JP4868514B2 - Apparatus and method for outputting result of estimation of biological state - Google Patents

Description

  The present invention relates to an apparatus and a method for outputting a result of estimating a sleep state as a biological state.

  Patent Document 1 describes a system for accurately determining the sleep state of a subject including mid-wakening. This system includes a body motion detection unit, a heartbeat or pulse detection unit, a determination unit that processes outputs obtained therefrom to determine midway arousal and sleep state, and a unit that outputs a determination result. Yes.

  Patent Document 2 describes a movable bed that supports a caregiver who is sleeping without feeling mental stress so that the user can turn over in a natural state. For this purpose, it is described that biological information such as the brain wave, heart rate, respiratory rate, eye movement, body motion number, etc. of the cared person is detected. In addition, it is described that using a sheet-like biological information detection sensor (sheet-like sensor) because wearing a sensor such as an electrode on a cared person to obtain such biological information has a physical sense of restraint. ing. This sheet-like sensor has a sheet-like insulator arranged opposite to each other, and its capacitance fluctuates with the heartbeat or breathing movement of the cared person. The signal (biological information signal) obtained by this sheet-like sensor includes frequency components of heartbeat and respiration. Furthermore, it is described that a biological information signal is analyzed to obtain a heart rate, a respiratory rate, and a body motion number, and a sleep depth is estimated from the biological information.

  In Patent Documents 3 and 4, voltage fluctuations based on the respiratory motion of the human body are measured at regular intervals, and the positive peak value of the voltage and the interval (time) between adjacent peaks are calculated from the measurement results. It is described that a sleep state is estimated by obtaining an average value of peak intervals, a value based on dispersion of peak intervals, and the like from the obtained peak values and peak intervals. Conventionally, there is a sleep polygraph (polysonograph, PSG) method that detects brain waves, eye movements, jaw electromyography, etc., and determines the sleep depth from the detected waveform as a method for detecting a sleep state change. The equipment is large, unsuitable for daily use, and requires qualified personnel. On the other hand, the method of estimating sleep depth with emphasis on respiratory rate, heart rate, body movement information, especially fluctuation in heart rate can obtain biological information without restraining the human body, but compared with polysomnogram It is described that the accuracy is quite bad.

  In Patent Document 5, whether or not arousal is determined based on the peak interval and peak value ratio of the respiratory motion waveform of a human, and whether the waveform is an average value of the peak-to-peak area or whether the sleep is deep or shallow due to dispersion And controlling the temperature in the bed, which is a sleeping environment, so that the human body temperature adjustment functions effectively according to the determined sleep state.

  Patent Document 6 describes that the environmental state in the bedroom and the physiological state of the sleeping person are detected, the health state of the sleeping person is detected, and the sleeping state is made comfortable. The physiological state detecting means detects a sleeping state of the sleeping person, such as a brain wave, heartbeat, breathing, body movement, skin temperature, myoelectric potential, blood pressure, sweating, skin potential, and snoring. The environmental condition detection means detects the temperature, humidity, wind speed, radiant heat, etc. in the bedroom or the bed, controls the air conditioner, the environment control means in the bed, and generates a sedative scent from the fragrance generator. Is described.

In Patent Document 7, a change in load corresponding to body movement due to sleep of a sleeper is generated as a respiratory signal, and the sleeper's apnea or hypopnea is determined based on a change in the frequency of the respiratory signal. It is described.
JP 2002-34955 A JP 2004-121837 A JP-A-2005-118151 JP 2006-20810 A JP 2006-198023 A JP 7-328079 A JP 2004-24684 A

  A highly accurate method for determining the sleep depth is a PSG method that detects an electroencephalogram and uses the detected waveform. On the other hand, in contrast to the PSG method, the method of trying to determine the sleep state using a sheet-like sensor that does not give a sleeping feeling to the sleeper is the peak value of the respiratory signal in addition to the respiratory rate, heart rate, and body movement information. In order to improve the accuracy, the peak interval, the average value of the peak interval, the variance of the peak interval, and the like are taken into consideration. However, the accuracy of estimating the sleep state is not high with respect to the determination result of the PSG method.

