JP2012000375A - Sleep state determining device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a sleep state determining device capable of determining the sleep state of a person with high accuracy in spite of no restraints.SOLUTION: The sleep state determining device includes body motion amount computing means for computing the time-series waveform data of the amount of body motion based on data obtained by detecting the body motion of a subject 5 during sleep by a video camera 2; body motion extracting means for extracting large body motion continuously made for a predetermined time or longer from the computed time-series waveform data of the body motion, and eliminating body motion in a body motion restrained section where the large body motion is restrained; and determining means for computing the stillness duration (t) of the large body motion every body motion restrained section to extract the time intervals of the large body motion, and determining which of multiple stages of sleep depth the sleep state of the subject is classified in, based on the time intervals of the large body motion and the body motion amount of the large body motion.

Description

  The present invention relates to a sleep state determination device that determines whether a sleep state is classified into one of a plurality of stages of sleep depth based on body movements during human sleep.

  Sleep is essential for humans and is important for maintaining brain and body function. However, in recent years, the number of people who suffer from sleep is rapidly increasing with the progress of highly stressed society. Poor sleep and sleep disturbances are known to cause mood and mood when waking up, and to reduce motivation and depression in older people, causing social activity to decline. It is one of the social problems. In developing children, as the quality and quantity of sleep decline, the learning function declines due to memory and concentration problems, the emotional control function declines, the motor ability declines, and irregular lifestyles become established. It is necessary to find and treat sleep disorders before falling into chronic insomnia. To that end, it is necessary to improve sleep in the long term by grasping one's daily sleep state well and managing daily sleep.

  Sleep is generally divided into shallow REM sleep and deep non-REM sleep, but more specifically, it is defined by multiple sleep depths. ) & Cares (A.Kales)), the sleep depth is defined by six states of awakening from entering the bed to falling asleep, REM sleep, and depths 1 to 4. Here, the depths 1 to 4 are obtained by further dividing the non-REM sleep into four stages. Among the non-REM sleeps, the depth 1 is the shallowest and the depth 4 is the deepest.

  When the person enters the floor, the sleep state shifts from the awake state to the light sleep state, and then shifts from the light sleep state to the deep sleep state. And it is common to shift to the state of shallow sleep again. By examining the transition of the sleep depth during sleep, it is possible to grasp the sleep state of the person.

  Therefore, various attempts have been made to detect a person's sleep state. For example, brain waves, eye movements, jaw electromyograms, electrocardiograms, etc. are measured simultaneously, and the measurement data is analyzed to determine the sleep depth. Thus, a sleep polygraph (PSG) method for evaluating a sleep state is known (see, for example, Patent Documents 1 and 2). Since this polysomnography method has many types of measurement data and high measurement accuracy, it has high reliability of evaluation.

Japanese Patent No. 2950038 JP 2004-305258 A

  However, since the measurement by the polysomnography method described above is to grasp the state of sleep by attaching a plurality of sensors to the measurement subject, the apparatus becomes large, and in measuring, hospitals, laboratories, etc. Since it can only be used in places with specialized equipment, it is not suitable for daily use at home as a health device. In addition, since multiple electrodes must be attached to the subject, the burden on the subject is great, the subject's sense of restraint often hinders the subject's sleep, and it is a great stress for children who are highly resistant to new experiences. Therefore, there is also a problem that it is difficult to measure an accurate sleep state.

  Furthermore, the polysomnographic method cannot accurately evaluate the sleep state of the subject unless it is by a doctor or other specialist, so that it is not necessary to have only an apparatus.

  The present invention has been made paying attention to the above-described problem, and an object of the present invention is to provide a sleep state determination device that can determine a person's sleep state with high accuracy while being unconstrained.

