Sleep estimation device, sleep estimation system, wearable device, and sleep estimation method

Abstract

A sleep stage is simply estimated. A sleep estimation device includes a first acquisition unit that acquires blood flow data, a generation unit that generates a frequency spectrum of the blood flow data by performing frequency analysis processing on the blood flow data, and a first determination unit that determines a sleep stage of a subject based on the frequency spectrum.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to estimation of a sleep stage of a subject. BACKGROUND

[0002] A technique of detecting a sleep stage is described in Patent Literature 1.

[0003] PRIOR ART DOCUMENTS

[0004] PATENT LITERATURE

[0005] Patent Literature 1: Japanese Patent Application Publication No. 2018-161432 SUMMARY

[0006] A sleep estimation apparatus according to an embodiment of the present disclosure includes: a first acquisition unit that acquires blood flow data representing blood flow of a subject; a generation unit that generates a frequency spectrum of the blood flow data by performing frequency analysis processing on the blood flow data; and a first determination unit that determines a sleep stage of the subject based on the frequency spectrum.

[0007] Further, a sleep estimation apparatus according to an embodiment of the present disclosure includes: an acquisition unit that acquires blood flow data representing blood flow of a subject; a generation unit that generates processing data representing a result of time-frequency analysis processing of the blood flow data by performing wavelet transform processing in which intensity in a given frequency band is more relatively emphasized than in other frequency bands or short-time Fourier transform processing on the blood flow data; and a determination unit that determines a sleep stage of the subject based on the processing data.

[0008] Further, a sleep estimation method according to an embodiment of the present disclosure includes: a first acquisition process of acquiring blood flow data representing blood flow of a subject; a generation process of generating a frequency spectrum of the blood flow data by performing frequency analysis processing on the blood flow data; and a first determination process of determining a sleep stage of the subject based on the frequency spectrum.

[0009] Further, a sleep estimation method according to an embodiment of the present disclosure includes: an acquisition process of acquiring blood flow data representing blood flow of a subject; a generation process of generating processing data representing a result of time-frequency analysis processing of the blood flow data by performing wavelet transform processing in which intensity in a given frequency band is more relatively emphasized than in other frequency bands or short-time Fourier transform processing on the blood flow data; and a determination process of determining a sleep stage of the subject based on the processing data. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 is a block diagram representing an example of a schematic configuration of a sleep estimation system according to Embodiment 1.

[0011] Figure 2is a graph showing an example of blood flow waveform data detected by a blood flow meter.

[0012] Figure 3 is a graph showing an example of a frequency spectrum generated by performing Fourier transform processing on blood flow waveform data.

[0013] Figure 4 is a graph showing an example of electrocardiogram waveform data detected by an electrocardiograph.

[0014] Figure 5 is a graph showing an example of a frequency spectrum generated by performing Fourier transform processing on electrocardiogram waveform data.

[0015] Figure 6 is an image showing an example of a result of wavelet transform processing on blood flow waveform data.

[0016] Figure 7 is a flowchart showing an example of a flow of processing performed by the sleep estimation device of Embodiment 1.

[0017] Figure 8 is a block diagram showing an example of a schematic configuration of the sleep estimation system of Embodiment 2. DETAILED DESCRIPTION

[0018] The determination (estimation) of the sleep stage of a subject related to the present disclosure will be described below. First, the principle that serves as the basis for the determination of the sleep stage of a subject related to the present disclosure will be described. Note that, in the present specification, the case where "A to B" is described indicates "A or more and B or less". Furthermore, in the present specification, blood flow waveform data will be described as an example of blood flow data used in the determination of the sleep stage.

