JP6518570B2 - Sleep stage estimation device, method and program - Google Patents

Description

  The present invention relates to a sleep stage estimation device, method and program for estimating a person's sleep stage.

  The sleep stage is an index that indicates the depth of sleep, and is defined by 2 to 6 stages according to international standards. A sleep progress map representing sleep stages in chronological order is important information for grasping the state of sleep in the field of medical care and healthcare. The sleep stage is determined for each time interval (about 30 seconds) based on R & K method which is an inspection determination rule, using a brain wave data or electromyogram data obtained by polygraph examination. However, since this polygraph test has a large physical load for the user, research on sleep stage estimation technology using heart rate data that can be acquired with a measuring device with a small physical load has been actively conducted in recent years.

  The conventional sleep stage estimation method using heart rate data, for example, generates a plurality of feature quantities related to the sleep stage from heart rate data, and estimates using machine learning (support vector machine, random forest, etc.) (See, for example, Non-Patent Document 1).

  In addition, techniques have been proposed to improve sleep stage estimation accuracy in consideration of the nature of the distribution of the sleep stage and transition probability depending on the elapsed time since going to bed (eg, non-patent) Reference 2).

Xiao, M. et al., "Sleep stages classification based on heart rate variability and random forest", Biomedical Signal Processing and Control, p 624-633, (2013). "Time-dependent sleep stage transition model using heart rate variability" DEIM Forum 2015 G6-5, http://db-event.jpn.org/deim2015/paper/24.pdf (confirmed on October 22, 2015).

  However, the autonomic nervous activity reflected in the heart rate is also changed by influences other than brain activity during sleep, such as blood pressure regulation and temperature regulation, movement of the body and respiration. For this reason, it is difficult for the technique described in Non-Patent Document 1 to estimate the sleep stage obtained from brain activity completely in accordance with the heartbeat.

  Further, the technique described in Non-Patent Document 2 has a disadvantage that the awakening estimation accuracy in the awake state is low.

  The present invention has been made in view of the above circumstances, and the object of the present invention is a sleep stage estimation apparatus which can estimate the sleep stage more accurately even if the heart beat is strongly affected by other than brain activity. , Providing a method and program.

In order to achieve the above object, one aspect of the present invention is a heartbeat fluctuation storage unit storing heartbeat fluctuation information in which heartbeat fluctuation features calculated for each first unit time from a set of electrocardiograms are summed up for each sleep stage; An acceleration storage unit storing acceleration information obtained by aggregating, according to a sleep stage, acceleration feature amounts calculated for each first unit time from a set of accelerations; and a second time axis of the electrocardiogram being longer than the first unit time An appearance frequency storage unit storing appearance frequency information in which the appearance frequency of the sleep stage for each first unit time is calculated for each time zone divided in the unit time, and for each of the divided time zones; And a transition frequency storage unit storing transition frequency information in which the transition frequency of the sleep stage for each first unit time is totaled is provided as a model database. Then, from the electrocardiogram acquired from the subject to be estimated, the feature amount of the heart rate fluctuation for each first unit time is calculated for each time zone divided by the second unit time, and from the subject to be estimated The acceleration feature quantity is calculated for each of the first unit time from the set of acquired accelerations, and whether the subject is awake or not is determined from the acceleration acquired from the estimation target subject, and the calculation is performed. The temporal change in the heart rate fluctuation feature amount and the temporal change in the acceleration feature amount, and the information stored in the heart rate fluctuation storage unit, the acceleration storage unit, the appearance frequency storage unit, and the transition frequency storage unit relative probability distribution of the sleep stage of the subject, if not determined that the wakefulness sleep stage and time of the time interval immediately preceding set for each of the first unit time, heart rate variability of the current time interval Set based on the probability by symptoms amount and the acceleration characteristic quantity, if it is determined that the awake state to estimate the probability distribution of the corrected as the mode of the probability distribution of the sleep stage is waking sleep stage of the subject It is something like that.

  According to the present invention, it is possible to provide a sleep stage estimation device, method and program that can estimate sleep stage more accurately even if the heartbeat is strongly influenced by other than brain activity.

