JP6932351B2 - Sleep stage determination device, sleep stage determination method and program - Google Patents

Description

The present invention relates to a sleep stage determination device for determining a REM sleep state, a sleep stage determination method, and a program for executing sleep stage determination.

In the medical field, the sleep stage of the subject is measured in order to diagnose the deterioration of sleep quality due to sleep disorders and the like. Human sleep stages are known to be classified into 6 stages from the viewpoint of sleep depth, and the 6 sleep stages are awakening, REM sleep, and non-REM sleep (stages 1 to 1) in order from the lightest sleep stage. It is called 4). Conventionally, these six stages of sleep are determined by attaching a large number of electrodes to the face or head of the person to be measured and measuring brain waves, eye movements, and jaw myoelectricity from the large number of electrodes. It was done by analyzing the results.

Such an examination at the sleep stage with a large number of electrodes attached to the face or head is usually an examination performed by staying at a medical institution and continuously attaching the electrodes to the body for a long period of time. Because it sleeps in a different situation from the above, it imposes a mental burden and a physical burden on the subject (patient). Further, the data acquired by using the electrodes needs to be analyzed and judged by a doctor having specialized knowledge and experience, and there is a problem that the sleep state cannot be easily judged.

In order to solve these problems, many sleep stage estimation methods that do not require diagnosis by a specialist doctor have been proposed.
For example, Patent Document 1 describes a technique for estimating a sleep stage from detection data of a mattress type pressure sensor by a method called Database-based Compact Genetic Algorithm, which is an improvement of a learning method using a genetic algorithm. The technique described in Patent Document 1 estimates the sleep stage based on the body movement and heartbeat of the subject detected by the mattress type pressure sensor. By estimating the sleep stage using such a mattress type pressure sensor, it is possible to estimate the sleep state of the person to be measured without imposing a burden on the person to be measured.

Japanese Unexamined Patent Publication No. 2014-239789

By the way, when considering the above-mentioned 6 stages of sleep, there are awakening, REM sleep, and 4 non-REM sleeps (stages 1 to 4) in order from the light sleep stage, and these 6 stages are detection data. There is a problem that it cannot be determined by a simple process such as dividing the above into six ranges. In particular, REM sleep is a sleep state completely different from non-REM sleep and awakening, and it has been very difficult to accurately determine REM sleep in the past.
As described in Patent Document 1, if the sleep stage can be determined by using a pressure sensor such as a mattress type pressure sensor, it is possible to diagnose the sleep state without imposing a burden on the person to be measured. However, the conventionally proposed method has a problem that the determination accuracy of REM sleep is not high, and as a result, the reliability of estimation of the sleep stage is not so good.

An object of the present invention is to make it possible to accurately determine the REM sleep of a person to be measured without imposing a burden on the person to be measured.

In order to solve the above problems, the sleep stage determination device of the present invention includes a data processing unit and a sleep stage determination unit.
The sleep stage determination unit determines that the start of the first REM sleep period is started when the rate of increase in the median heart rate of the subject for each first period is equal to or higher than a predetermined threshold, and every first period. When the rate of decrease in the median heart rate of REM sleep decreases, the end of the first REM sleep period is determined, and the body movements of the subject are frequency-analyzed for each second period. A plurality of decision trees for which a judgment is obtained are generated by machine learning, and it is determined that the second period is the second REM sleep period from the majority decision of whether or not the determination tree is REM sleep obtained from the generated plurality of decision trees. ..
The sleep stage determination unit includes a period in which the period determined to be the first REM sleep period and the period determined to be the second REM sleep period overlap, and the period determined to be the second REM sleep period within a certain interval is If present, the period determined to be the second REM sleep period within the fixed interval is determined as the REM sleep period.

In addition, the sleep stage determination method of the present invention includes a first REM sleep period determination process, a second REM sleep period determination process, and a sleep stage determination process.
The first REM sleep period determination process determines that the start of the first REM sleep period is started when the rate of increase in the median heart rate of the subject for each first period is equal to or greater than a predetermined threshold, and the first REM sleep period is determined. When the rate of decrease in the median heart rate for each period decreases, the end of the first REM sleep period is determined.
The second REM sleep period determination process is an analysis of data obtained by frequency-analyzing the body movements of the subject for each second period, and generates a plurality of determination trees for obtaining the probability value of REM sleep by machine learning. From the majority decision of whether or not it is REM sleep obtained from a plurality of determination trees, it is determined that the second period is the second REM sleep period.
The sleep stage determination process includes a period in which the period determined to be the first REM sleep period and the period determined to be the second REM sleep period overlap, and the period determined to be the second REM sleep period within a certain interval is If present, the period determined to be the second REM sleep period within the fixed interval is determined as the REM sleep period.