  The object of the present invention is a sleep state determination apparatus for determining a sleep state based on a body movement at the time of sleep of a subject, and based on data obtained by detecting the body movement at the time of sleep of the subject. A body movement amount calculating means for calculating time series waveform data of body movement amount as information relating to the size of the body, and extracting a large body movement continuously performed for a predetermined time or more from the calculated time series waveform data of the body movement amount , Body motion extracting means for removing the body motion in the body motion suppression section in which the large body motion is suppressed, and the large body motion by calculating the stationary duration t of the large body motion for each body motion suppression section. And determining means for determining whether the sleep state of the subject is classified into one of a plurality of stages of sleep depth based on the time interval of the large body motion and the amount of body motion of the large body motion Sleep state with It is accomplished by a constant system.

  In a preferred embodiment of the present invention, the determination means sets 0 as the body movement stationary duration at any unit time of sleep time when it is outside the body movement suppression section, and within the body movement suppression section. In some cases, the rest duration t is set, and based on the body motion rest duration and the body motion amount for each unit time, it is determined whether the sleep state is classified into one of a plurality of stages of sleep depth. It is characterized by that.

  In a further preferred embodiment of the present invention, the determination means obtains an average of the body movement amount at each time up to a predetermined time before the unit time as the body movement amount for each unit time, and the average body movement amount and It is characterized in that it is determined whether the sleep state is classified into any one of a plurality of stages of sleep depth based on the body motion stationary duration.

  In a further preferred embodiment of the present invention, the determination means uses discriminant analysis as a method for determining to which sleep depth the sleep state for each unit time is classified.

  In a further preferred embodiment of the present invention, the body motion amount calculating means includes a converting means for converting a moving image obtained by an imaging means for imaging a human body motion during sleep into a predetermined number of still images, Detection means for detecting a change in density value in the image accompanying the body movement of the subject by performing differential processing on each still image constituting the moving image, and a center position coordinate of the body movement of the subject by the change in the density value And a body movement amount calculating means for calculating the body movement amount according to a change in the position of the center position coordinate of the body movement of the subject.

  ADVANTAGE OF THE INVENTION According to this invention, the sleep state determination apparatus which can determine a person's sleep state with high precision, without restraint can be provided.

It is explanatory drawing which shows schematic structure of the sleep state determination apparatus which concerns on one Embodiment of this invention. It is a block diagram which shows schematic structure of an information processing part. It is a block diagram which shows schematic structure of a sleep state determination part. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject A, and the actual measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject B, and the actual measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject C, and the actual measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject D, and the actual measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject E, and the actual measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject F, and the measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject G, and the actual measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the body movement amount measurement data by the information processing part of the test subject H, and the actual measurement data of the sleep depth by the sleep polygraph method. It is explanatory drawing for demonstrating a body movement stationary duration. It is the graph which compared the determination result of the sleep depth by the sleep state determination apparatus which concerns on this embodiment, and the measurement data of the sleep depth by the sleep polygraph method. It is the graph which compared the determination result of the sleep depth by the sleep state determination apparatus which concerns on this embodiment, and the measurement data of the sleep depth by the sleep polygraph method. It is explanatory drawing for demonstrating the frequency | count of body movement. It is the graph which compared the determination result of the sleep depth by the sleep state determination apparatus which concerns on other embodiment, and the measurement data of the sleep depth by the sleep polygraph method.

  Embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 1 shows a schematic configuration of a sleep state determination apparatus 1 according to an embodiment of the present invention. As shown in FIG. 1, the sleep state determination device 1 is connected to a video camera 2. The video camera 2 is fixed by a tripod, and is installed so that, for example, the whole body of the subject 5 lying on the bed 10 enters the frame. The video camera 2 captures a moving image of the subject 5 during sleep, and outputs the captured moving image to the sleep state determination device 1. The output moving image is stored in the sleep state determination apparatus 1. Note that the video camera 2 preferably has an infrared photographing function capable of photographing a video at night or in the dark.