  One embodiment of the present invention is an analysis apparatus that outputs a result of estimating a biological state. The analysis apparatus includes a means for acquiring a respiratory signal including a plurality of respiratory peaks each including an exhalation part and an inspiration part, and a means for extracting and recording the expiration time of the expiration part included in each of the plurality of respiratory peaks in a memory And a means for including a plurality of expiration time changes recorded in the memory as a first element corresponding to an increase or decrease in the intensity of the low frequency component of the electroencephalogram as a determination element, and outputting the estimation result Means.

  Each of a plurality of respiration peaks included in the respiration signal is composed of an inhalation part for inhaling and an exhalation part for exhaling. Among these, the time of the expiration part, that is, the increase and decrease of the expiration time, and the low frequency component of the brain wave during sleep, for example, the amplitude component of the spectrum of 3 Hz or less (especially 0.5 Hz to 3.0 Hz) called δ wave The point of view considered to have a correlation between the increase and decrease of the sum of the values obtained. Therefore, by extracting the expiration time from the respiratory signal and determining the sleep state, an element corresponding to a low frequency component of an electroencephalogram, particularly an electroencephalogram that is affected by the sleep state, can be included in the determination element. For this reason, since the waveform component of the electroencephalogram or a component corresponding thereto can be added to the determination of the sleep stage without directly detecting the electroencephalogram, it does not give the sleeping person a sense of restraint with respect to the PSG method, for example, a sheet The sleep state can be accurately determined using the state sensor.

  Preferably, the estimating means statistically processes a plurality of expiration times recorded in the memory, and an increase or decrease in the expiration time subjected to the statistical processing is a first factor. The first factor is the increase or decrease of the sum or average of a plurality of expiration times included in a predetermined number of respiratory peaks, or the average increase or decrease of a plurality of expiration times included in a plurality of respiratory peaks included in a predetermined time interval. It is preferable that Noise due to components such as body motion other than respiration included in the respiration signal can be removed. Moreover, it is known from past experiments that the change in the sleep depth is repeated at a cycle of about 90 minutes, and the fluctuation of the electroencephalogram does not change greatly in about one breath. Therefore, by obtaining the average of the expiration time of the respiration peak repeated within a few minutes, for example, about 5 minutes, there is a possibility that a judgment element having a better correlation with the low frequency component of the electroencephalogram may be obtained. .

  Another aspect of the present invention includes the above analysis device, a seat type sensor capable of detecting a load change of a user in a lying state, and a device that generates a respiratory signal from the output signal of the sensor. A living body monitoring system. It is possible to provide a system that can estimate a sleep stage without tying a sleeping person to a sensor and applying almost no burden. A sheet-type sensor includes a plurality of pressure-sensitive elements assembled on a sheet-like support member.

  Still another aspect of the present invention is an environment control system that includes the analysis device described above and a device that controls at least a part of the living environment based on the output of the analysis device. The sleep stage or sleep state is estimated, and the temperature, scent, brightness, etc. of the bedroom are controlled based on the result, and more comfortable sleep and awakening can be provided by controlling the bed.

Still another aspect of the present invention is a method for outputting a result of estimating a biological state. The method includes the following steps.
a1. Obtaining a respiratory signal including a plurality of respiratory peaks each including an expiratory portion and an inspiratory portion;
a2. Extracting the expiration time of the expiration part included in each of a plurality of respiratory peaks and recording it in a memory.
a3. A sleep state is estimated by including a plurality of expiration time changes recorded in a memory per predetermined time as a first element corresponding to an increase or decrease in intensity of a low frequency component of an electroencephalogram.
a4. Output the estimated result.

  By recording the extracted expiration time in the memory, the increase / decrease can be determined using a plurality of expiration times recorded in the memory per predetermined time, and the sleep state is estimated using the increase / decrease as a first factor. be able to. Therefore, the sleep stage or sleep state of the sleeper can be estimated from the expiration time in a pipeline manner, and the result can be obtained almost in real time. Preferably, the estimating step includes statistically processing a plurality of expiration times recorded in the memory, and setting the increase / decrease of the expiration time subjected to the statistical processing as a first element.

