610 Frequency calculation Means 620 Frequency slope time series analysis calculation means 630 Frequency analysis means 640 Regression line calculation means 650 Branch point identification means 660 Determination means

Claims (30)

A biological state estimation device that estimates a biological state using a biological signal collected by a biological signal measuring means,
A frequency calculating means for obtaining a time series waveform of a frequency from a time series waveform of a biological signal at a predetermined measurement time obtained by the biological signal measuring means;
In the time-series waveform of the frequency of the biological signal obtained by the frequency calculation means, movement calculation is performed for obtaining a slope of the frequency for each predetermined time window set with a predetermined overlap time, and obtained for each time window. A frequency gradient time series analysis calculating means for outputting a time series change of the frequency gradient as a frequency gradient time series waveform;
Frequency analysis means for performing frequency analysis on the frequency slope time series waveform obtained from the frequency slope time series analysis calculation means and outputting an analysis waveform indicating the relationship between logarithmic power spectral density and logarithmic frequency;
For the analysis waveform output by the frequency analysis means, a regression line calculation means for obtaining a regression line for each predetermined frequency range;
A branch point specifying means for specifying, as a branch point, a point at which the slope of the regression lines in adjacent frequency ranges changes by a predetermined angle or more from a plurality of regression lines obtained by the regression line calculation means;
The position of the branch point specified by the branch point specifying means, the shape of each regression line obtained by the regression line calculation means, and the magnitude of the amplitude of the analysis waveform of the logarithmic power spectral density obtained by the frequency analysis means And a determination unit that determines the state of the living body using one or more elements.   The determination means includes the analysis of the position of the branch point specified by the branch point specifying means, the shape of each regression line obtained by the regression line calculation means, and the logarithmic power spectral density obtained by the frequency analysis means. The biological state estimation apparatus according to claim 1, wherein the state of one or more elements of the amplitude of the waveform is compared with the awakening state of the elements and the determination time point, and the state is determined based on the relative change. The determination means has no branch point and the slope of each regression line approximates 1 / f ⁿ , or one branch point consisting of the intersections of any of the regression lines, When the position is on the long cycle side with a frequency of 0.01 Hz or less and the slope of the regression line on the short cycle side approximates 1 / f ⁿ with this branch point as the boundary, it is determined that the reference is healthy and awake The biological state estimation device according to claim 2.   The determination means has two or more branch points consisting of intersections of any one of the regression lines, and the position of one branch point is on the long period side with a frequency of 0.01 Hz or less, and the one branch point The living body state estimation apparatus according to claim 2 or 3, wherein a drowsiness state is determined when a logarithmic power spectral density value is greater than or equal to a predetermined value between the first branch point and the other branch point. Depending on whether the slope of the regression line on the long-period side approximates 1 / f ⁿ , is close to a horizontal state, or rises to the right, in a state of accepting, enduring, The biological state estimation apparatus according to claim 4, wherein the biological state estimation apparatus determines that the state is a counterattack. The state of accepting sleepiness in turn depending on whether the slope of the regression line closer to the short period than the other branch point is close to 1 / f ⁿ , close to a horizontal state, or rising to the right, The living body state estimation apparatus according to claim 4, wherein the biological state estimation apparatus determines that the state is endured or counterattacked.   The determination means further reduces the logarithmic power spectral density value corresponding to the frequency of the function adjustment signal in the regression line on the long period side from the branch point of 0.01 Hz or less, compared to when awakening, Means for determining a sleep onset symptom when the value of the logarithmic power spectral density corresponding to the frequency of the activity adjustment signal rises and is higher than the value of the logarithmic power spectral density corresponding to the frequency of the function adjustment signal The biological state estimation apparatus of any one of Claims 1-6. The determination means has four or more branch points composed of intersections of any of the regression lines, two of which are located on the long period side with a frequency of 0.01 Hz or less, and the other two The branch point is at a frequency of 0.006 Hz or more and is located on the short cycle side from the branch point on the long cycle side,
There is a difference greater than or equal to a predetermined value in terms of logarithmic power spectral density between the two branch points on the long period side,