The figure which showed an example of the appearance frequency of the sleep stage according to the time zone. The figure which shows an example of the appearance probability of a sleep step transition pattern. BRIEF DESCRIPTION OF THE DRAWINGS The block diagram which shows the function structure of the sleep stage estimation apparatus which concerns on one Embodiment of this invention. The flowchart which shows the process sequence and process content in the 1st learning phase which concern on the same embodiment. The flowchart which shows the process sequence and process content in the 2nd learning phase which concern on the same embodiment. The flowchart which shows the process sequence and process content in the estimation phase which concern on the same embodiment. The figure which shows an example of the heart rate fluctuation table according to the sleep stage concerning the embodiment. The figure which shows an example of the sleep stage appearance frequency table according to time concerning the embodiment. The figure which shows an example of the sleep stage transition frequency table according to time concerning the embodiment. The figure which shows an example of the acceleration table according to sleep stage concerning the embodiment. The figure which shows an example of the observation method of time-sequential heart rate data which concerns on the embodiment, and the output method of a sleep stage estimated value. The figure which shows an example of the Lorentz plot concerning the embodiment. The figure which displayed the type of the heartbeat fluctuation feature-value based on the embodiment as a list.

Hereinafter, embodiments according to the present invention will be described with reference to the drawings.
(principle)
This invention focuses on the fact that the appearance frequency and transition probability of the sleep stage change according to the elapsed time after sleep onset, and can estimate the more likely sleep stage even when the heart beat is strongly affected by other than brain activity. It is For example, there is a low possibility of sudden awakening and falling again from a deep sleep after sleep onset, but the state may be estimated with a probability of 0.52 when estimated based on the heart rate alone. In the present invention, the estimation accuracy is improved by reducing the estimation probability for such rare occurrences of the transition and the estimation confidence.

  FIG. 1 shows an example of the result of calculating the number of appearances of each sleep stage every 30 minutes using sleep stage data for 45 persons. As apparent from this example, the appearance frequency of REM (REM sleep) in 30 minutes from the onset of sleep onset (Fig. 1 (a)) is 180 minutes and 360 minutes from the onset of sleep shown in FIG. 1 (b) and (c) respectively. It is less than the number of occurrences of REM in 30 minutes after the lapse.

  Moreover, FIG. 2 shows an example of the result of having calculated the appearance probability of the transition pattern of each sleep stage for every 30 minutes using the sleep stage data for 45 persons similarly. As shown in the figure, the transition probability from REM to NREM (non-REM sleep) in 30 minutes from the onset of sleep onset (Fig. 2 (a)) is 150 minutes elapsed from the onset of sleep shown in Figs. 2 (b) and (c). It can be seen that it is higher than the transition probability from REM to NREM in 30 minutes after and after 300 minutes.

  Therefore, in the present invention, the appearance frequency and transition probability of the sleep stage according to the elapsed time are incorporated into the generation model of the sleep stage by the learning phase, and the feature quantity of the user's heartbeat data is determined in the estimation phase, And the sleep stage is estimated based on the generation model of the sleep stage. In this way, it is possible to estimate the sleep stage more accurately even if the heart beat is strongly affected by other than brain activity.

First Embodiment
<Configuration>
FIG. 3 is a block diagram showing a functional configuration of a sleep stage estimation apparatus according to an embodiment of the present invention.
The sleep stage estimation apparatus 1 according to the present embodiment includes, for example, a server computer or a personal computer, and includes a control unit 10, a storage unit 20, and an input / output interface unit 30. Among them, the input / output interface unit 30 takes in the electrocardiogram data output from the electrocardiograph 2 and the acceleration data output from the three-axis accelerometer 5 mounted on the chest, and also an input unit such as a keyboard or a touch panel. Operation data and display data are input / output to / from the display unit 4 such as 3 and a liquid crystal device.

  The storage unit 20 uses a writable and readable non-volatile memory such as a hard disk drive (HDD) or a solid state drive (SSD) as a storage medium, and is a storage unit necessary to implement this embodiment. , And the estimated data storage unit 22.

The model database 21 includes a heart rate fluctuation table 211, an appearance frequency table 212, a transition frequency table 213, and an acceleration table 214.
The heart rate fluctuation table 211 is used to store the heart rate feature amount level and the appearance frequency in association with the identification information (heart rate feature vector ID) of the heart rate feature vector for each sleep stage. An example is shown in FIG.

  The appearance frequency table 212 is used to store sleep stages and appearance frequencies that appear in the time zone according to the time zone. An example is shown in FIG.