  The sleep state determination apparatus 1 calculates body movement amount time series waveform data of body movement amount, which is information related to the magnitude of body movement (body movement) during sleep of the subject 5 based on a moving image from the video camera 2. A sleep state determination unit 4 (determining a sleep state of the subject 5 during sleep based on the information processing unit 3 (shown in FIG. 2) including the means 35 and the time-series waveform data (body movement amount data) of the body movement amount. 3), and can be configured by a portable computer, for example.

  The information processing unit 3 detects the magnitude of the movement of the subject during sleep due to body movement by performing a difference process of moving images input from the video camera 2. The difference process is a process for detecting a motion of an object by calculating a difference between successive images. In this embodiment, it is possible to determine the sleep state in a completely unconstrained state in which the sensors are separated from the subject by detecting the body motion amount data during the sleep of the subject by the difference processing of the moving image. ing. Hereinafter, a method of calculating the amount of body movement for each time series will be specifically described.

  As shown in FIG. 2, the information processing unit 3 includes a still image conversion unit 30, an ROI setting unit 31, a gray scale conversion unit 32, a density value change detection unit 33, a center position coordinate calculation unit 34, Body movement amount calculating means 35.

  The still image conversion means 30 converts the moving image input from the video camera 2 into a still image (for example, a BMP format still image) for each predetermined number of frames constituting the moving image. In the present embodiment, the frame rate (for example, 30 [fps]) of the moving image input from the video camera 2 is resampled to 1 [fps]. This is because during sleep, the subject's movement due to body movement is small, and 1 [fps] is considered to be a sufficient sampling rate to capture the subject's movement. Thereby, the amount of information stored in the sleep state determination device 1 can be reduced.

  The ROI setting unit 31 continuously reads BMP format still images output from the still image conversion unit 30 and sets an ROI (Region of Interest) for speeding up the processing. ROI refers to a region of interest, and a region of interest, that is, a region necessary for processing is set and cut out from the entire still image. By performing the ROI process, it is possible to remove excess noise and speed up the process by reducing the image size of the still image.

  The gray scale conversion unit 32 converts the gray scale of each still image that has been subjected to the ROI processing, and can detect a change in the shading of the image between frames by differential processing.

In the moving image input from the video camera 2, the density value of each pixel in the image changes with the movement of the subject due to the body movement. The density value change detecting means 33 detects a change in density value of each pixel in the image by taking a difference between two adjacent frames. Specifically, for each still image of each frame that has been grayscaled, a time derivative of the density value of all the pixels is calculated, and a change in the density value of each pixel is detected. The image size of the still image is n (1, 2,..., N) in the x direction and m (1, 2,..., M) in the y direction. 240 (height)) pixels, and if the density determinant of each frame at time t is D (t), the density determinant D (t) and the time of the density determinant D (t) The differential matrix D ′ (t) is calculated by the following mathematical formulas (1) and (2), respectively. A change in density of a pixel at a certain coordinate position (x, y) is expressed by d ′ _xy (t).

The thus obtained change in the density value of each pixel is input to the center position coordinate calculation means 34. The center position coordinate calculation means 34 estimates the center position coordinates of the movement of the body during sleep of the subject from the change in the concentration value. Specifically, if the subject moves due to body movement in the difference image, the density value decreases at the coordinates before the movement and the density value increases at the coordinates after the movement. d ′ _xy (t) is classified into the following three states using a constant ε. In addition, the constant ε in the following conditional expression is an allowable error.

(1) Concentration value increases | d ′ _xy (t) |> ε and d ′ _xy (t) <0
(2) Density value decreases | d ′ _xy (t) |> ε and d ′ _xy (t)> 0
(3) No change in density value | d ′ _xy (t) | ≦ ε

Next, (1) the average value (p _x , p _y ) of the coordinate positions of all the pixels whose density values have increased, and (2) the average value (q _x , q _y ). Then, the coordinates (p _x , p _y ) are the center coordinates of the aggregate of the pixels having the increased density value, the coordinates (q _x , q _y ) are the center coordinates of the aggregate of the pixels having the decreased density value, and the coordinates (p By calculating the center coordinates of _x , p _y ) and coordinates (q _x , q _y ), the center position coordinates of body movement during sleep of the subject are calculated. The center position coordinates (x, y) of the body movement of the subject in each of the x direction and the y direction are calculated by the following mathematical formulas (3) and (4), respectively.

  The body movement center position coordinates (x, y) thus obtained are input to the body movement amount calculation means 35. The body motion amount calculation means 35 calculates a value L corresponding to the distance that the subject moved by body motion during sleep in the image at time t from the time-series change in the position of the center position coordinate (x, y) of the subject's body motion. (T) is calculated from the following equation (5).

  Next, since the time differentiation of the movement distance represents the speed, a value v (t) corresponding to the speed at which the subject moved due to body movement during sleep is calculated by the following formula (6).

  Using the speed v (t) thus obtained, the body movement amount calculating means 35 calculates the body movement amount p (t) at the time t during sleep. The body motion amount p (t) at time t is calculated by the product of the velocity v (t) and the weight m of the subject, that is, p (t) = m × v (t). Time series waveform data of body movement is measured.

  4 to 11 show time series waveform data (lower part) of body movement amount during sleep measured by the information processing unit 3 of 8 healthy preschool children (including men and women) A to H as subjects. 4 to 11, the horizontal axis represents the measurement time [sec] and the vertical axis represents the amount of body movement [pic]. In addition, in FIG. 4 to FIG. 11, actual measurement data (upper stage) of the sleep depth measured by the polysomnography method for these subjects A to H is also shown for comparison of the sleep depth.

  Next, the sleep state determination unit 4 classifies the sleep state of the subject during sleep based on the time series waveform data of the body movement measured by the information processing unit 3 into one of a plurality of stages of sleep depth. As shown in FIG. 3, GMs extraction means 40, body movement stationary duration extraction means 41, average body movement amount extraction means 42, sleep depth determination means 43, and sleep depth transition locus Generating means 44.

  There are two types of body movements when sleeping: large body movements (coarse body movements) such as turning over and fine body movements (fine body movements). Here, when body movements with a duration of 2 seconds or more are classified as Gross Movements (GMs), such GMs occur most frequently in the awake state (Wake), while the sleep becomes deeper, that is, REM. The frequency of appearance decreases as sleep and non-REM sleep (depth 1 to depth 4) continue. The present inventors have found that it is possible to discriminate the sleep state during sleep into a plurality of stages (sleep depth) by paying attention to the GMs.

  That is, as described above, when SWS (Slow Wave Sleep) in which the sleep state is deep sleep at depth 3 and depth 4, GMs are in a suppressed state, and the time interval at which GMs occurs takes a large value. . Therefore, it is possible to extract the SWS during sleep by extracting a section in which GMs are suppressed and detecting the duration in which the body motion is stationary (body motion stationary duration). The inventors have found. Furthermore, when the sleep state is in the awake state (Wake), the GMs are in the non-suppressed state and move greatly due to body movement, so it is considered that the body movement amount takes a large value. Therefore, the present inventors have found that Wake during sleep can be extracted by taking into account the magnitude of body movement (body movement amount) during GMs.

  Thus, in the present embodiment, in view of the fact that the time interval of GMs and the magnitude of body movement (body movement amount) during GMs are greatly related to the sleep state, the time series waveform of the body movement amount during sleep of the subject For each time data t, for each time data t, by extracting “body motion stationary duration” and “body motion amount” which are two variables related to body motion having a strong relationship with sleep depth, about the sleep state during sleep, It is possible to discriminate by at least three levels of sleep depths: awake state (Wake), Light & REM (REM sleep, depth 1, 2), and SWS (depth 3, 4). Hereinafter, a method for determining the sleep state will be specifically described.