Further, the program of the present invention causes a computer to execute a procedure for executing the first REM sleep period determination process, the second REM sleep period determination process, and the sleep stage determination process of the sleep stage determination method. ..

According to the present invention, the duration of REM sleep can be accurately determined based on the pressure data during sleep of the subject obtained from the pressure sensor.

It is a block diagram which shows the structural example of the sleep stage determination apparatus by one Embodiment of this invention. It is a figure which shows the example of the determination state of the sleep stage by the example of one Embodiment of this invention. It is a block diagram which shows the hardware configuration example of the sleep stage determination apparatus of one Embodiment of this invention. It is a flowchart which shows the determination processing example of the heartbeat component by one Embodiment of this invention. It is a flowchart which shows the determination processing example of all frequency components by one Embodiment of this invention. It is a figure which shows the determination processing example of REM sleep by one Embodiment of this invention. It is a flowchart which shows the determination processing example of REM sleep by one Embodiment of this invention. It is a figure which shows the determination example of REM sleep by one Embodiment of this invention. It is a figure which shows the example which determined the REM sleep from the actual detection data by one Embodiment of this invention.

Hereinafter, an example of an embodiment of the present invention (hereinafter, referred to as “this example”) will be described with reference to the accompanying drawings.
[1. Configuration of sleep stage determination device]
FIG. 1 is a block diagram showing a configuration of the sleep stage determination device 10 of this example. FIG. 2 is a diagram showing an example of a state in which a sleep stage is determined using the sleep stage determination device 10 of this example.
The sleep stage determination device 10 of this example acquires the body movement of the person to be measured as pressure data by the mattress sensor 2. As shown in FIG. 2, for example, the mattress sensor 2 is laid on the mattress of the bed 1 in which the subject A sleeps, and detects the body movement of the upper body of the subject A during sleep as a change in pressure. .. The mattress sensor 2 is placed on the mattress below the person A to be measured as an example. For example, the mattress sensor 2 may be built in the mattress.
FIG. 2 shows an example in which the sleep stage determination device 10 is installed beside the bed 1 and the mattress sensor 2 and the sleep stage determination device 10 are connected by a cable. For example, the pressure data acquired by the mattress sensor 2 is wirelessly transmitted. It may be transmitted to the sleep stage determination device 10 in another room.

As shown in FIG. 1, the sleep stage determination device 10 includes a biological data acquisition unit 11, a biological data processing unit 12, a sleep stage determination unit 13, and an output unit 14.
The biological data acquisition unit 11 acquires the pressure data output by the mattress sensor 2. The pressure data acquired by the biological data acquisition unit 11 is supplied to the data processing unit 12. The biological data processing unit 12 analyzes the supplied pressure data. Specifically, the pressure data is sampled and converted into digital data, and the components of each frequency are acquired by a high-speed Fourier transform process (hereinafter, referred to as “FFT process”) of the digitalized pressure data. Then, the state of heartbeat and the state of respiration are discriminated from the state of each acquired frequency component. That is, the heart rate and the respiration rate are discriminated from the frequency components corresponding to the heartbeat and the respiration among the frequency components obtained by FFT processing the pressure data.
Further, the biological data processing unit 12 uses the frequency component obtained by the FFT processing to perform data processing necessary for the analysis of the sleep stage.

Then, the result analyzed by the biological data processing unit 12 is supplied to the sleep stage determination unit 13. The sleep stage determination unit 13 determines the sleep state of the person to be measured A based on the result analyzed by the biological data processing unit 12. The sleep stage determination unit 13 determines the sleep state of REM sleep and the sleep state other than REM sleep. Further, the sleep stage determination unit 13 determines five stages of awakening and non-REM sleep (stages 1 to 4) as sleep states other than REM sleep. In the following description, the process of determining the sleep state of REM sleep and the sleep state other than REM sleep will be mainly described, and the sleep states other than REM sleep will be divided into 5 stages (awakening and non-REM sleep stages 1 to 1). The description of the process for determining in 4) will be omitted. The sleep stage of non-rem sleep is considered in association with the depth of sleep, and indicates that sleep becomes deeper in the order of stages 1 to 4. For example, stage 2 sleeps deeper than stage 1, and when it is described as "stage 2 or less", it indicates stages 2 to 4.