The transition frequency table 213 is used to store sleep phase transition patterns and their appearance frequency that occur in the time zone according to the time zone. FIG. 9 shows an example thereof.
The acceleration table 214 is used to store the acceleration feature amount level and the appearance frequency in association with identification information (acceleration feature vector ID) of the acceleration feature vector for each sleep stage. An example is shown in FIG.
The estimated data storage unit 22 is used to store data representing the sleep stage estimation result.

  The control unit 10 is provided with a central processing unit (CPU: Central Processing Unit) and a memory, and as a control processing function necessary to carry out the present embodiment, the heartbeat fluctuation feature value calculation unit 11 and sleep stage The heart rate fluctuation analysis unit 12, the hourly sleep stage appearance frequency analysis unit 13, the hourly sleep stage transition frequency analysis unit 14, the sleep stage acceleration analysis unit 15, the acceleration feature quantity calculation unit 16, and the posture determination unit 17 And a sleep stage probability estimation unit 18 and an estimated data output control unit 19.

  The heart rate fluctuation analysis unit 12 classified by sleep stage, the sleep phase appearance frequency analysis unit 13 classified by time, the sleep stage transition frequency analysis unit 14 classified by time, and the acceleration analysis unit 15 classified by sleep stage constitute a model creating unit 10A.

  These control processing functions are all realized by causing the CPU to execute a program stored in a program memory (not shown).

  The heart rate variability feature amount calculation unit 11 calculates a heart rate variability feature amount based on the electrocardiogram data acquired from the electrocardiograph 2 via the input / output interface unit 30. Specifically, an RRI (R-R Interval) representing a peak interval is detected from the electrocardiogram waveform, and the detected RRI is divided every n seconds, and the frame length N [sec], the slide width n [sec ], The feature vector ht = {ht1, ht2, ..., htk} of the heart rate fluctuation is calculated. Here, t represents an elapsed time, and is counted for n seconds as t = {n, 2 * n, 3 * n,.

  Based on the time information t, the heart rate variability analysis unit 12 for the sleep stage calculates the feature vector ht of the heart rate variability calculated by the heart rate variability feature value calculation unit 11 in the first learning phase described later. The distribution of each heartbeat feature value of each sleep stage is calculated in correspondence with the correct label of the sleep stage for every n seconds input in. Then, a process of storing the distribution of each heartbeat characteristic value calculated for each sleep stage in the heartbeat fluctuation table 211 in the model database 21 is performed.

  In the first learning phase to be described later, the hourly sleep stage appearance frequency analysis unit 13 divides the time series data of the sleep stage correct answer label every n seconds input in the input unit 3 every q minutes (for example, 30 minutes) Then, the number of appearances of each sleep stage i in each divided time zone m is calculated. Then, processing of storing the calculated number of appearances of each sleep stage i in the appearance frequency table 212 in the model database 21 is performed.

  In the first learning phase to be described later, the hourly sleep stage transition frequency analysis unit 14 performs every n seconds in each time zone m divided by q minutes in the same manner as the above-described hourly sleep stage appearance frequency analysis unit 13. Calculate the number of appearances of each sleep stage transition. Then, processing is performed to store the calculated number of occurrences of each sleep stage transition in the transition frequency table 213 in the model database 21.

  The acceleration feature quantity calculation unit 16 calculates an acceleration feature quantity based on acceleration data acquired from the three-axis accelerometer 5 via the input / output interface unit 30 in the second learning phase described later.

Specifically, the D value and the awakening determination value for each fixed section are calculated using the algorithm of Cole et al. (Non-Patent Document 3) shown below. The D value and the awakening determination value are values obtained by the algorithm of Cole et al., And these are set as acceleration feature vectors at = {at1, at2, ..., atw}.
Roger J. Cole, Daniel F. Kripke, William Gruen, Daniel J. Mullaney, J. Christian Gillin. "Automatic Sleep / Wake Identification From Wrist Activity" Sleep, 15 (5): 461-469, (1992).

  The posture determination unit 17 determines the posture based on acceleration data acquired from the three-axis accelerometer 5 via the input / output interface 30 in an estimation phase to be described later. Specifically, a binary determination of the sleeping posture or the posture in which the body is raised is performed. For example, when the accelerometer 5 is mounted on the chest such that the y-axis of the 3-axis accelerometer 5 is horizontal to the subject's body and the lower part of the body has a positive value, the gravity acceleration value of the y-axis is If it exceeds a certain level, it is judged as the posture that raised the body.