[0019] [Principle]

[0020] Figure 2 is a graph showing an example of blood flow waveform data detected by a blood flow meter. In Figure 2 , the vertical axis represents a value proportional to the blood flow per unit time [unit: dimensionless], and the horizontal axis represents the measurement time [unit: min]. The blood flow meter can acquire the raw waveform data W1 and the processed waveform data W2 shown in Figure 2 as the blood flow waveform data. The processed waveform data W2 is waveform data obtained by processing the raw waveform data W1 so as to easily acquire the peak value of the R wave. The processed waveform data W2 is generated, for example, by performing smoothing processing on the raw waveform data W1. In the processed waveform data W2, the time interval between adjacent peak values (blood flow at the positions shown by the inverted triangular shapes in the graph) represents the heart rhythm interval (RRI: R-R Interval). In Figure 4Also in the raw waveform data W12 of the heart rhythm shown, the time interval between adjacent peaks (blood flow at positions shown by inverted triangles in the figure) represents the heart rhythm interval (RRI).

[0021] The blood flow meter that detects blood flow waveform data is a sensor that can detect blood flow waveform data representing blood flow of an object by receiving scattered light generated by irradiating light to a blood vessel of the object. The blood flow meter includes: a light emitting section that irradiates light to a blood vessel of an object; and a light receiving section that receives the scattered light.

[0022] Generally, if laser light is irradiated to a fluid, the irradiated laser light is scattered to generate scattered light due to (i) a scatterer included in the fluid and moving with the fluid and (ii) a stationary object such as a tube through which the fluid flows. Generally, the scatterer brings about inhomogeneity of complex refractive index in the fluid.

[0023] The scattered light generated by the scatterer moving with the fluid brings about a wavelength shift due to a Doppler effect corresponding to the flow rate of the scatterer. On the other hand, the scattered light generated by the stationary object does not bring about a wavelength shift. These scattered lights cause interference of light, and thus light beats are observed.

[0024] The blood flow meter can be a sensor that utilizes this phenomenon. That is, the blood flow meter can be a laser Doppler blood flow meter that detects light beats brought about by the scattered light generated in blood as a fluid by irradiating laser light to a blood vessel of an object, as blood flow waveform data.

[0025] More specifically, the acquired light receiving signal can be analyzed by a processor included in the blood flow meter to calculate frequency analysis data representing signal intensity of each frequency of the light receiving signal. As an example, the processor can analyze the acquired light receiving signal using a method such as FFT (Fast Fourier Transformation).

[0026] The processor can further generate blood flow waveform data representing a variation pattern of blood flow of the object based on the frequency analysis data. As an example, the processor can calculate the first moment and X of the acquired frequency analysis data as the blood flow waveform data. More specifically, the processor can calculate the first moment and X of the acquired frequency analysis data using the following equation. The processor can calculate the first moment and X in a certain frequency band (for example, 1 to 20 kHz) using the following equation.

[0027] X =∑ fx x P(fx)

[0028] Here, "fx" is a frequency, and "P(fx)" is a value of signal intensity at the frequency fx.

[0029] The first moment and X calculated by the processor based on the frequency resolution data can be values proportional to the blood flow of the subject. The processor can generate pattern data indicating a pattern of variation in the blood flow per hour of the subject by calculating the first moment and X for each of a plurality of the frequency resolution data. Further, the processor can generate blood flow waveform data using data included in the frequency resolution data, among data, which is included in a part of the frequency band. The processor can output the generated blood flow waveform data.

[0030] The blood flow waveform data can include, in addition to the blood flow, data related to at least one of a cardiac output and a coefficient of variation of vascular motion (vasomotion). The cardiac output is the amount of blood sent out by one heartbeat of the heart. The vasomotion is a contraction and expansion motion of the blood vessel that is naturally generated and rhythmic. The coefficient of variation of the vasomotion is a value indicating variation in the blood flow generated based on the vasomotion as a deviation.

[0032]

[0031] Further, the blood flow waveform data can include a pulse wave.