The biological condition estimation apparatus according to claim 1, wherein a disease state is determined when there is a difference greater than or equal to a predetermined value in logarithmic power spectral density between the two branch points on the short cycle side. The slope of the regression line connecting the other branch point on the long cycle side and one branch point on the short cycle side and the regression line on the short cycle side further than the other branch point on the short cycle side is 1 / f ⁿ 9. The method according to claim 8, wherein the stress, which is associated with the disease, is judged as being in an accepting state, a tolerating state, or a countering state in order depending on whether it is approximate, close to a horizontal state, or rises to the right. Biological state estimation device.   Furthermore, the said determination means is a biological condition estimation apparatus of Claim 1 which determines an active state by the magnitude | size of the amplitude of the said analysis waveform in the short cycle side more than 0.006Hz.   The determination means determines that cancer is suspected when at least one of the regression lines set on the basis of the short-period branch point located at a frequency of 0.006 Hz or more is rising. The biological state estimation apparatus according to any one of claims 7 to 9, which is determined to be present.   The determination means is a regression line set based on the short-period branch point located at a frequency of 0.006 Hz or higher in the analysis process of the biological signal measured in the awake state using the biological signal measuring means. The biological condition estimation apparatus according to claim 11, wherein when there are at least two regression lines rising upward, it is determined that there is a suspicion of cancer.   The biological signal measuring means is brought into contact with the back portion, and using the biological signal obtained from the back portion, the frequency calculating means, the frequency inclination time series analysis calculating means, the frequency analyzing means, the regression line calculating means, the branch point specifying means, and The biological state estimation apparatus according to claim 1, wherein the biological state is estimated by the determination unit.   The biological signal measuring means is further brought into contact with at least one part of the abdomen and chest, and the biological waveform obtained from at least one part of the abdomen and chest is compared with the analytical waveform obtained from the back part to determine the state of the biological body The biological state estimation device according to claim 12, wherein   The determination means further compares an analysis waveform obtained by frequency analysis of electrocardiogram data obtained from an electrocardiograph with an analysis waveform obtained from the back, abdomen or chest, and between these analysis waveforms, a logarithm The biological state estimation apparatus according to any one of claims 1 to 14, further comprising means for determining whether there is a predetermined difference or more in the value of power spectral density.   The determination means is a case where the analysis waveform of the waveform of the electrocardiogram is high, the analysis waveform based on a biological signal collected from the analysis waveform obtained from the back, abdomen or chest is low, and there is a predetermined difference or more between the two. The biological state estimation device according to claim 15, wherein the sleep state is determined when the difference is within a predetermined range.   The determination means further compares the analysis waveform obtained by frequency analysis of fingertip volume pulse wave data, compares the analysis waveform based on the fingertip volume pulse wave, and a biological signal obtained from the back, abdomen or chest The living body state estimation apparatus according to any one of claims 1 to 16, further comprising means for determining whether or not there is a predetermined difference or more in logarithmic power spectral density value between the analysis waveform and the analysis waveform.   The determination means further compares the analysis waveform obtained by frequency analysis of fingertip volume pulse wave data, and compares the stress caused by the disease or the amplitude of the analysis waveform based on the fingertip volume pulse wave. The living body state estimation apparatus according to claim 17, wherein the degree of sleepiness is determined in a receiving state, a enduring state, or a counterattack state.   The determination means has a high analysis waveform of the waveform of the electrocardiogram, a low analysis waveform based on a biological signal collected from an analysis waveform obtained from the back, abdomen, or chest, and there is a predetermined difference or more between them. The biological state estimation apparatus according to any one of claims 15 to 18, which determines that there is a cardiac hypertrophy tendency. The determination means further compares the analysis waveform obtained by frequency analysis of fingertip volume pulse wave data, compares the analysis waveform based on the fingertip volume pulse wave, and the analysis waveform of the biological signal obtained from the back The slope of the regression line on the short cycle side of the analysis waveform of the fingertip plethysmogram approximates 1 / f ⁿ , whereas the slope of the analysis waveform of the biological signal obtained from the back is horizontal The living body state estimation apparatus according to any one of claims 1 to 19, wherein the biological state estimation apparatus determines that it has a sign of atrial fibrillation when close to. A computer program set in a biological state estimation device that estimates a biological state using a biological signal collected by a biological signal measuring means,
A frequency calculation step for obtaining a time series waveform of a frequency from a time series waveform of a biological signal at a predetermined measurement time obtained by the biological signal measuring means;