  First, the GMs extracting means 40 extracts only GMs in which body movement continues for 2 seconds or more from the time series waveform data of the body movement amount during sleep of each subject measured by the information processing unit 3. The body movement with a duration of 2 seconds or less is removed, and corrected body movement amount time-series waveform data (corrected body movement amount data) is created. The corrected body movement amount data thus obtained is input to the body movement stationary duration calculation means 41 and the average body movement amount calculation means 42, respectively.

The body motion stationary duration extracting means 41 extracts “body motion stationary duration” at each time data t from the inputted corrected body motion amount data. First, a “body motion section” is defined for the corrected body motion amount data. As shown in FIG. 12, the body motion section refers to a section from when the body motion starts to be suppressed in a state where the body motion is suppressed until the body motion is suppressed again. Within each section, the time at which the body movement is suppressed t _0, the body movement occurs body movement from the state is suppressed start time and t _MTin, the time at which the body movement is suppressed again t _MTout, and by setting each The body movement stationary duration MT _rest at time t in each section is expressed by the following mathematical formula (7).

  Next, the average body movement amount calculating means 42 extracts the “body movement amount” at each time data t from the input corrected body movement amount data. Here, in the present embodiment, the average body motion amount calculation unit 41 extracts “average body motion amount” as the body motion amount in each time data t. The “average body movement amount” is a simple average value of body movement amounts at each time from a target time to a time before a predetermined time (60 [sec] in the present embodiment). In this way, the simple average value of the body movement amount for 60 seconds immediately before each time data t is extracted as the body movement amount in each time data t because the sleep depth in each time data t is performed immediately before that. It is based on the knowledge that it depends on the body movement. In addition, the reference range of the simple average value is set to 60 [sec] as a comparison target. The sleep epoch measurement epoch in the polysomnography method in which the sleep depth of the subject is determined is 30 [sec], and at least one epoch is used. This is because the reference range is set to include the whole.

  The “body motion stationary duration” and “average body motion amount” for each time data t obtained in this way are input to the sleep depth determination means 43. The sleep depth discrimination means 43 discriminates the sleep depth for each time series for the sleep state of the subject based on the “body motion stationary duration” and the “average body motion amount” in each input time data t. In this embodiment, linear discriminant analysis is used as an algorithm for discriminating the sleep depth.

  Linear discriminant analysis is an algorithm for discriminating a data group into two or more groups (discriminant group) to be discriminated. In this embodiment, the above-mentioned “body motion stationary duration” and When a scatter diagram of each time data t is drawn using the “average body movement amount” and the “body movement stationary duration” and “average body movement amount”, each time data t is expressed as awake state (Wake), shallow. The optimal functions (linear discriminant functions) for dividing into three discriminant groups of sleep state (Light & REM) and deep sleep state (SWS) are obtained. And when unknown data is input, the sleep depth for every time series during sleep is discriminated by determining which group group the unknown data belongs to based on a linear discriminant function. Hereinafter, a method for obtaining the discriminant analysis function will be described.

  The sleep depth discriminating means 43 first includes, as teacher data for determining a discriminant analysis function among the measured body movement data (FIGS. 4 to 11) of the subjects A to H, for example, subjects B, C, D, Select E, G, H body movement data. With respect to the body motion amount data selected as the teacher data, the “body motion rest duration” and “average body motion amount” for each time data t extracted by the body motion rest duration extracting means 41 and the average body motion calculating means 42. ”As an independent variable, and for comparison of sleep depth, based on the measured data of sleep depth for each subject B, C, D, E, G, H measured by the polysomnographic method for each time series A linear discriminant represented by the following formulas (8) and (9) is calculated with “sleep depth (wake state (Wake), Light & REM, SWS)” for each extracted time data t as an objective variable. The calculated values Z1 and Z2 of this linear discriminant indicate discrimination scores, respectively, and grouping is performed according to whether Z1 and Z2 are positive or negative.

Z1 = a ₁ x ₁ + a ₂ x ₂ + C1 (8)
_{Z2 = a 3 x 1 + a} 4 x 2 + C2 ··· (9)

First, assuming that the equation (8) is a linear discriminant function for discriminating the discriminant group Wake (wake state) and the other discriminant groups (Light & REM and SWS), the linear discriminant function is located at a position farthest from the second group. This linear discriminant function is the best reference line that separates the two groups by being drawn and being drawn farthest from the second group. In order to best divide the two groups, the sum of squares S _T (referred to as “total variation”) of the difference between the discrimination score Z1 _{i of} all the discrimination groups and the average Z1 _all, and the discrimination score Z1 _j of the discrimination group Wake The correlation ratio F1 (= S _wake ÷ S _T ) with the square sum S _wake (referred to as “variation between groups”) of the difference from the average Z1 _wake is maximized.

Therefore, first, the total variation _{S T} and group variability _{S wake,} the following equation (10), is calculated by using respectively (11). In the following formulas (10) and (11), n _all indicates the number of discrimination scores of all discrimination groups, and n _wake indicates the number of discrimination scores of the discrimination group Wake.

Then, after calculating the correlation ratio F1 _(a 1, a _2), to separate and determine group Wake and other discrimination group optimally, correlation ratio F1 _(a 1, a ₂₎ so that is maximized A ₁ and a ₂ , that is, as shown in the following formulas (12) and (13), F1 (a ₁ and a ₂ ) are partially differentiated by a ₁ and a ₂ , respectively, so that the values become 0 Obtain a ₁ and a ₂ .

∂F1 (a ₁ , a ₂ ) / ∂a ₁ = 0 (12)
∂F1 (a ₁ , a ₂ ) / ∂a ₂ = 0 (13)

Next, to calculate the constant term C1, it calculates the respective average values X ₁ and X ₂ of the body movement quiescent time duration and the average body motion amount for all discrimination group. Since the linear discriminant function needs to draw the center of the discriminant group Wake and the center of the other discriminant group into two, the center of the discriminant group Wake and the center of the other discriminant group are divided into two. It is necessary to pass through the center (X ₁ , X ₂ ) of the entire discriminant group. When the linear discriminant function Z1 = a ₁ x ₁ + a ₂ x ₂ + C1) passes through the center (X ₁ , X ₂ ) of the entire discriminant group, 0 = a ₁ X ₁ + a ₂ X ₂ + C1 holds. Thus, as a result of calculating C1, a linear discriminant function Z1 for discriminating the awake state (Wake) from the other discriminant groups (Light & REM and SWS) is obtained.

Next, assuming that the above equation (9) is a linear discriminant function for discriminating between the discriminant group SWS and the other discriminant groups (Light & REM and Wake), the linear discriminant function Z2 is the most from the two groups. This linear discriminant function Z2 becomes the best reference line for dividing the two groups by being drawn to a far position and being drawn to the farthest position from the second group. Therefore, in order to calculate the linear discriminant function Z2 that best separates the two groups, and the total variation _{S T,} between the sum of squares _{S SWS} ( "Group of the difference between the discriminant score Z2 _k discrimination group SWS and its average _{Z2 SWS} A ₃ and a ₄ are calculated so that the correlation ratio F2 (= S _SWS ÷ S _T ) with _{respect to} “variation”) is maximized.

  Similarly, by calculating the constant term C2 of the linear discriminant function Z2, a linear discriminant function Z2 for discriminating the discriminant group SWS from the other discriminant groups (Light & REM and Wake) is obtained.

Based on the linear discriminant functions Z1 and Z2 obtained in this way, the sleep depth discriminating means 43 uses the “body motion stationary duration” of each time data t and the body motion amount data during sleep of the subject who wants to discriminate the sleep depth and The “average body movement amount” is input to x ₁ and x ₂ in the above formulas (8) and (9) to calculate discrimination scores Z1 and Z2, respectively. First, the discrimination score Z1 is set to the discrimination threshold “0”. In comparison, it is determined whether each time data t belongs to the discrimination group Wake or other discrimination groups (Light & REM and SWS). Then, for each time data t discriminated in the discrimination group of Light & REM and SWS, the discrimination score Z2 is compared with the discrimination threshold “0” to determine whether each time data t belongs to the discrimination group SWS. It is determined whether it belongs to the group Light & REM. This makes it possible to determine whether the sleep state of the subject during sleep is “wake state (Wake)”, “light sleep state (Light & REM)”, or “deep sleep state (SWS)”. Is possible.