[0032] Figure 3 is a graph indicating an example of a spectrum generated by performing Fourier transform processing on the blood flow waveform data shown in Figure 2 The vertical axis indicates the intensity of the spectrum [unit: dB], and the horizontal axis indicates the frequency [unit: Hz]. The Fourier transform processing is an example of frequency resolution processing, and is processing of generating a spectrum of waveform data that does not include changes in time.

[0033] The spectrum FW1 corresponding to each sleep stage and the spectrum FW2 are shown in Figure 3 The spectrum FW1 is a spectrum generated as a result of performing Fourier transform processing on the original waveform data W1. The spectrum FW2 is a spectrum generated as a result of performing Fourier transform processing on the heart rate interval (RRI) of the processed waveform data W2.

[0034] The above-described sleep stages are classified into three stages of wakefulness, rapid eye movement sleep, and non-rapid eye movement sleep. The non-rapid eye movement sleep can be further classified into period 1 (N1), period 2 (N2), and period 3 (N3) in order from the lightest sleep stage. The rapid eye movement sleep is sleep accompanied by rapid eye movement (REM). The non-rapid eye movement sleep is sleep not accompanied by rapid eye movement.

[0035] This classification is based on brainwave data detected by an EEG meter installed on the subject. Brainwaves are categorized into four types according to wavelength, from longest to shortest: beta waves, alpha waves, theta waves, and delta waves. Beta waves are, for example, brainwaves with frequencies around 38–14 Hz. Alpha waves are, for example, brainwaves with frequencies around 14–8 Hz. Theta waves are, for example, brainwaves with frequencies around 8–4 Hz. Delta waves are, for example, brainwaves with frequencies around 4–0.5 Hz.

[0036] A person falls asleep when theta and delta waves are dominant compared to beta and alpha waves. Here, "dominant" means that the proportion of a certain wave increases in the measured brain waves. It is known that the dominant brain waves change periodically within the range of theta and delta waves during sleep. Furthermore, when the proportion of theta waves in the brain waves is less than a given value, the person is in REM sleep; when the proportion of theta waves is greater than a given value, and when delta waves are dominant, the person is in non-REM sleep. Period 1 is, for example, a state where alpha waves are less than 50% and various low-amplitude frequencies are mixed together. Period 2 is, for example, a state where low-amplitude theta and delta waves are irregularly present but without high-amplitude slow waves. Period 3 is, for example, a state where slow waves below 2 Hz and 75 μV account for more than 20%. The state where slow waves below 2 Hz and 75 μV account for more than 50% can also be called Period 4.

[0037] exist Figure 3 In this context, wakefulness is represented as "WK", rapid eye movement (REM) sleep as "RM", non-rapid eye movement (NREM) sleep stage 1 as "N1", NREM sleep stage 2 as "N2", and NREM sleep stage 3 as "N3".

[0038] like Figure 3 As shown in spectra FW1 and FW2, significant intensity variations are observed in the 0.2–0.3 Hz band (a given band) of the spectrum obtained from subjects in sleep stages 2 and 3. In other words, within this spectrum, the intensity in a first range R1 within the 0.2–0.3 Hz band is greater than the intensity in a second range R2 outside the first range R1 by a given value. Hereinafter, the intensity in the first range R1 will be referred to as the first intensity, and the intensity in the second range R2 will be referred to as the second intensity.

[0039] The first intensity can be, for example, the maximum intensity in a frequency band of 0.2 to 0.3 Hz. The second intensity can be, for example, the maximum intensity in a second range R2 other than a first range Rl including the maximum intensity. The given value is set, for example, to a degree that can determine the presence of the characteristic waveform Sh in the frequency band of 0.2 to 0.3 Hz. The characteristic waveform Sh can be, for example, a waveform that is convex upward and is wide to some extent (for example, a waveform having a full width at half maximum of 0.03 Hz or more). In Figure 3 In FIG. 6, an example of the first range Rl and the second range R2 in the frequency spectrum FWl and FW2 corresponding to period 3 is shown.