  FIG. 1 shows an example of a bedroom home system. The home system 50 includes a biological information detection unit 10 including a sensor sheet 2 installed on a bed or a futon in a bedroom, and a control unit 20 for the bedroom. The control unit 20 is connected to the home LAN 60, and can control the environment in the bedroom by controlling the air conditioner 61, the light 62, and the aroma device 63 of the bedroom connected to the LAN 60. Therefore, the home system 50 has a function as an environment control system. Further, the control unit 20 sends information detected by the biological information detection unit 10 through the home LAN 60 and a result of analyzing the information to the monitoring unit 58. It is also possible to send to the external monitoring unit 59 via an external network connected to the home LAN 60 via the gateway 66, for example, the Internet 65. Therefore, the home system 50 has a function as a living body monitoring system.

  The biological information detection unit 10 uses a pressure sensor (pressure sensor) 7 as a pressure sensitive element, and a sensor sheet 2 in which a plurality of pressure sensitive elements 7 are arranged in an array and signals from the plurality of pressure sensitive elements 7 are received. And a data processing unit 3 that collects and sends the data to the control unit 20. The sensor sheet 2 is composed of a plurality of sub-sheets 2a, 2b and 2c. Each of the sub-sheets 2a, 2b and 2c uses a thin plastic sheet 4 as a base material. The sensor sheet 2 is configured by attaching (assembling) a plurality of pressure-sensitive elements 7 to each sheet 4 on a sheet-like support member to constitute a sheet-type sensor, and the plurality of pressure-sensitive elements 7 are arranged at appropriate intervals. Are arranged regularly.

  The sheet 4 is also provided with wiring 8 for taking out signals from the plurality of pressure sensitive elements 7. Therefore, by placing the sensor sheet 2 on a bed or the like, a large number of pressure sensitive elements 7 can be arranged on the bed. For this reason, the signal (load signal) from the pressure-sensitive element 7 is a change in load applied to the bedding of the body movement of the sleeping person 9 without attaching a sensor or electrode directly to the subject (sleeping person) 9 lying on the bed. Can be converted to For this reason, by analyzing the load signal from the pressure-sensitive element 7, it is possible to monitor the sleeping state of the sleeping person (subject) 9 and other states.

  The data processing unit 3 generates a respiration signal from the load signal. For example, Patent Document 7 discloses that a body associated with respiration is subjected to fast Fourier transform (FFT) for frequency analysis of a signal from each pressure-sensitive element 7 and the power spectrum in the respiratory frequency component (δ wave component). It describes that a plurality of pressure-sensitive elements 7 that detect a load change according to movement are extracted. Further, it is described that a breathing curve (breathing signal) 39 is generated by adding signals that fall within a predetermined phase difference using the pressure sensitive element 7 having the largest power spectrum among them as a reference element.

  The control unit 20 can be configured using an appropriate hardware resource, for example, a computer including a memory (including a semiconductor memory such as a register and a RAM, a hard disk), a CPU, a display, and various interfaces. The control unit 20 includes a function as an analysis unit 28 that estimates a sleep state. In addition, the control unit 20 includes a function as the environment control unit 30 that outputs a control signal to the air conditioner 61, the light 62, the fragrance 63, and the like in the bedroom based on the estimated result via the home LAN 60. Furthermore, the control unit 20 includes a display 29 for displaying and outputting the estimated result and the environmental control status.

  The analysis unit 28 receives the respiratory signal 39 from the information processing unit 3 of the biological information detection system 10, extracts the expiration time from the input interface 23 stored in the memory 25 and the respiratory signal stored in the memory 25, and stores it in the memory 25. The first analysis function 21 for storing, the second analysis function 22 for estimating the sleep state by including the expiration time as a determination element, and the output interface 24 for outputting the estimated result are included. The second analysis function 22 further includes a function 26 for statistically processing the expiration time stored in the memory 25, and a first increase / decrease corresponding to the increase / decrease in the intensity of the low frequency component of the electroencephalogram. And a function 27 for estimating the sleep state by including it as a determination element. The output interface 24 outputs the estimated sleep state to the environment control unit 30 and sends it to the monitoring units 58 and / or 59 via the home LAN 60.