In the time-series waveform of the frequency of the biological signal obtained by the frequency calculating step, the movement calculation for obtaining the slope of the frequency is performed for each predetermined time window set with a predetermined overlap time, and is obtained for each time window. A frequency slope time series analysis step of outputting the time series change of the slope of the frequency as a frequency slope time series waveform;
A frequency analysis step for performing a frequency analysis on the frequency gradient time series waveform obtained from the frequency gradient time series analysis calculation step and outputting an analysis waveform indicating a relationship between the logarithmic power spectrum density and the logarithmic frequency;
For the analysis waveform output by the frequency analysis step, a regression line calculation step for obtaining a regression line for each predetermined frequency range;
A branch point specifying step for specifying a point where the slope of the regression lines in the adjacent frequency range changes more than a predetermined angle as a branch point from a plurality of regression lines obtained by the regression line calculation step;
The position of the branch point specified by the branch point specifying step, the shape of each regression line determined by the regression line calculation step, and the amplitude of the analysis waveform of the logarithmic power spectral density determined by the frequency analysis means A computer program that causes a computer to execute a determination step of determining a state of a living body using one or more elements.   The determination step includes the analysis of the position of the branch point specified by the branch point specifying step, the shape of each regression line obtained by the regression line calculation step, and the logarithmic power spectral density obtained by the frequency analysis step. The computer program according to claim 21, wherein the state of one or more elements of the amplitude of the waveform is compared with the arousal state of those elements and the determination time point, and the state is determined based on the relative change. In the determination step, when there is no branch point and the slope of each regression line is approximated to 1 / f ⁿ , or there is one branch point formed by the intersection of any one of the regression lines, When the position is on the long cycle side with a frequency of 0.01 Hz or less and the slope of the regression line on the short cycle side approximates 1 / f ⁿ with this branch point as the boundary, it is determined that the reference is healthy and awake The computer program according to claim 22.   In the determination step, there are two or more branch points composed of intersections of any one of the regression lines, and the position of one branch point is on the long period side with a frequency of 0.01 Hz or less, and the one branch point 24. The computer program according to claim 22 or 23, wherein a drowsiness state is determined when there is a difference greater than or equal to a predetermined value in logarithmic power spectral density between the first branch point and the other branch point. Depending on whether the slope of the regression line on the long-period side approximates 1 / f ⁿ , is close to a horizontal state, or rises to the right, in a state of accepting, enduring, The computer program according to claim 24, wherein the computer program is determined to be counterattacking. The state of accepting sleepiness in turn depending on whether the slope of the regression line closer to the short period than the other branch point is close to 1 / f ⁿ , close to a horizontal state, or rising to the right, 25. The computer program according to claim 24, wherein the computer program is determined to be in a state of being endured or countering.   In the determination step, the value of the logarithmic power spectral density corresponding to the frequency of the function adjustment signal is further reduced in the regression line on the longer period side than the branch point of 0.01 Hz or less, compared to when awakening, A step of determining a sleep-onset phenomenon when the value of the logarithmic power spectral density corresponding to the frequency of the activity adjustment signal is higher than the value of the logarithmic power spectral density corresponding to the frequency of the function adjustment signal. The computer program according to any one of claims 21 to 26. In the determination step, there are four or more branch points composed of intersections of any of the regression lines, two of which are located on the long period side with a frequency of 0.01 Hz or less, and the other two The branch point is at a frequency of 0.006 Hz or more and is located on the short cycle side from the branch point on the long cycle side,
There is a difference greater than or equal to a predetermined value in terms of logarithmic power spectral density between the two branch points on the long period side,
The computer program according to claim 21, wherein a disease state is determined when there is a difference of a predetermined value or more in logarithmic power spectral density between the two branch points on the short cycle side. The slope of the regression line connecting the other branch point on the long cycle side and one branch point on the short cycle side and the regression line on the short cycle side further than the other branch point on the short cycle side is 1 / f ⁿ 29. The state according to claim 28, wherein the stress, which is related to the disease, is determined as being in an accepting state, a tolerating state, or a countering state in order according to whether it is approximate, close to a horizontal state, or rises to the right. Computer program.   The computer program according to claim 21, wherein the determination step determines an active state based on an amplitude of the analysis waveform on a short cycle side of 0.006 Hz or more.