[0040] On the other hand, in the frequency spectrum obtained from a subject in a state of wakefulness, rapid eye movement sleep, and a sleep stage of period 1, no significant intensity change is observed in the frequency band of 0.2 to 0.3 Hz, and furthermore, the characteristic waveform Sh described above is not observed.

[0041] As a result of diligent research, the inventors found that when a frequency spectrum in which a significant intensity change (characteristic waveform Sh) is observed in the frequency band of 0.2 to 0.3 Hz is obtained, the subject is highly likely to be in a sleep stage of period 2 or period 3. That is, the inventors found that when a frequency spectrum having the characteristic waveform Sh is obtained, the subject is highly likely to be in a sleep stage of period 2 or period 3. Furthermore, the inventors found that, in particular, in the frequency spectrum of the blood flow waveform data detected by the blood flow meter (for example, a laser Doppler blood flow meter), a significant intensity change is observed in the frequency band of 0.2 to 0.3 Hz. The inventors arrived at these insights to develop a sleep estimation device that can improve the accuracy of the determination of the sleep stage of a subject.

[0042] (Comparison with electrocardiogram waveform data)

[0043] Furthermore, the following differences are observed between the electrocardiogram waveform data (electrocardiogram) detected by the electrocardiograph and the blood flow waveform data detected by the blood flow meter.

[0044] Figure 4 is a graph showing an example of the electrocardiogram waveform data detected by the electrocardiograph. In Figure 4 In FIG. 5, the vertical axis represents the intensity of the heart rhythm [unit: dB], and the horizontal axis represents the measurement time [unit: min]. In Figure 4 In FIG. 5, as the electrocardiogram waveform data, the original waveform data Wl l and Wl2 of the heart rhythm are shown.

[0045] Figure 5is a chart showing an example of a spectrum generated by performing Fourier transform processing on electrocardiogram waveform data. The vertical axis represents the intensity of the spectrum [unit: dB], and the horizontal axis represents the frequency [unit: Hz].

[0046] In Figure 5 spectra FW11 and FW12 corresponding to each sleep stage are shown. The spectrum FW11 is a spectrum generated as a result of performing Fourier transform processing on the original waveform data W11. The spectrum FW12 is a spectrum generated as a result of performing Fourier transform processing on the heart rate interval (RRI). The sleep stage corresponding to the spectra FW11 and FW12 is determined based on the brain wave data acquired from the brain wave meter equipped to the subject.

[0047] As Figure 5 shown in the spectra FW11 and FW12 obtained by transforming the electrocardiogram waveform data detected from the subject in the sleep stages of the period 2 and the period 3, no significant intensity change (characteristic waveform Sh) is seen in the frequency band of 0.2 to 0.3 Hz. With respect to the spectrum FW12 corresponding to the sleep stages of the period 2 and the period 3, in the frequency band of 0.2 to 0.3 Hz, it becomes a shape convex upward. But with respect to the spectrum FW12 corresponding to the sleep stage of the period 1, it also becomes the same shape as the spectrum FW12 corresponding to the period 2 and the period 3. Therefore, the spectrum FW12 corresponding to the period 2 and the period 3 does not see a significant intensity change in the frequency band of 0.2 to 0.3 Hz.

[0048] As a result of diligent research, the inventors found that the significant intensity change seen in the frequency band of 0.2 to 0.3 Hz is a phenomenon unique to the spectrum of the blood flow waveform data. From this, the inventors found that by using the spectrum transformed not from the electrocardiogram waveform data but from the blood flow waveform data, it is possible to determine the sleep stage of the subject with high accuracy, in particular, with high probability that the subject is in the sleep stage of the period 2 or the period 3.

[0049] (Regarding the frequency band)

[0050] Depending on the blood flow meter used and the individual differences of the subject, etc., there is a possibility that some expansion will occur in the frequency band in which the above-mentioned significant intensity change is seen. If this is taken into account, in the frequency band of, for example, 0.15 to 0.4 Hz of the spectrum FW1 and FW2 corresponding to the period 2 or the period 3, there is a sufficient possibility that the significant intensity change not seen in the electrocardiogram waveform data can be seen. In the following description, it is described that the frequency band in which the above-mentioned significant intensity change is seen is 0.2 to 0.3 Hz.

[0051] [Wavelet transform processing]

[0052] In the above-described principle, the use of a spectrum obtained by performing Fourier transform processing as frequency resolution processing is described. The sleep stage of the subject can be determined based on a spectrum obtained by performing wavelet transform processing as frequency resolution processing. Wavelet transform processing is an example of time-frequency resolution processing. Time-frequency resolution processing is processing of generating a spectrum of waveform data that includes a change in time. Wavelet transform processing is processing of generating a spectrum of waveform data using a mother wavelet as an arbitrary reference waveform.

[0053] The mother wavelet used in wavelet transform processing is defined as follows. In the following equation, "t" denotes a time variable, "a" denotes a scale parameter (a parameter that causes the mother wavelet to expand or contract in the time axis direction), and "b" denotes a translation parameter (a parameter that causes the mother wavelet to move parallel to the time axis direction).

[0054] [Math. 1]

[0055]

[0056] Further, a function for performing wavelet transform processing is defined as follows. In the following equation, "f(t)" denotes waveform data, and "*" denotes a conjugate complex number. By substituting the mother wavelet whose values of "a" and "b" are adjusted into the following equation, a spectrum of the waveform data can be generated.

[0057] [Math. 2]

[0058]

[0059] By using wavelet transform processing, the intensity (hereinafter referred to as a target intensity) in the frequency band of 0.2 to 0.3 Hz can be relatively more emphasized than other frequency bands. As described in the above-described principle, when a spectrum of a waveform Sh that has characteristics in the frequency band of 0.2 to 0.3 Hz is obtained, the likelihood that the subject is in a sleep stage of period 2 or period 3 is high. Therefore, by using wavelet transform processing to emphasize the target intensity, the likelihood that the determination of whether the subject is in a sleep stage of period 2 or period 3 is performed more accurately and well can be improved.

[0060] In wavelet transform processing, as described above, a mother wavelet that emphasizes the intensity of the object can be used. The mother wavelet with increased object intensity can be set by adjusting the values ​​of "a" and "b" mentioned above. Alternatively, Morlet can also be used as the mother wavelet. In this case, the scale parameter "a" has a relationship of "ω = 2π / a" and represents a local angular frequency. Since the angular frequency "ω" is represented as "ω = 2πf", this "f (frequency)" portion can be set to 0.2–0.3 Hz to perform wavelet transform processing. The object intensity can be the overall intensity of the 0.2–0.3 Hz frequency band, or it can be the intensity within a portion of the frequency band (e.g., the first range R1).

[0061] Furthermore, the wavelet transform processing of the blood flow waveform data can generate intensity variation data representing the change in intensity over time in each frequency band within a given time period. The given time can be experimentally determined to allow for accurate assessment of the subject's sleep stage. In this embodiment, the given time can be set, for example, to 2.5 minutes.

[0062] Unlike Fourier transform, wavelet transform can generate a spectrum that includes the temporal variation of intensity, thus allowing for a greater amount of data compared to Fourier transform. Generally, in the generation of fully learned models as described below, a larger amount of data (more features to learn) results in a more accurate fully learned model. Therefore, using intensity variation data is effective in generating fully learned models.

[0063] Figure 6 This is an image representing an example of the result of wavelet transform processing on blood flow waveform data. Figure 6 In the middle, it is shown that by becoming Figure 3 The spectrum FW1 of period 3 shown is an image generated by performing wavelet transform processing on the blood flow waveform data (original waveform data W1). This image is an example of intensity variation data that emphasizes the intensity of the object. