Sleep detection device and sleep detection method

Abstract

A sleep determination device that can accurately determine a REM sleep period during a sleep period without performing frequency analysis. [Solution] The system includes a data collection unit (S11) that collects heart rate data representing the heart rate per unit time during the subject's sleep period; a first calculation unit (S12) that treats the collected heart rate data during the sleep period as time series data and calculates a predicted trend value for each specified time period from the time series data as an hourly predicted trend value; a second calculation unit (S13) that calculates a trend value throughout the sleep period from the time series data as a period trend value; a third calculation unit (S14) that calculates the difference between the hourly trend value and the period trend value as an hourly sleep depth predicted value for each specified time period; and a REM determination unit (S16) that determines the REM sleep period during the sleep period based on the hourly sleep depth predicted value.

Description

[Technical Field]

[0001] The present invention relates to a sleep determination device and method for determining the REM sleep period during sleep based on the changes in a subject's heart rate. [Background technology]

[0002] Conventionally, a sleep stage determination device described in Patent Document 1 is known. This sleep stage determination device has a biosensor (pressure sensor) embedded in the mat on which the subject lies and outputs a signal corresponding to the pressure the subject exerts on the mat. The sleep stage determination device extracts a heart rate signal that fluctuates in accordance with the heart rate from the body signals that fluctuate in accordance with various bodily movements (heart rate, respiratory movement, body movement) output from the biosensor, converts it into data, and generates heart rate data. The sleep stage determination device performs frequency analysis (FFT: Fast Fourier Transform) on the heart rate data to convert it into components in the frequency domain, extracts frequency components in the range of periods from 2.5 seconds to 150 minutes (medium frequency components) from those components in the frequency domain, and converts the extracted frequency components into time domain data (IFFT: Inverse Fourier Transform). Then, the sleep stage determination device determines that the time domain data that exceeds a predetermined threshold is the REM sleep stage.

[0003] Such a sleep stage determination device can determine the duration of REM sleep during a sleep period without requiring subjects to be restrained in bed by attaching electrodes or other devices. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2017-164397 [Overview of the project] [Problems that the invention aims to solve]

[0005] The sleep stage determination device described above converts heart rate data (time-domain data) into frequency components (FFT: Fast Fourier Transform), and then inversely transforms that frequency-domain data (IFFT: Inverse Fourier Transform) to return it to time-domain data. Converting heart rate data into frequency components in this way (frequency analysis) is based on the premise that a constant wave repeats, and therefore cannot necessarily capture the actual state of heartbeats, which include complex level fluctuations and periodic fluctuations during sleep. For this reason, conventional sleep stage determination devices that include such frequency analysis processing have limitations in their detection accuracy.

[0006] This invention has been made in view of these circumstances, and provides a sleep determination device and sleep determination method that can accurately determine the REM sleep period during sleep without performing frequency analysis. [Means for solving the problem]

[0007] The sleep determination device according to the present invention includes a data acquisition unit that collects heart rate data representing the heart rate per unit time during the subject's sleep period, a first calculation unit that converts the collected heart rate data during the sleep period into time series data and calculates a predicted trend value for each predetermined time period from that time series data as an hourly predicted trend value, and a first calculation unit that converts the time series data throughout the sleep period Expressed as a time function ru A second calculation unit calculates the trend value as a period trend value, and the first calculation unit The aforementioned Every scheduled time The calculated hourly predicted trend value and the period trend value calculated by the second calculation unit The value of each time interval marked by the predetermined time. The system comprises a third calculation unit that calculates the difference between the two values ​​as an hourly sleep depth prediction value for each predetermined time interval, and a REM sleep determination unit that determines the REM sleep period during the sleep period based on the hourly sleep depth prediction value calculated by the third calculation unit.

[0008] With this configuration, heart rate data representing the heart rate per unit time during the subject's sleep period is collected. The collected heart rate data during the sleep period is treated as time series data, and a predicted trend value for each predetermined time interval is calculated from this time series data as an hourly predicted trend value. Furthermore, the time series data of the heart rate data during the sleep period is used to calculate the trend throughout the sleep period. Time function represented The trend value is calculated as the period trend value. Then, the hourly predicted trend value and the period trend value The value of each time interval marked by the predetermined time. The difference between this value and the predicted hourly sleep depth is calculated, and the REM sleep period during the sleep period is determined based on this predicted hourly sleep depth.