The output unit 14 outputs the sleep stage determined by the sleep stage determination unit 13. The output unit 14 is composed of, for example, a display device, and displays the sleep stage. Alternatively, the output unit 14 may be configured as a recording device to record the sleep state overnight.

[2. Hardware configuration example of sleep stage determination device]
FIG. 3 shows an example of hardware configuration when the sleep stage determination device 10 is configured by a computer device.
The computer device C includes a CPU (Central Processing Unit) C1, a ROM (Read Only Memory) C2, and a RAM (Random Access Memory) C3 connected to the bus C8. Further, the computer device C includes a non-volatile storage C4, a network interface C5, an input device C6, and a display device C7.

The CPU C1 reads out from the ROM C2 the program code of the software that realizes each function included in the biological data processing unit 12 and the sleep stage determination unit 13 of the sleep stage determination device 10 and executes the program code. Regarding the FFT process for frequency analysis of pressure data, the CPU C1 executes the FFT process by reading the program for executing the corresponding process from the ROM C2. Variables, parameters, etc. generated during the arithmetic processing are temporarily written in the RAM C3.

As the non-volatile storage C4, for example, HDD (Hard disk drive), SSD (Solid State Drive), flexible disk, optical disk, optical magnetic disk, CD-ROM, CD-R, magnetic tape, non-volatile memory and the like are used. Be done. In this non-volatile storage C4, in addition to the OS (Operating System) and various parameters, a program for making the computer device C function as the sleep stage determination device 10 is recorded. Further, the data about the sleep stage determined by the sleep stage determination unit 13 is recorded in the non-volatile storage C4.

For the network interface C5, for example, a NIC (Network Interface Card) or the like is used, and various data can be transmitted / received via a LAN (Local Area Network) to which terminals are connected, a dedicated line, or the like. For example, the computer device C acquires the pressure data output by the mattress sensor 2 via the network interface C5. The input device C6 is composed of a device such as a keyboard, for example, and the input device C6 sets a period for determining the sleep stage by the sleep stage determination device 10, and instructs the display form of the sleep stage. The display device C7 displays the sleep stage determined by the sleep stage determination device 10.
It should be noted that the sleep stage determination device 10 is configured from a computer device as an example, and data processing such as FFT processing may be performed by preparing dedicated hardware.

[3. REM sleep judgment processing]
Next, the REM sleep determination process performed by the sleep stage determination device 10 of this example will be described.
The sleep stage determination device 10 of this example performs REM sleep determination processing (first REM sleep period determination processing and second REM sleep period determination processing) by two types of methods, and REM sleep by the two types of methods. The judgment result of the above is comprehensively judged, and finally the judgment result of the period of REM sleep is obtained. Here, the two types of methods are referred to as method 1 and method 2.
The details of the two types of methods will be described later, but the method 1 is a REM sleep determination method focusing on heart rate fluctuation, and the method 2 is a REM sleep determination method by machine learning using all frequency components. The sleep stage determination device 10 determines REM sleep with a hybrid structure using these two methods.

First, the REM sleep determination process according to Method 1 will be described.
In this method 1, the medium frequency component of the heart rate variability is estimated in real time from the heart rate information for a certain period of time using a trigonometric function model (TFM), and the sleep depth is estimated from the calculated heart rate variability. Here, REM sleep is estimated using the physiological viewpoint that the heart rate fluctuates during REM sleep.

FIG. 4 is a flowchart showing the determination process of REM sleep by the method 1.
First, the biological data processing unit 12 acquires a heartbeat value from the data frequency-analyzed by the FFT process. The biological data processing unit 12 calculates the median HR at regular time intervals (here, 5 minutes) for the detected heartbeat (step S111).
Then, the biological data processing unit 12 calculates the fluctuation amount ΔHR every 5 minutes of the detected median heart rate HR, and compares the fluctuation amount ΔHR every 5 minutes with the threshold value TH1 (step S112).
Here, when the median heart rate for 5 minutes immediately before the estimated time is ΔHR _curr and the median heart rate from 5 minutes to 10 minutes before is ΔHR _prev , the amount of fluctuation is determined by the following equation [Equation 1]. Obtain ΔHR.