  FIG. 2 is a flowchart showing the processing in the analysis unit 28. In step 71, a respiration signal 39 including a plurality of respiration peaks is acquired by the input interface 23 and stored in the memory 25. As shown in FIG. 3A, the respiration signal 39 includes a plurality of peaks (respiration peaks) 38 indicating respiration. Breathing consists of inspiration and expiration. Therefore, each respiration peak 38 has an inhalation portion 36 and an exhalation portion 37, and the respiration time tb is the sum of the inspiration time ti and the exhalation time te. That is, the respiration peak 38 includes a minimum-maximum-minimum. When the minimum-maximum-minimum is defined as one cycle (respiration cycle, respiration curve), a plurality of cycles are repeated during respiration. Therefore, the respiration signal 39 is a signal having a plurality of respiration cycles.

In step 72, the first analysis function 21 extracts the expiration time te of the expiration portion 37 included in each of the respiratory peaks 38 and records it in the memory 25. In the respiration signal 39, the switching points between the inspiratory portion 36 and the expiratory portion 37 are a maximum point and a minimum point, and the position can be obtained from waveform differentiation. In this process, when the expiration time te is extracted, the breath peak 38 exceeding a predetermined amplitude is cut as body movement noise. It also cuts shoulder noise. The shoulder noise is such that a small change as shown in FIG. Shoulder noise can be removed by deleting vertices that satisfy the following conditions.
(MX _i −MN _i−1 ) + (MX _{i + 1} −MN _i ) <Cs (MX _{i + 1} −MN _i−1 )
Deletes vertices MX _i and MN _i ,
(MX _i −MN _i ) + (MX _{i + 1} −MN _{i + 1} ) <Cs (MX _i −MN _{i + 1} )
Delete vertices MX _{i + 1} and MN _i
However, Cs is a threshold value, for example, 1.2.

  There are several methods for discriminating between the exhalation part 37 and the inhalation part 36. As a method depending on the detection principle, for example, in the case of a thermistor type nasal airflow sensor, the temperature rises at the time of exhalation. Therefore, the exhalation portion 37 and the inhalation portion 36 can be determined by acquiring temperature data together from the sensor side. . As a general method, the expiratory portion 37 and the inspiratory portion 36 can be determined from an empirical rule that the expiratory time is longer than the inspiratory time in the adjacent expiratory and inspiratory air due to the action of the autonomic nerve. Then, the expiration time te is calculated from the difference between the time at the maximum point and the time at the minimum point of the respiratory signal 39 and stored in the memory 25.

  In step 73, the average value in units of 5 minutes of the expiration time te stored in the memory 25 is calculated by the function 26 for statistical processing of the second analysis function 22. The average value in units of 5 minutes indicates that the expiration time te at a certain time t0 is obtained as an average value of a plurality of expiration times te obtained 5 minutes before that time t0. It does not mean that only significant data can be obtained. The averaging time is not limited to 5 minutes. It is sufficient that the time is sufficient to remove noise. In the PSG method, EEG data is subjected to frequency analysis every 5 minutes. For this reason, calculating the average value for 5 minutes is meaningful in comparison with the PSG method.

In step 74, the estimating function 27 estimates a sleep state with a step value based on the obtained 5-minute average value tea of the expiration time te. For estimating the sleep stage, it is possible to use not only the expiration time te but also the parameters disclosed in the above patent documents. The determination of the expiration time tea includes a method using an average value itself and a method using a standardized numerical value. In this specification, standardization means that the i-th value tea (i) in units of 5 minutes is processed (converted) by the following formula (1) using an average value A and a standard deviation B for a certain time.
(Tea (i) -A) / B (1)

  The normal (average) value (reference value) during sleep of the value thus standardized is zero. The conversion that has been standardized is to use the average value A and standard deviation B of one sleep. If the analysis unit 28 is repeatedly used for a specific user, an average value of a plurality of sleeps can be used based on the past data of the user. On the other hand, when it is initially set or for an individual user, a real-time estimated value cannot be obtained if an attempt is made to standardize from the beginning. Therefore, it is desirable to set a general value as a reference value and shift to a determination based on a standardized value after a few hours have passed.