  The sleep stage probability estimation unit 18 calculates the heart rate variability feature vector calculated by the heart rate variability feature amount calculation unit 11 based on the electrocardiogram data input from the electrocardiograph 2 in the estimation phase to be described later, and the accelerometer 5. The acceleration feature vector calculated by the acceleration feature amount calculation unit 16 based on the inputted acceleration data, the information stored in each of the tables 211 212 213 and 214 of the model database 21, the posture determination unit 17 Sleep stage St every n seconds is estimated using the awake state determined from the acceleration data according to. Then, processing for storing data representing the estimation result in the estimation data storage unit 22 is performed.

  The estimated data output control unit 19 reads out data representing the sleep stage estimation result in the corresponding time zone of the corresponding user from the estimated data storage unit 22 according to the output instruction inputted in the input unit 3. Then, display data for displaying the read estimation data is generated, and the generated display data is output to the display unit 4 via the input / output interface unit 30 and displayed.

<Operation>
Next, an operation of estimating the sleep stage by the device configured as described above will be described.
(1) Learning phase In the first and second learning phases, while measuring the subject's EEG, EMG, ECG waveform and acceleration, the measurer such as a medical worker visually observes the subject's sleep stage every n seconds. The observation is performed, and the observation result is input by the input unit 3 as a "sleep phase correct answer label".

  Then, the sleep stage estimation device 1 calculates the feature quantity of the heart rate fluctuation and the feature quantity of the acceleration from the electrocardiogram waveform and the acceleration, and based on the inputted correct label of the sleep stage, in each sleep stage The appearance frequency of the feature quantity of the heart rate fluctuation and the appearance frequency of the feature quantity of the acceleration in each sleep stage are calculated and stored in the heart rate fluctuation table 211 and the acceleration table 214, respectively. The frequency is calculated and the calculated value is stored in the appearance frequency table 212 and the transition frequency table 213, respectively.

FIG. 4 is a flowchart showing the processing procedure and processing contents in the sleep stage estimation device 1 in the first learning phase.
When the measurement in the first learning phase is started, the subject's electrocardiogram data output from the electrocardiograph 2 at step S11 is controlled via the input / output interface unit 30 under the control of the heart rate fluctuation feature quantity calculation unit 11. Taken in. Then, based on the electrocardiogram data acquired at step S12, the feature value of the heart rate fluctuation is calculated as follows.

  That is, first, RRI (RR Interval) representing the peak interval of the electrocardiogram waveform is detected from the electrocardiogram data, and the detected RRI is divided every n seconds. Then, the feature vector ht = {ht1, ht2, ..., htk} of the heart rate fluctuation is calculated as the frame length N [sec] and the slide width n [sec]. Here, t represents an elapsed time, and is counted for n seconds as t = {n, 2 * n, 3 * n,. A specific example is shown in FIG.

  The heart rate variability feature value includes, for example, an average value of heart rate variability RRI (R-R Interval), a standard deviation of RRI, a root of a root mean square of differences between RRIs before and after (RMSSD), and a time difference between RRIs before and after The occurrence rate (pNN50) of the number of times the LRI becomes 50 ms or more, the length L of the long side of the shape drawn by Lorentz plot of RRI, and the length T, L of the short side of the shape drawn by Lorentz plot of RRI Sympathetic activity index (CSI: Cardiac Sympathetic Index) determined from T, parasympathetic activity index (CVI: Cadiac Vagel Index) determined from L and T, high frequency component (HF) of the power spectrum waveform of RRI, or low frequency At least one of the components (LF) or their feature quantities Other feature quantities calculated from

  An example of Lorentz plot of RRI is shown in FIG. 12, and the definition of each of the heart rate fluctuation feature values described above is shown in FIG. The detailed calculation method of the heart rate fluctuation feature amount is described, for example, in the non-patent document 1.

  The values of the feature quantities of the heart rate variability feature vector ht may be, for example, normalized to a range of 0 to 1 and then discrete at L level. For example, the first element of the feature vector when t = n (first n second interval) is expressed as hn, 1 = {0, 1,..., L}.