In step S112, the variation amount ΔHR obtained in this way every 5 minutes is compared with the threshold value TH1. In the comparison in step S112, the threshold value TH1 is set to 0.04, and it is determined whether or not the condition of ΔHR> 0.04 is satisfied.
If the condition of ΔHR> 0.04 is not satisfied in step S112 (NO in step S112), the biometric data processing unit 12 waits until the corresponding condition is satisfied.
Then, when it is determined that the condition of ΔHR> 0.04 is satisfied (YES in step S112), the biological data processing unit 12 determines that the REM sleep period (REM sleep period according to the method 1) has started (step S113). After that, the biological data processing unit 12 determines whether or not the reduction rate of the fluctuation amount ΔHR has decreased (step S114). In addition, "decrease rate" means, for example, that the value of ΔHR changes from 0.07 to 0.06. Here, when the decrease rate of the fluctuation amount ΔHR does not decrease (NO in step S114), the biological data processing unit 12 waits until the decrease rate of the fluctuation amount ΔHR decreases. Further, when the decrease rate of the fluctuation amount ΔHR decreases (YES in step S114), the biological data processing unit 12 determines the end of the REM sleep period (REM sleep period according to the method 1) (step S115), and in step S112. Returning to the comparison between the fluctuation amount ΔHR and the threshold value TH1.
As described above, in the method 1, the period of REM sleep can be determined based on the fluctuation amount of the heart rate.

Next, the REM sleep determination process according to the method 2 will be described.
This method 2 is a method of determining REM sleep and non-REM sleep by using Random Forests, which is a machine learning method based on ensemble learning. In this method 2, the pressure data acquired by the mattress sensor 2 for a certain period (here, 30 seconds) is subjected to FFT processing to perform frequency analysis to determine REM sleep.

FIG. 5 is a flowchart showing the determination process of REM sleep by the method 2.
First, the biometric data processing unit 12 divides the biometric data acquired by the biometric data acquisition unit 11 at regular intervals (1 epoch section: 30 seconds), and performs frequency analysis on the biometric data at regular intervals by FFT processing. By doing so, biometric data expressed in frequency is generated (step S121). Then, all frequency components of the FFT-processed biological data are normalized so that the sum of all frequency components becomes 1. Note that normalization is performed here as an example, and data that is not normalized may be handled.
Then, the biological data processing unit 12 uses the pressure data at regular intervals to perform binary determination by a machine learning method of random forest (step S122). The details of binary judgment by machine learning of random forest will be described later.

After that, the biological data processing unit 12 determines whether or not the condition of REM sleep is satisfied by the binary determination in step S122 (step S123). If the condition of REM sleep is not satisfied by this determination (NO in step S123), the process returns to the process of step S121, and the biometric data processing unit 12 moves to the process for the next fixed period.
Then, when the condition of REM sleep is satisfied by this determination (YES in step S123), the biological data processing unit 12 sets the corresponding fixed period (30 seconds) into the REM sleep period (REM sleep period according to the method 2). The setting (step S124) is performed, and then the process returns to the process of step S121, and the process proceeds to the next fixed period process.

FIG. 6 is a diagram for explaining a binary determination state by machine learning of a random forest.
First, as shown on the upper side of FIG. 6, the biological data processing unit 12 acquires the pressure data D1 for 30 seconds. The vertical axis of the graph of the pressure data D1 is the sensor value, and the horizontal axis is the time (30 seconds).
The biometric data processing unit 12 performs FFT processing on the pressure data D1 to obtain FFT data D2 showing the intensity value for each frequency component. The vertical axis of the graph of FFT data D2 is the intensity value, and the horizontal axis is the frequency (0.016 Hz to 5.0 Hz). Here, the FFT data D2 is 300-dimensional vector data containing only frequency components of 5 Hz or less.

The biological data processing unit 12 generates a plurality of decision trees, which are prediction models, from the data of the 300-dimensional vector by using a machine learning algorithm called a random forest. That is, as shown in FIG. 6, a plurality of decision trees T ₁ , T ₂ , ..., _TN (N is an arbitrary integer) are generated. Each decision tree branches based on what the state of each frequency component is, and obtains the judgment of each sleep stage (here, the REM sleep stage). For example, as shown in the lower left of FIG. 6, determining the tree T ₁ of a certain stage Sa, normalized values of the frequency components 0.3Hz will determine 0.25 or less, in the determination, normalization value When is less than 0.25, it is determined to be unconfirmed, and when it is 0.25 or more, it is determined to be REM sleep. Then, in the undetermined case (when the normalized value of the 0.3 Hz component is smaller than 0.25), whether or not the normalized value of the frequency component 1.5 Hz is greater than 0.05 in the next step Sb. When the normalized value is greater than 0.05, it is determined to be non-REM sleep, and when the normalized value is 0.05 or less, it is determined to be REM sleep.

In this way, the decision tree T by analyzing the FFT data D2 based on _1, to obtain a judgment q ₁ whether REM sleep. Further, by analyzing the FFT data D2 based on the decision tree T ₂ under a condition different from that of the decision tree T _{1 , the determination q 2 of} whether or not it is REM sleep is obtained. In this way, for each decision tree T ₁ , T ₂ , ..., _{TN under different conditions, the determination q 1} , q ₂ , ..., Q _N of whether or not it is REM sleep is obtained.
Then, biometric data processing unit 12, a plurality of decision trees _{T 1,} T 2, · · ·, determines whether REM sleep obtained for each _{T N q 1, q 2,} ···, the _{q N} majority Then, it is determined whether or not the current period of 30 seconds is REM sleep.

FIG. 7 is a flowchart showing a processing example of determining the REM sleep period by using the section determined to be REM sleep by the method 1 and the section determined to be REM sleep by the method 2. The process of determining the REM sleep period is determined by the sleep stage determination unit 13.
First, the sleep stage determination unit 13 determines whether or not there is a section determined to be REM sleep by the method 1 (step S131). In this determination, if there is no section determined to be REM sleep by the method 1 (NO in step S131), the sleep stage determination unit 13 determines that there is no REM sleep period within the searched period, and determines that there is no REM sleep period. The process of searching for a period ends.
Then, when there is a section determined to be REM sleep by the method 1 (YES in step S131), the sleep stage determination unit 13 determines the section determined to be REM sleep by the method 1 and the section determined to be REM sleep by the method 2. It is determined whether or not there is a section that overlaps with (step S132).

Further, if there is a section in which the section determined to be REM sleep by the method 1 and the section determined to be REM sleep by the method 2 overlap in the judgment of step S132 (YES in step S132), the REM sleep is determined by the method 1. All the periods including the determined section and the section determined to be REM sleep by the method 2 exist within a certain period (here, within 2 minutes) are determined as the REM sleep period (step S133). If the section determined to be REM sleep by method 1 and the section determined to be REM sleep by method 2 do not overlap in the determination of step S132 (NO in step S132), it is determined to be REM sleep by method 1. The section is determined as the REM sleep period (step S134).

FIG. 8 is an explanatory diagram showing an example of a processing state for determining the period of REM sleep from the determination result in the method 1 and the determination result in the method 2 shown in the flowchart of FIG. 7.
The example of FIG. 8A shows a case where there is a section determined to be REM sleep by the determination by the method 1 in a certain period, and there is no section determined to be REM sleep by the determination by the method 2 in the period overlapping the section. In the case of the state shown in FIG. 8A, the sleep stage determination unit 13 determines the section determined to be REM sleep by the method 1 as the REM sleep section. The confirmation process of FIG. 8A corresponds to the process of determining REM sleep in step S134 of the flowchart of FIG. 7.

The example of FIG. 8B shows a case where there is a section determined to be REM sleep by the determination by the method 1 in a certain period, and there is a section determined to be REM sleep by the determination by the method 2 in the period overlapping the section. In this case, the sleep stage determination unit 13 determines all the sections determined to be REM sleep by the determination by the method 1 as the REM sleep period, and also determines the REM sleep period for the sections determined to be REM sleep by the determination by the method 2. Is confirmed. The section determined to be REM sleep by the determination by the method 2 here is defined as a continuous REM sleep period during which there is a section determined to be REM sleep within 2 minutes by the method 2. The confirmation process of FIG. 8B corresponds to the process of determining REM sleep in step S133 of the flowchart of FIG. 7.

The example of FIG. 8C shows a case where there is a section determined to be REM sleep by the determination by the method 2 in a certain period, and the corresponding section is not determined to be REM sleep in the method 1. In this case, the sleep stage determination unit 13 does not determine the corresponding section as REM sleep.

FIG. 9 shows an example of determining the sleep state of the person to be measured. The horizontal axis of the characteristic BW and the characteristic HR in FIG. 9 indicates the time, and in this FIG. 9, the characteristic BW and the characteristic HR during sleep for several hours are shown.
The characteristic BW shown on the upper side of FIG. 9 shows an example in which the sleep state (sleep stage) of the subject is measured by an electroencephalograph (that is, an example measured by a device different from the sleep stage determination device 10 of this example). .. On the vertical axis of the graph of this characteristic BW, the positions indicated by A indicate arousal, and the positions indicated by the numbers 1, 2, 3 and 4 indicate stages 1 to 4 of non-REM sleep. In the characteristic BW shown in FIG. 9, the period of REM sleep is not determined, but REM sleep exists between stage 1 of non-REM sleep and awakening.

Here, the sections R1, R2, R3, R4, and R5 shown overlaid on the characteristic BW of FIG. 9 are REM sleep periods determined by the sleep stage determination device 10 of this example from the determination results of both method 1 and method 2. be.

The characteristic HR shown at the bottom of FIG. 9 shows the change in heart rate. The characteristic HR indicating the heart rate is acquired by the sleep stage determination device 10 based on the pressure data detected by the mattress sensor 2 shown in FIG. Within the characteristic HR showing the fluctuation of the heart rate, the positions of the peaks that are temporarily higher than the heart rate before and after the heart rate ra1, ra2, ra3, ra4, ra5 (almost the central part of the part indicated by the broken line circle in FIG. 9). ) Corresponds to the section determined to be REM sleep by the method 1.
The sections ra1, ra2, ra3, ra4, and ra5 determined to be REM sleep in the method 1 are positions where the heart rate of the characteristic HR peaks, and are relatively short sections.

Further, each of the plurality of sections rb shown in FIG. 9 is a section determined to be REM sleep by the method 2. In FIG. 9, the section rb determined to be REM sleep by the method 2 is shown only in the vicinity of the REM sleep periods R4 and R5, and the other sections are not shown.
In the case of this example, one section rb determined to be REM sleep by the method 2 is 30 seconds, and when the 30-second section rb exists within 2 minutes, the REM sleep determined by the method 2 is continuous. Assuming that, one REM sleep period (for example, period R4 or R5) is determined. However, in the case of this example, a portion where the period in which the 30-second section rb exists within 2 minutes does not overlap with the sections ra1 to ra5 determined to be REM sleep by the method 1 is determined not to be REM sleep. do.

In this way, the method 1 for determining REM sleep based on the heartbeat and the method 2 for determining REM sleep based on the machine learning of the frequency distribution of the pressure data detected by the mattress sensor 2 are combined to obtain REM sleep. By determining the period, it becomes possible to determine REM sleep with extremely high accuracy.
Specifically, in the method 1 for determining REM sleep based on the heartbeat, as can be seen from the sections ra1 to ra5 in FIG. 9, a relatively short period is determined as REM sleep, and the actual REM sleep It is likely that it does not match the length.
On the other hand, in the method 2 for determining REM sleep from the machine learning of the frequency distribution of the detected data, for example, even in the period between the section R4 and the section R5 in FIG. There was a section rb to be determined, and there was a possibility that the accuracy of determining REM sleep was not so high.

Here, in the case of this example, by combining the method 1 and the method 2, there is at least a section determined to be REM sleep by the method 1, and the section rb determined to be REM sleep by the method 2 is a certain period ( By determining the REM period R1 to R5 when it exists within (2 minutes), it becomes possible to determine the REM sleep period with extremely high accuracy.
Moreover, in the case of this example, both the heartbeat data for determining the method 1 and the frequency distribution data of the body movement for determining the method 2 are obtained by the mattress sensor 2 of the person to be measured. It is pressure data based on body movement, and unlike the case of measuring brain waves, it becomes possible to accurately determine REM sleep without imposing a burden on the person to be measured.

In the above-described embodiment, both the heartbeat data for determining the method 1 and the frequency distribution data of the body movement for determining the method 2 are output from the same sensor (mattress sensor 2). Although it was obtained from the data, for example, the heartbeat data was acquired from another sensor so that only the data of the frequency distribution of the body movement for determining the method 2 was obtained from the mattress sensor 2. May be good.

Further, the heartbeat determination period (5 minutes in step S111) for determining the method 1 and the frequency distribution aggregation period for determining the method 2 (step S121) described in the above-described embodiment example. The length of (30 seconds) in) shows a suitable example and is not limited to these values. Further, as for the threshold value described in the above-described embodiment, the above-mentioned value is an example, and other values may be set.

1 ... bed, 2 ... mattress sensor, 10 ... sleep stage determination device, 11 ... biometric data acquisition unit, 12 ... biometric data processing unit, 13 ... sleep stage determination unit, 14 ... output unit, A ... subject, C ... Computer device, C1 ... CPU, C2 ... ROM, C3 ... RAM, C4 ... Non-volatile storage, C5 ... Network interface display, C6 ... Input device, C7 ... Display device, C8 ... Bus

Claims (5)

When the rate of increase in the median heart rate for each first period of the subject is equal to or greater than a predetermined threshold, it is determined that the first REM sleep period has started, and the median heart rate for each first period is determined. When the rate of decrease is reduced, the end of the first REM sleep period is determined and
The body movement of the person to be measured is frequency-analyzed every second period, and a plurality of decision trees for obtaining a determination of REM sleep are generated by machine learning by analyzing the frequency-analyzed data, and the generated multiple decision trees are used. From the obtained majority decision on whether or not it is REM sleep, the data processing unit that determines that the second period is the second REM sleep period, and
When the period determined to be the first REM sleep period and the period determined to be the second REM sleep period overlap, and there is a period determined to be the second REM sleep period within a certain interval. , A sleep stage determination device including a sleep stage determination unit that determines a period determined as a second REM sleep period within a certain interval as a REM sleep period. The sleep stage determination device according to claim 1, wherein the sleep stage determination unit determines that the first REM sleep period that does not overlap with the second REM sleep period is also a true REM sleep period. It is provided with a biological data acquisition unit that acquires output data of a pressure sensor that detects the body movement of the person to be measured.
The heartbeat is obtained from the output data of the pressure sensor acquired by the biometric data acquisition unit, and the output of the pressure sensor acquired by the biometric data acquisition unit is also obtained for the body movement for determining the second REM sleep period. The sleep stage determination device according to claim 1 or 2, which is obtained from the data. When the rate of increase in the median heart rate for each first period of the subject is equal to or greater than a predetermined threshold value, it is determined that the first REM sleep period has started, and the median heart rate for each first period is determined. When the decrease rate of the first REM sleep period is determined, the first REM sleep period determination process for determining the end of the first REM sleep period and
In the analysis of the data obtained by frequency-analyzing the body movement of the person to be measured for each second period, a plurality of decision trees for obtaining the determination of REM sleep were generated by machine learning, and the REM sleep obtained from the generated plurality of decision trees was generated. From the majority decision of whether or not, the second REM sleep period determination process for determining that the second period is the second REM sleep period, and
When the period determined to be the first REM sleep period and the period determined to be the second REM sleep period overlap, and there is a period determined to be the second REM sleep period within a certain interval. , A sleep stage determination method including a sleep stage determination process for determining a period determined to be a second REM sleep period within a certain interval as a REM sleep period. When the rate of increase in the median heart rate for each first period of the subject is equal to or greater than a predetermined threshold value, it is determined that the first REM sleep period has started, and the median heart rate for each first period is determined. The first REM sleep period determination procedure for determining the end of the first REM sleep period and the procedure for determining the end of the first REM sleep period when the decrease rate of
The body movement of the person to be measured is frequency-analyzed every second period, and a plurality of decision trees for obtaining a judgment of REM sleep are generated by machine learning by analyzing the frequency-analyzed data, and the generated multiple decision trees are used. From the obtained majority decision on whether or not it is REM sleep, the second REM sleep period determination procedure for determining that the second period is the second REM sleep period, and the second REM sleep period determination procedure.
When the period determined to be the first REM sleep period and the period determined to be the second REM sleep period overlap, and there is a period determined to be the second REM sleep period within a certain interval. , The sleep stage determination procedure for determining the period determined to be the second REM sleep period within the fixed interval as the true REM sleep period,
A program that causes a computer to run.

Images