An example of a method for determining the sleep stage based on the standardized value Te (second) of the average value tea of the expiration time is as follows.
Te <−0.5s ・ Awakening or REM sleep −0.5s ≦ Te <0.8s ・ ・ Shallow non-REM sleep 0.8s ≦ Te ・ ・ ・ ・ ・ ・ ・ Deep non-REM sleep

An example of a method for determining the sleep stage based on a general value from the average value tea (second) of the expiration time is as follows.
tea <2.2s ・ Awakening or REM sleep 2.2s ≦ tea <3.2s ・ Shallow non-REM sleep 3.2s ≦ tea ・ ・ ・ ・ ・ ・ Deep non-REM sleep

  In step 75, the estimated value of the sleep stage is output from the output interface 24. Therefore, the environment control function 30 controls the environment so as to lead to deep non-REM sleep in the early stages of sleep, and controls the environment so that it occurs at the time of awakening or REM sleep at the stage of waking up. The process according to a user's sleep state can be performed.

  In the flowchart shown in FIG. 2, steps 71, 72 and 73 are performed in series. However, since these processes use the memory 25 as a buffer, each process is performed at an independent timing. It is possible to execute. Therefore, it is possible to output the user's sleep stage in real time with almost no delay.

  FIG. 4 shows the presence or absence of correlation between some information obtained from the respiratory signal and the low frequency component (δ wave component) of the electroencephalogram (EEG). The vertical axis represents the signal intensity and merely shows a tendency to increase or decrease. Changes in the breathing time tb and the inspiratory time ti are small, and no correlation with the low frequency component of EEG is positively recognized. On the other hand, the change in expiration time te is relatively large, and a stable correlation with the low frequency component of EEG is recognized.

  FIG. 5 to FIG. 8 further show the correlation using standardized data. FIG. 5 shows the standardized value Te of the 5-minute average value of the expiration time and the standardized value of the low frequency component of EEG (the 5-minute addition value of the amplitude component of the δ wave spectrum, the same applies to FIGS. 6 and 7). ing. The tendency of increase / decrease in the standardized value Te of the expiration time and the low frequency component of the EEG is almost the same. In particular, a good correlation is seen in the early stages of sleep. Therefore, it can be seen that by setting an appropriate threshold value, it can be used as data suggesting the sleep stage in the same manner as the low frequency component of EEG.

  On the other hand, FIG. 6 shows the standardized value of the average respiratory rate for 5 minutes and the low frequency component of EEG. It is difficult to find a law in relation to increase / decrease. FIG. 7 shows the standardized value of the ratio of the average expiration time for 5 minutes to the average inspiration time and the low frequency component of EEG. Also in this figure, it is difficult to find a law property in relation to increase / decrease.

  FIG. 8 shows a comparison between the sleep stage obtained by the PSG method and the sleep stage obtained by the analysis unit 28 described above. FIG. 8A shows the sleep stage obtained by the PSG method. REM sleep is sleep in which rapid eye movement is observed. The electroencephalogram is mainly a relatively fast θ wave. In humans, it is said that REM sleep occupies 1 hour and a half to 2 hours out of 6 to 8 hours of sleep.

  Stage 1 (S1) to stage 4 (S4) are collectively called non-REM sleep. Stage 1 (S1) is in a somnolence state. On the brain wave, the α wave observed at awakening decreases, and a low-amplitude potential is observed. Stage 2 (S2) is a stage where a sleep spindle is observed on the electroencephalogram. Stage 3 (S3) is a stage in which low-frequency δ waves increase, and is about 20% to 50%. Stage 4 (S4) is a stage in which the δ wave is 50% or more.

  FIG. 8B shows a state in which the above-described determination criterion is applied to the standardized value Te. FIG. 8C shows the sleep stage output from the analysis unit 28. It can be seen that the sleep stage obtained by the PSG method almost coincides with the sleep stage and that the sleep stage can be accurately determined.