Further, in parallel with the first learning phase, the sleep stage estimation device 1 also executes the processing of the second learning phase.
FIG. 5 is a flowchart showing the processing procedure and processing contents in the sleep stage estimation device 1 in the second learning phase.
When measurement in the second learning phase is started, under the control of the acceleration feature quantity calculation unit 16, first, in step S21, the subject's acceleration data output from the accelerometer 5 is fetched via the input / output interface unit 30. Be Then, based on the acquired acceleration data in step S22, the acceleration feature amount calculation unit 16 calculates the feature amount of acceleration as follows.

  That is, first, using the algorithm described in the non-patent document 1 from the acceleration data, the D value every n seconds, the awakening determination value, etc. are calculated, and the acceleration feature vector at = {at1, at2, ..., atw} Ask for

The values of the feature quantities of the acceleration feature vector at may be, for example, normalized to a range of 0 to 1 and then discrete at Z level. For example, the first element of the feature vector when t = n (first n second interval) is expressed as an, 1 = {0, 1,..., Z}.

  Subsequently, based on the acceleration feature vector at which is calculated by the acceleration feature quantity calculation unit 16, the acceleration analysis unit 15 analyzes the acceleration of each sleep stage as follows. That is, the calculated acceleration feature vector at is associated with the n-second sleep stage correct answer label input from the input unit 3 based on the time information t. Then, the distribution of each acceleration feature at each sleep stage is calculated.

  In one of the first learning phases, the heart rate fluctuation analysis unit 12 according to the sleep stage based on the heart rate fluctuation feature vector ht calculated by the heart rate fluctuation feature amount calculation unit 11 shows that As analyzed. That is, the calculated heart rate fluctuation feature vector ht is associated with the n-second sleep stage correct answer label input by the input unit 3 based on the time information t. Then, the distribution of each heart rate feature of each sleep stage is calculated.

  Next, in step S13, the appearance frequency of each sleep stage for each time zone (m minutes) is totalized as follows by the time-based sleep stage appearance frequency analysis unit 13. That is, first, time series data of the sleep stage correct answer label every n seconds input in the input unit 3 is divided every q minutes (for example, 30 minutes), and appearance of each sleep stage i in each divided time zone m The number is calculated. At this time, m = 1 is represented as an interval of 0 minutes to (q × 1) minutes, and m = 2 is represented as an interval of (q × 1) minutes to (q × 2) minutes.

  Then, the calculated number of appearances of each sleep stage i is stored in the appearance frequency table 212 together with the information indicating the sleep stage in association with the time zone, as shown in FIG. 8, for example.

  Subsequently, in step S14, the appearance frequency of each sleep stage transition pattern for each time zone (m minutes) is totalized as follows by the hourly sleep stage transition frequency analysis unit 14. That is, first, in each time zone m divided every q minutes, the number of appearances of each sleep stage transition every n seconds is calculated.

  For example, in a certain time zone m, when the first 0 to n second interval sleep stage is "wake up (WAKE)" and the n to 2 n second interval sleep stage is "REM sleep (REM)", the n to 2 n second interval The transition of the sleep stage of is counted as a transition pattern of “wake up (WAKE) → rem sleep (REM)”.

  In addition, it is not necessary to count the transition pattern of the sleep stage of the first 0 to n second interval. Then, the number of appearances of each sleep stage transition calculated as described above is stored in the transition frequency table 213 together with the information representing the sleep stage transition pattern in association with the time zone, as shown in FIG. 9, for example.

  Subsequently, in step S15, the heart rate variability analysis unit 12 for each sleep stage counts the heart rate variability feature amount for each sleep stage, and the calculated appearance frequency is, for example, the heart rate feature vector ID and the heart rate as shown in FIG. The feature level is associated with each sleep stage and stored in the heart rate fluctuation table by sleep stage 211, and the first learning phase is completed.

  For example, assuming that t_wake = {n, 5 * n, 8 * n} when the correct answer label in the sleep stage is "wake up (WAKE)", the count becomes ht_wake, 1 = 1 for all times. Is stored. At this time, since the discrete values of the heart rate variability feature amount are 0 ≦ l ≦ L, they are counted for each value.