[0009] In the sleep determination device according to the present invention, the REM sleep determination unit includes a comparison determination unit that determines whether the hourly sleep depth prediction value is greater than a predetermined threshold, and based on the determination result of the comparison determination unit, the device can be configured to determine that the period during which the hourly sleep depth prediction value is greater than the threshold is the REM sleep period.

[0010] With this configuration, At regular intervals The hourly predicted trend value and the period trend value of The value at each time interval, which is set within the predetermined time interval. The difference between the predicted hourly sleep depth and the predicted hourly sleep depth is determined to determine whether the predicted hourly sleep depth is greater than a predetermined threshold. The period during which the predicted hourly sleep depth is greater than the predetermined threshold, i.e., the trend of the heart rate estimated for each predetermined period (which can be represented by the predicted hourly trend value), is determined to be the trend of the heart rate throughout the entire sleep period (the period trend value). (Time function) A period that is larger than a certain threshold (which can be expressed by) is determined to be the REM sleep period.

[0011] The hourly predicted trend value μ mentioned above. t This is the time-series data of heart rate y t from, μ t =μ t-1 +δ t-1 δ t =δ t-1 +ζ t ζ t ~Normarl(0,σ 2 ζ ) δ t : The first-order difference (μ t -μ t-1 ) at time point t ζ t : Normal white noise According to Under the following assumptions, δ t The likelihood of δ t ~Normal(δ t-1 , σ 2 ζ ) The likelihood of the heart rate value y t is y t ~Normal(μ t , δ 2 v ) Can be calculated by iterative calculation.

[0012] The above-mentioned period trend value Yt uses the following matrices X and θ, X: Each row is 1, y t , y t 2 θ: [c, b, a] T However, T is the transpose, a, b, c: Parameters (X T ... X)θ = X T Y t From, using θ calculated by the least squares method, As a function of time Can be calculated.

[0013] ... The sleep determination method according to the present invention includes a data collection step of collecting heartbeat data representing the heartbeat per unit time during the sleep period of the subject, a first calculation step of using the collected heartbeat data during the sleep period as time series data and calculating a predicted trend value for each predetermined time as an hourly predicted trend value from the time series data, and from the time series data through the sleep period time Expressed as a function It should be noted that there seems to be some incomplete or unclear parts in the original text (such as the ellipsis in the middle), but the translation is carried out as accurately as possible according to the existing content.A second calculation step in which the trend value is calculated as the period trend value, and by the first calculation step Every predetermined time The calculated hourly predicted trend value and the period trend value calculated in the second calculation step The value of each time interval marked by the predetermined time. A third calculation step calculates the difference between the above as the predicted hourly sleep depth value for each predetermined time interval, and the hourly sleep depth calculated by the third calculation step degree Forecast Measurement The configuration includes a REM determination step that determines the REM sleep period during the sleep period based on the above.

[0014] Furthermore, in the sleep determination method according to the present invention, the REM determination step includes a comparison determination step that determines whether the hourly sleep depth prediction value is greater than a predetermined threshold, and based on the determination result in the comparison determination step, the period during which the hourly sleep depth prediction value is greater than the threshold is determined to be the REM sleep period. [Effects of the Invention]

[0015] The sleep determination device and sleep determination method according to the present invention can accurately determine the REM sleep period during sleep without performing frequency analysis. [Brief explanation of the drawing]

[0016] [Figure 1] Figure 1 is a block diagram showing the sleep determination device according to the present invention. [Figure 2] Figure 2 is a graph showing the trend of heart rate data, which represents the heart rate every minute. [Figure 3] Figure 1 is a flowchart showing the processing flow in the sleep stage determination unit of the sleep detection device. [Figure 4] Figure 4 is a graph showing the changes in the hourly predicted trend value μt obtained based on heart rate data during the sleep period. [Figure 5]Figure 5 compares the results of determining the REM sleep period from hourly sleep depth predictions (a) with the results of determining the sleep stage based on sleep depth estimation using an electroencephalograph (EEG) (b). [Figure 6] Figure 6 is a diagram that numerically shows the REM sleep determination results in an embodiment of the present invention and the REM sleep determination results based on sleep depth estimation using an electroencephalograph. [Modes for carrying out the invention]

[0017] Embodiments of the present invention will be described below with reference to the drawings.