  The analysis unit 28 and the analysis method described above associate the expiration time with the low frequency component of the electroencephalogram based on the fact that the increase and decrease of the expiration time included in the respiratory signal is highly correlated with the increase and decrease of the low frequency component of the electroencephalogram. By judging, the sleep stage is judged. As a result, as described above, the sleep stage can be obtained with an accuracy comparable to that of the PSG method using the respiratory signal. In addition, since the input information for estimating the sleep state may be a respiratory signal, in the apparatus, system and method according to the present invention, it is not necessary to attach an electrode or the like to the sleeper, and the sleeper's restraint can be reduced. Therefore, a system that provides more comfortable sleep can be provided.

  Although the sleep state can be accurately estimated from the expiration time, the respiration signal includes a lot of information that is considered to be related to sleep such as the number of breaths as disclosed in the above-mentioned patent document. Therefore, in addition to the expiration time, other factors included in the respiratory signal can be added as a sleep state determination element, or the main element of the determination can be selected based on the transition of sleep. There is a possibility of improvement.

The figure which shows schematic structure of the home system for bedrooms. The flowchart which shows the method of acquiring and analyzing an expiration time. FIG. 3A shows an example of a respiratory signal, and FIG. 3B shows an example of a respiratory signal having a shoulder. The figure which shows the correlation with some information contained in a respiration signal, and the low frequency component of an electroencephalogram. The figure which shows the correlation with an average expiration time and an EEG low frequency component. The figure which shows the correlation with an average respiration frequency and an EEG low frequency component. The figure which shows the correlation with the expiration time / inhalation time and an EEG low frequency component. FIG. 8A shows a sleep stage obtained by the PSG method, FIG. 8B shows a change in expiration time, and FIG. 8C shows a sleep stage obtained by the above analysis method.

Explanation of symbols

2 sensor sheet, 3 information processing unit, 7 pressure sensitive element 10 biological information detection unit, 20 control unit for bedroom 28 analysis unit, 30 environment control unit 50 home system for bedroom

Claims (8)

Means for obtaining a respiratory signal including a plurality of respiratory peaks each including an expiratory portion and an inspiratory portion;
Means for extracting and recording the expiration time of the expiration part included in each of the plurality of respiratory peaks in a memory;
Means for including a plurality of expiration time changes recorded in the memory as a first element corresponding to an increase or decrease in intensity of a low frequency component of an electroencephalogram as a determination element to estimate a sleep state;
An analysis device that outputs a result of estimating a biological state, and means for outputting the estimated result.   2. The analysis apparatus according to claim 1, wherein the estimating means statistically processes a plurality of expiration times recorded in the memory, and uses the increase / decrease of the expiration time subjected to the statistical processing as the first element.   2. The estimation unit according to claim 1, wherein the estimating means increases or decreases the sum or average of a plurality of exhalation times included in a predetermined number of respiration peaks or a plurality of exhalations included in a plurality of respiration peaks included in a predetermined time interval. An analysis apparatus in which an increase / decrease in an average of time is the first element. An analysis device according to claim 1;
A sheet-type sensor that can detect a load change of the user in a lying state,
And a device for generating the respiration signal from an output signal of the seat type sensor.   5. The living body monitoring system according to claim 4, wherein the sheet type sensor includes a plurality of pressure sensitive elements assembled to a sheet-like support member. An analysis device according to claim 1;
And an apparatus for controlling at least a part of the living environment based on the output of the analysis apparatus. Obtaining a respiratory signal including a plurality of respiratory peaks each including an expiratory portion and an inspiratory portion;
Extracting the exhalation time of the exhalation part contained in each of the plurality of breathing peaks and recording it in a memory;
Including a plurality of expiration time increases / decreases recorded in the memory per predetermined time as a first element corresponding to an increase / decrease in the intensity of a low frequency component of an electroencephalogram, and estimating a sleep state;
A method of outputting a result of estimating a biological state, the method including outputting an estimated result.   8. The method according to claim 7, wherein the estimating step includes statistically processing a plurality of expiration times recorded in the memory, and setting the increase or decrease in the expiration time statistically processed as the first element.

Images