  In the other second learning phase, finally, in step S23, the acceleration feature quantity for each sleep stage is summed up by the sleep stage acceleration analysis unit 15, and the appearance frequency as the calculation result is shown, for example, in FIG. As described above, the acceleration feature vector ID and the acceleration feature amount level are stored in the acceleration table 214 for each sleep stage in association with each sleep stage, and the second learning phase is completed.

  For example, assuming that t_wake = {n, 5 * n, 8 * n} when the correct answer label in the sleep stage is "wake up (WAKE)", the count becomes at_wake, 1 = 1 for all times. Is stored. At this time, since the discrete values of the acceleration feature amount are 0 ≦ l ≦ M, they are counted for each value.

(2) Estimation phase In the estimation phase, only the electrocardiogram data and acceleration data of the person to be estimated are measured, and the heart rate fluctuation feature amount calculated from the measured electrocardiogram data and acceleration data, the acceleration feature amount, and the posture determination unit 17 Based on the heart rate fluctuation table 211, the acceleration table 214, the appearance frequency table 212, and the model data stored in the transition frequency table 213, the sleep stage St of the person to be estimated is estimated every n seconds.

  FIG. 6 is a flowchart showing the processing procedure and the processing content. First, in step S31, the heart rate fluctuation feature quantity calculation unit 11 and the acceleration feature quantity calculation unit 16 cause the electrocardiogram in one sleep of the person to be estimated from the electrocardiograph 2 and the accelerometer 5 via the input / output interface unit 30. Data and acceleration data are captured. Then, in step S32 and step S33, the heart rate variability feature amount and the acceleration feature amount are extracted every n seconds from the acquired electrocardiogram data and the acceleration data in parallel by the heart rate variability feature amount calculation unit 11 and the acceleration feature amount calculation unit 16. It is calculated. The calculation method is the same as in the case of the learning phase.

  On the other hand, in step S38 executed in parallel, the posture determination unit 17 determines the posture every n seconds from the acquired acceleration data. That is, a binary determination of the sleeping posture or the posture in which the body is raised is performed. Awake in the case of a posture that raised the body.

  In the following step S39, it is determined whether the result of the binary determination is "WAKE" indicating awakening.

  On the other hand, under the control of the sleep stage probability estimation unit 15 following the step S33, estimation of the sleep stage of the person to be estimated is performed as follows. That is, first, in step S34, sleep phase variables every n seconds are randomly initialized. Also, the Gibbs sampling counter g is initialized to "1".

  Next, in step S35, the sleep step variable every n seconds is updated by Gibbs sampling, and the following processing is repeatedly performed until it is determined in step S36 that the Gibbs sampling counter g has reached a preset Gibbs sampling repetition number. Be done.

(A) As an initial value, time t = 0 and sleep stage S0 = “WAKE” are set.
(B) While the time t is smaller than the sleep time T, the following processing is performed.
(B-1) In step S39, when the posture determination unit 17 does not determine awakening, in step S35, the sleep stage S_ {t, g} is a sleep stage St−1 of time t−1, time The sampling is performed according to the value of P (St = i | ht, at, St-1 = j, Ct = m) using the heart rate fluctuation feature amount ht and the acceleration feature amount at at time C.
(B-2) If it is determined that the posture determination unit 17 is awake in step S39, the sleep step S_ {t, g} is set to "WAKE" in the following step S40.

(B-3) Time t = t + 1.
(C) The Gibbs sampling counter g is incremented by "+1".

  Subsequently, in step S37, from each of the g sleep stage sampling history sets {S_ {t, g}} for each time t, the appearance distribution of each sleep stage is summed up and output as the sleep stage probability distribution at time t. When outputting the sleep stage, the sleep stage which becomes the mode value in the sampling history set {S_ {t, g}} is stored in the estimated data storage unit 22 as an estimation result.

  P (St = i | ht, at, St-1 = j, Ct = m) is obtained by the following equation.

  Here, θ mi represents the probability that the sleep stage becomes i in the time zone m, and is calculated by the following equation.

  Here, Mmi represents the number of times the sleep stage i appears in the time zone m.

  Further, λmji represents the probability that the sleep stage transitions from j to i in the time zone m, and is calculated by the following equation.

  Here, Mmji represents the number of times that the sleep stage transitions from j to i in time zone m.

  Also, φ i h tk represents the appearance probability of each heartbeat feature at sleep stage i,

Calculated by

  Furthermore, πiatw represents the sleep probability of each acceleration feature at sleep stage i,

Calculated by
When the kth element htk of the heart rate variability feature vector newly obtained at time t in the sleep stage i is lk, Milk refers to the model database and the appearance frequency when htk satisfies the value lk Represents the number of times.

  When the w-th element of the acceleration feature vector newly obtained at time t in the sleep stage i is zw, Mizw refers to the model database 21 and the appearance frequency when atw satisfies the value zw Represents the number of times.

  The data representing the estimation result of the sleep stage of the estimation target person stored in the estimation data storage unit 22 is read out from the estimation data storage unit 22 under the control of the estimation data output control unit 19. Then, display data is generated based on the read estimation data, and the display data is supplied to the display unit 4 through the input / output interface unit 30 and displayed. The lower part of FIG. 11 shows an example of the display result.

<Effect of the embodiment>
As described above, in one embodiment, in the learning phase, the heart rate variability feature amount and the acceleration feature amount for each sleep stage are calculated from the electrocardiogram data and the acceleration data of the subject and stored in the heart rate variability table 211 and the acceleration table 214 At the same time, the appearance frequency and transition frequency of the sleep stage according to the passage of time are obtained and stored in the appearance frequency table 212 and the transition frequency table 213, respectively. Then, in the estimation phase, the heart rate fluctuation feature amount calculated from the electrocardiogram data of the estimation target person and the acceleration data, the acceleration feature amount, the awakening determination result by the posture determination unit 17, and the sleep stored in the respective tables 211, 212, 213 Based on the model data of the stage, the sleep stage of the person to be estimated is estimated.

  Therefore, even if the heartbeat changes due to influences other than brain activity during sleep, such as blood pressure regulation and temperature regulation, movement of the body and breathing, etc., it is possible to estimate the sleep stage accurately. In addition, since the body movement is large at the time of awakening, the awakening can be estimated more accurately by considering the acceleration. Then, by incorporating the awakening determination by the posture determination, the awakening state when the user is raised from the bed can be reliably detected.

  Furthermore, by providing the learning phase, it is possible to construct the optimal heartbeat fluctuation table 211, the acceleration table 214, the appearance frequency table 212, and the transition frequency table 213.

  When it is considered that the sleep stage estimation device 1 is realized by, for example, a portable information device such as a smartphone or a tablet personal computer and an application program mounted thereon, the user of the portable information device as a subject is awakened If the detection accuracy in the state is reduced, it may lead to the user's reliability of the application program being significantly reduced. Therefore, by reliably determining that the user is in the awakening state from the body movement or posture based on the acceleration data, it is possible to reliably suppress the decrease in the reliability of the user for the application program.

[Other embodiments]
The present invention is not limited to the above embodiment. For example, although the case where the storage unit 20 was provided in the sleep stage estimation apparatus 1 was described as an example in the above embodiment, the storage unit 20 is provided in a database server etc. provided in the cloud, and the sleep stage estimation apparatus 1 and its database Data may be written and read by communicating with the server.

  In the above embodiment, the heart rate fluctuation feature amount is calculated based on the electrocardiogram data obtained by the electrocardiograph, but the pulse wave is measured and the heart rate fluctuation feature amount is calculated based on the measurement data. You may do so.

  Furthermore, data representing the estimation result may be transmitted to another terminal or server via a communication line, and the control function of the other control unit, the processing procedure and processing content by the control function, and the display of the estimation result The format and the like can be variously modified and implemented without departing from the scope of the present invention.

  In short, the present invention is not limited to the above embodiment as it is, and in the implementation stage, the constituent elements can be modified and embodied without departing from the scope of the invention. In addition, various inventions can be formed by appropriate combinations of a plurality of constituent elements disclosed in the embodiments. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, components in different embodiments may be combined as appropriate.

  DESCRIPTION OF SYMBOLS 1 ... sleep stage estimation apparatus, 2 ... electrocardiograph, 3 ... input part, 4 ... display part, 5 ... accelerometer, 10 ... control unit, 10 A ... model preparation part, 11 ... heart rate fluctuation feature quantity calculation part, 12 ... Heart rate fluctuation analysis unit by sleep stage, 13 ... Sleep stage appearance frequency analysis unit by hour, 14 ... Sleep stage transition frequency analysis unit by hour, 15 ... Acceleration analysis unit by sleep stage, 16 ... Acceleration feature quantity calculation unit, 17 ... Posture Determination unit 18 Sleep stage probability estimation unit 19 Estimated data output control unit 20 Storage unit 21 Model database 22 Estimated data storage unit 211 Heart rate fluctuation table 212 Appearance frequency table 213 Transition frequency table, 214: acceleration table, 30: input / output interface unit.

Claims (3)

A heart rate fluctuation storage unit storing heart rate fluctuation information in which heart rate fluctuation feature values calculated for each first unit time from a set of electrocardiograms are tabulated by sleep stage;
An acceleration storage unit storing acceleration information obtained by aggregating, according to sleep stages, acceleration feature amounts calculated for each of the first unit times from a set of accelerations;
The appearance frequency information in which the appearance frequency of the sleep stages for each of the first unit times is stored is stored for each time zone obtained by dividing the time axis of the electrocardiogram by the second unit time longer than the first unit time. Appearance frequency storage unit,
A transition frequency storage unit storing transition frequency information in which the frequency of transition of sleep stages for each of the first unit time is calculated for each of the divided time zones;
A heart rate variability feature amount calculating unit that calculates a feature amount of heart rate variability for each first unit time for each time zone divided by the second unit time from an electrocardiogram acquired from a subject to be estimated;
Acceleration feature amount calculation means for calculating an acceleration feature amount for each of the first unit time from a set of accelerations acquired from a subject to be estimated;
A posture determination unit that determines whether or not the subject is awake from the acceleration obtained from the subject to be estimated, based on the posture of the subject;
The calculated temporal change of the heart rate fluctuation feature amount and the temporal change of the acceleration feature amount, and the information stored in the heart rate fluctuation storage unit, the acceleration storage unit, the appearance frequency storage unit, and the transition frequency storage unit And the sleep stage and time zone of the immediately preceding time interval set for each of the first unit times when the posture determination unit does not determine the awakening state with respect to the probability distribution of the sleep stage of the subject based on The subject is set based on the heart rate fluctuation feature amount of the current time interval and the probability based on the acceleration feature amount, and when it is determined to be awake state, the mode of the probability distribution of the sleep stage is corrected to become awakening and the subject's sleep A sleep stage estimation apparatus comprising: estimation means for estimating a stage probability distribution. A heart rate variability storage unit storing heart rate variability information in which heart rate variability features calculated for each first unit time from a set of electrocardiograms are accumulated for each sleep stage; and for each first unit time from a set of accelerations An acceleration storage unit storing acceleration information obtained by summing up the acceleration feature amount for each sleep stage, and the second time interval obtained by dividing the time axis of the electrocardiogram by a second time unit longer than the first time unit; An appearance frequency storage unit storing appearance frequency information in which the appearance frequency of the sleep stage for each unit time is stored, and the transition frequency of the sleep stage for each first unit time for each divided time zone A sleep method estimation method executed by a sleep stage estimation apparatus including a model database including: a transition frequency storage unit storing the aggregated transition frequency information;
A heart rate variability feature amount calculating process for calculating a feature amount of heart rate variability for each of the first unit times for each time zone divided by the second unit time from an electrocardiogram acquired from a subject to be estimated;
An acceleration feature amount calculating step of calculating an acceleration feature amount for each first unit time from a set of accelerations acquired from a subject to be estimated;
A posture determination step of determining whether or not the subject is awake from the acceleration obtained from the subject to be estimated, based on the posture of the subject;
The calculated temporal change of the heart rate fluctuation feature amount and the temporal change of the acceleration feature amount, and the information stored in the heart rate fluctuation storage unit, the acceleration storage unit, the appearance frequency storage unit, and the transition frequency storage unit Based on the probability distribution of the sleep stage of the subject based on the sleep stage and the time zone of the immediately preceding time interval set for each of the first unit time when the awakening state is not determined in the posture determination process ; The subject is set based on the heart rate fluctuation feature amount of the current time interval and the probability based on the acceleration feature amount, and when it is determined to be awake state, the mode of the probability distribution of the sleep stage is corrected to become awakening and the subject's sleep A sleep stage estimation method comprising: estimating a stage probability distribution. The program which makes the computer with which the said sleep stage estimation apparatus equips perform the process which each means which the sleep stage estimation apparatus of Claim 1 comprises performs.