[0018] A sleep determination device according to one embodiment of the present invention is configured as shown in Figure 1.

[0019] In Figure 1, the sleep detection device includes a sensor pad 60 installed on a bed 50 and an information processing device 10 such as a personal computer. The subject S (person) lies on the bed 50 with their back pressing against the sensor pad 60. A sensor (piezoelectric element) is installed inside the sensor pad 60, and a detection signal corresponding to the movement (vibration) of the subject S transmitted to the sensor through the sensor pad 60 is output from the sensor (piezoelectric element).

[0020] The information processing device 10 is a computer composed of hardware and software, and includes a biosignal processing unit 11, a storage unit 12, and a sleep stage determination unit 13. The biosignal processing unit 11 receives a detection signal from a sensor (piezoelectric element) in the sensor pad 60 that changes in accordance with the subject S's physical movements (body movements such as turning over in bed, respiratory movements, and heartbeats). The biosignal processing unit 11 extracts a heart rate signal that fluctuates in level according to the heartbeat by applying impedance conversion and filtering to the detection signal from the sensor pad 60 (sensor). The biosignal processing unit 11 then generates heart rate data from the heart rate signal, representing the number of heartbeats per unit time, for example, every minute. The heart rate data (heart rate / minute) over a predetermined period of time including the sleep period changes, for example, as shown in Figure 2. The biosignal processing unit 11 stores the heart rate data (heart rate / minute) over the predetermined period of time including the sleep period in the storage unit 12.

[0021] The sleep stage determination unit 13 uses the heart rate data (heart rate / minute) stored in the memory unit 12 to determine the REM sleep period during the sleep period. Specifically, the sleep stage determination unit 13 performs the processing according to the procedure shown in Figure 3, for example.

[0022] In Figure 3, the sleep stage determination unit 13 acquires heart rate data from the memory unit 12 during the sleep period (from the start of sleep to the time of waking) (S11: data acquisition unit (data acquisition step)). The heart rate data y during the sleep period t This is used as time-series data, and from this time-series data, a predicted trend value is obtained at predetermined intervals, for example, every minute, as the hourly predicted trend value μ. t This is calculated as (S12: First calculation unit (first calculation step)). This hourly predicted trend value μ t This can represent the estimated heart rate trend at predetermined intervals, for example, every minute, without being affected by heart rate variability (period) of up to twice the length of the predetermined interval (1 minute).

[0023] Hourly predicted trend value μ t Specifically, μ t =μ t-1 +δ t-1 δt =δ t-1 +ζ t ζ t ~Normarl(0,σ 2 ζ ) δ t : First-order difference at time t (μ t -μ t-1 ) size ζ t Regular white noise According to, Under the following assumptions, δ t The likelihood of is δ t ~normal (δ t-1 , σ 2 ζ ) Heart rate value y t The likelihood of is y t ~Normal(μ t , δ 2 v ) It can be calculated through iterative calculation.

[0024] As a result of the above calculation, for example, the fluctuating heart rate data y as shown in Figure 2. t (Time-series data) shows the fluctuating hourly predicted trend value μ as shown in Figure 4. t You can obtain this.

[0025] Next, the sleep stage determination unit 13 uses the time-series data (heart rate data) to determine the sleep stage throughout the sleep period. Expressed as a time function Trend value Period Trend value Y t The trend value Y for this period is calculated as follows (S13: Second calculation unit (second calculation step)). t This can represent the trend in heart rate throughout the entire sleep period, unaffected by heart rate variability (periods) beyond the duration of sleep. Period Trend Value Y t This is done using the following matrices X and θ: X: 1 each row, y t , y t 2 θ:[c, b, a] THowever, T is transposed. a, b, c: Parameters (X T X)θ=X T Y t Using θ calculated by the least squares method, As a function of time It can be calculated.

[0026] The sleep stage determination unit 13 is Every minute The hourly predicted trend value μt (see S12) mentioned above and Expressed as a time function Period Trend Value Yt The value for each time interval within that one minute. to difference (μt-Yt) is calculated as the hourly sleep depth prediction value every minute (S14: Third calculation unit (third calculation step)). Then, the sleep stage determination unit 13 determines the REM sleep period during the sleep period based on the hourly sleep depth prediction value (μt-Yt) (REM determination unit, REM determination step).

[0027] The REM sleep period is determined as follows (S15-S20).

[0028] The sleep stage determination unit 13 determines the start time of the sleep period t start Set (S15), and at that point t start Hourly sleep depth prediction (μt start -Yt start The sleep stage determination unit 13 determines whether the hourly sleep depth prediction value (μtstart-Yt) is greater than a predetermined threshold TH (S16: comparison determination unit (comparison determination step) · REM determination unit (REM determination step)). start If ) is greater than the threshold TH (μt start -Yt start >TH: S16 YES), at that point t start This is determined to be the REM sleep period (S17). On the other hand, the hourly sleep depth prediction value (μt start -Yt start ) If the threshold TH is less than or equal to (μt start -Yt start ≤TH:S16 (NO), the sleep stage determination unit 13 determines the sleep stage at that time t startIt is determined that it is not the REM sleep period (NonRem) (S18).

[0029] Thereafter, the sleep stage determination unit 13 confirms that t has not reached the end time t end of the sleep period (NO in S19), increments t by 1 minute (1 min) each time (t = t + 1 min: S20), and as described above, repeatedly determines whether the hourly sleep depth prediction value (μ t -Y t ) at that time t is greater than the threshold value TH (S16: comparison determination unit · REM determination unit). In this process, when the hourly sleep depth prediction value (μ t -Y t ) is greater than the threshold value TH (μ t -Y t > TH: YES in S16), it is determined that the time t at that time is the REM sleep period Rem (S17). On the other hand, when the hourly sleep depth prediction value (μ t -Y t ) is less than or equal to the threshold value TH (μ t -Y t ≦ TH: NO in S16), it is determined that the time t at that time is not the REM sleep period (NonRem) (S18).

[0030] In the process of the above-described processing, when t updated by 1 minute each time reaches the end time t end of the sleep period (YES in S19), the sleep stage determination unit 13 ends a series of processing.

[0031] As a result of the above-described processing, an hourly sleep depth prediction value (μ t -Y t ) that varies as shown in Fig. 5(a) during the sleep period is obtained. From the relationship between the hourly sleep depth prediction value (μ t -Y t ) and the threshold value TH (μ t ―Y t > TH), periods T1, T2, T3, T4, and T5 during the sleep period are determined as the REM sleep period. This determination result generally agrees with the sleep depth estimation result (depth 4) by the electroencephalogram shown in Fig. 5(b). In Fig. 5(b), W represents wakefulness, depth 4 represents REM sleep, and depths 3 to 0 represent non-REM sleep.

[0032] Furthermore, a numerical comparison of the estimation results using the above-described process (example) and the sleep depth estimation results using an electroencephalograph (EEG) is shown in Figure 6. In Figure 6, the EEG sleep depth estimation results estimated that 87 out of 355 time points (5 hours and 55 minutes of sleep) were REM sleep (Rem), whereas the EEG sleep depth estimation results estimated that 84 out of 355 time points were REM sleep (Rem) (precision = 96.6%). Also, the EEG sleep depth estimation results estimated that 244 out of 355 time points were non-REM sleep (NonRem), whereas the EEG sleep depth estimation results estimated that 229 out of 355 time points were non-REM sleep (NonRem) (precision = 93.9%).

[0033] According to the sleep determination device of the present invention as described above, the REM sleep period during sleep can be accurately determined without performing frequency analysis.

[0034] Furthermore, in the process described above, the threshold TH used to determine the REM sleep period (see S16 in Figure 4) can be determined to an appropriate value based on, for example, the sleep depth determination results from an electroencephalograph.

[0035] In the process described above, the heart rate data represented the heart rate every minute, but it is not limited to that; it may also represent the heart rate at predetermined intervals other than every minute. Similarly, the hourly predicted trend value μ t Furthermore, the hourly sleep depth prediction values ​​may also represent values ​​for each predetermined time interval other than one minute (predicted trend value, sleep depth prediction value).

[0036] Hourly predicted trend value μ t The value was obtained by iterative calculation according to the formula described above, but is not limited to this. A trend value representing the heart rate trend may be obtained at predetermined intervals, for example, every minute, according to other calculation methods.

[0037] Also, This is a trend value expressed as a time function. The period trend value Yt was obtained by calculation according to the formula described above, but is not limited to this; a trend value representing the heart rate trend throughout the entire sleep period may be obtained by other calculation methods.

[0038] Although embodiments of the present invention have been described above, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments described above can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments are included in the scope and spirit of the invention, as well as in the invention described in the claims. [Industrial applicability]

[0039] The sleep determination device and sleep determination method according to the present invention have the effect of being able to accurately determine the REM sleep period during sleep without performing frequency analysis, and are useful as a device and method for determining the REM sleep period during sleep based on the heartbeat of a subject. [Explanation of Symbols]

[0040] 10 Information Processing Devices 11. Biosignal Processing Unit 12 Storage section 13 Sleep stage determination unit 50 beds 60 Sensor Pads

Claims

1. A data acquisition unit that collects heart rate data representing the heart rate per unit time during the subject's sleep period, A first calculation unit collects heart rate data during the sleep period and uses it as time-series data, and calculates predicted trend values ​​for predetermined time intervals from this time-series data as hourly predicted trend values. A second calculation unit calculates a period trend value from the aforementioned time series data as a time function over the sleep period, A third calculation unit calculates the difference between the hourly predicted trend value calculated by the first calculation unit at predetermined intervals and the period trend value calculated by the second calculation unit at each time interval in the predetermined interval as the hourly sleep depth predicted value for each predetermined interval, A sleep determination device comprising: a REM sleep determination unit that determines the REM sleep period during the sleep period based on the hourly sleep depth prediction value calculated by the third calculation unit.

2. The REM determination unit, Includes a comparison determination unit that determines whether the predicted hourly sleep depth value is greater than a predetermined threshold, The sleep determination device according to claim 1, wherein, based on the determination result of the comparison determination unit, the device determines that the period during which the predicted hourly sleep depth is greater than the threshold is the REM sleep period.

3. The first calculation unit processes the time-series data y of the heart rate. ｔ From hourly predicted trend value μ ｔ of m ｔ =μ ｔ－１ +d ｔ－１ d ｔ =d ｔ－１ +g ｔ g t ~Normarl(0,σ ２ ζ ) σ ｔ : First-order difference at time t (μ ｔ -μ ｔ－１ ) size ζ ｔ : Regular white noise According to, Under the following assumptions, δ ｔ The likelihood of is δ ｔ ~Normal (δ ｔ－１ , σ ２ ζ ) Heart rate value y ｔ The likelihood of is y ｔ ~Normal (μ ｔ , δ ２ ｖ ) A sleep determination device according to claim 1 or 2, which calculates by iterative calculation.

4. The second calculation unit uses the following matrices X and θ: X: Each row 1, y ｔ y ｔ ２ θ: [c, b, a] Ｔ T is transposed a, b, c: Parameters The trend value Y for the aforementioned period ｔ The following equation (X Ｔ X)θ=X Ｔ Y ｔ The sleep determination device according to claim 3, wherein θ is calculated by the least squares method and then calculated as a function of time.

5. A data collection step to collect heart rate data representing the heart rate per unit time during the subject's sleep period, A first calculation step involves collecting heart rate data during the sleep period and converting it into time-series data, and calculating predicted trend values ​​for predetermined time intervals from that time-series data as hourly predicted trend values, A second calculation step involves calculating a period trend value from the aforementioned time series data as a time function over the sleep period, A third calculation step calculates the difference between the hourly predicted trend value calculated at predetermined intervals by the first calculation step and the period trend value calculated by the second calculation step at each time interval in the predetermined time, as the hourly sleep depth predicted value for each predetermined time interval. A sleep determination method comprising: a REM determination step for determining the REM sleep period during the sleep period based on the hourly sleep depth prediction value calculated by the third calculation step.

6. The aforementioned REM determination step is, The comparison determination step includes determining whether the predicted hourly sleep depth value is greater than a predetermined threshold, The sleep determination method according to claim 5, wherein, based on the determination result in the comparison determination step, the period during which the predicted hourly sleep depth is greater than the threshold is determined to be the REM sleep period.

Images