  The sleep depth discrimination data thus obtained is input to the sleep depth transition locus generating means 44. The sleep depth transition trajectory generating means 44 generates a time transition trajectory of the sleep depth and outputs it to a monitor or the like.

  FIG. 13 and FIG. 14 show the sleep state determination result (lower stage) determined by the sleep state determination apparatus 1 according to the present embodiment and the sleep depth measurement data (upper stage) measured by the polysomnography method for the test subject A and the test subject F. ). Referring to FIGS. 13 and 14, the sleep depth is determined except for the fact that Wake misidentifies in the vicinity of 26000 [sec] in FIG. 13 and 12000 [sec] in FIG. 14 and in a short section in each part. The value and the actual measurement value show extremely approximate fluctuations, and thus it can be said that the sleep depth determination device 1 of the present invention can determine the transition of the approximate sleep depth, and the sleep depth is accurately reproduced. It is confirmed that

  As described above, according to the sleep depth determination device 1 according to the present invention, the sleep state is determined from the body movement data of the test subject detected by the video camera 2, so that the device becomes a large scale. Therefore, it is suitable as a sleep depth determination device in a general home. In addition, since the sleep state can be determined in a completely unconstrained state in which the subject is not restrained by sensors, the subject can be measured with no burden and an accurate sleep state can be measured.

  In the present embodiment, linear discriminant analysis is used as an algorithm for discriminating the sleep depth of each time series by the sleep depth discriminating means 43 with respect to the sleep state of the subject. Discriminant analysis by a method (nonlinear discriminant analysis such as Mahalanobis pan-distance, multiple discriminant analysis, canonical discriminant analysis, etc.) may be used. Further, in the present embodiment, body motion data is measured in an unconstrained state for the subject by detecting the body motion data during sleep of the subject from a moving image obtained by photographing the subject during sleep with the video camera 2. However, the method for measuring the body movement data of the subject without restriction is not necessarily limited thereto. For example, a pressure vibration sensor may be laid under the bed mattress, and the body motion data of the subject may be measured without restriction from the output value of the pressure vibration sensor. Various methods can be adopted as long as the data can be measured without restriction.

  As mentioned above, although one Embodiment of this invention was described, the specific aspect of this invention is not limited to the said embodiment. For example, in this embodiment, the sleep state of the subject during sleep is determined in three stages (sleep depth): an awake state (Wake), a shallow sleep state (Light & REM), and a deep sleep state (SWS). In addition to this, it is configured to further distinguish the light sleep state (Light & REM) into a REM sleep state (REM) and a light sleep state (Light, depth 1 and depth 2 non-REM sleep). You can also.

  In human sleep, in the REM sleep state (REM), muscle tone of skeletal muscles is suppressed, and thus large body movements are reduced. Here, this large body movement refers to Localized Movements (LMs) and GMs whose duration exceeds 0.5 seconds. On the other hand, Twitch Movements (TMs) having a duration of 0.5 seconds or less are known to appear swarms particularly in the REM sleep state (REM). In addition, GMs in the REM sleep state (REM) are characterized by a long movement of the body movement and a large movement with a sleeping collar compared to the light sleep state (Light). It is done.

  Thus, when the sleep state is in the REM sleep state (REM), the GMs are in a suppressed state compared to the light sleep state (Light), and the time interval at which the GMs occurs takes a large value. Further, since TMs is in an unsuppressed state, the number of body movements takes a large value. Under such knowledge, the present inventors, for each time data t discriminated as Light & REM, the “body motion stationary duration” which is strongly related to discriminating Light and REM from body motion data and By extracting “the number of body movements” as an independent variable and performing discriminant analysis using the sleep depth obtained by the polysomnography method as a target variable, each time data t discriminated as Light & REM can be discriminated as Light and REM. I thought that. As the “number of body movements”, in this embodiment, as shown in FIG. 15, body movement data for the previous 60 seconds is extracted from a certain time t, and the number of body movement peaks is counted. This sum is used as the number of body movements at time t (three times in the illustrated example).

  Therefore, for each time data t discriminated as Light & REM, “body motion stationary duration” and “number of body motions” are extracted from the body motion data, and discriminant analysis based on Mahalanobis distance using these two parameters. A discriminant analysis function optimal for dividing each time data t into two discriminant groups (sleep depth) of a light sleep state (Light) and a REM sleep state (REM) was constructed.

  The graph of FIG. 16 shows a sleep depth determination value (lower row) determined by the above-described discriminant analysis based on the Mahalanobis distance after the sleep state is determined to three sleep depths by the sleep state determination device 1 of the present invention. ) And a measured value of sleep depth by the polysomnographic method (upper stage). As shown in the figure, it was confirmed that the determination value of the sleep depth according to the present embodiment determined from the body movement shows a variation that is very close to the actual measurement value obtained by the sleep polygraph method.

DESCRIPTION OF SYMBOLS 1 Sleep state determination apparatus 2 Video camera 3 Information processing part 4 Sleep state determination part 30 Still image conversion means 31 ROI setting means 32 Gray scale conversion means 33 Density value change detection means 34 Center position coordinate calculation means 35 Body movement amount calculation means 40 GMs Extraction means 41 Body movement stationary duration extraction means 42 Average body movement amount extraction means 43 Sleep depth determination means 44 Sleep depth transition locus generation means

Claims (5)

A sleep state determination device that determines a sleep state based on a body movement during sleep of a subject,
Body motion amount calculating means for calculating time series waveform data of body motion amount, which is information related to the size of the body motion, based on data obtained by detecting body motion during sleep of the subject;
Body motion extraction for extracting a large body motion continuously performed for a predetermined time or more from the calculated time series waveform data of the body motion amount and removing a body motion in a body motion suppression section in which the large body motion is suppressed. Means,
For each body motion suppression section, the stationary duration t of the large body motion is calculated to extract the time interval of the large body motion, and based on the time interval of the large body motion and the body motion amount of the large body motion, A sleep state determination apparatus comprising: determination means for determining whether a subject's sleep state is classified into one of a plurality of stages of sleep depth. For each unit time of the sleep time, the determination means sets the body movement stationary duration as 0 when it is outside the body movement suppression section, and when it is within the body movement suppression section, its rest duration t. Set each one
The sleep state determination apparatus according to claim 1, wherein the sleep state determination device determines whether the sleep state is classified into one of a plurality of stages of sleep depth based on the body movement stationary duration and the amount of body movement for each unit time.   The determination means obtains an average of the body movement amount at each time up to a time before the predetermined time as the body movement amount for each unit time, and based on the average body movement amount and the body movement stationary duration, The sleep state determination apparatus according to claim 2, wherein the sleep state is determined as one of a plurality of stages of sleep depth.   The sleep state determination apparatus according to claim 2 or 3, wherein the determination unit uses discriminant analysis as a method of determining to which sleep depth the sleep state for each unit time is classified.   The body motion amount calculating means includes a converting means for converting a moving image obtained by an imaging means for capturing a human body motion during sleep into a predetermined number of still images, and each still image constituting the moving image. By performing difference processing, a detection unit that detects a change in density value in the image accompanying the body movement of the subject, a center position calculation unit that calculates a center position coordinate of the body movement of the subject by the change in the density value, The sleep state determination apparatus according to any one of claims 1 to 4, further comprising body movement amount calculation means for calculating the body movement amount based on a change in a central position coordinate of body movement of the subject.

Images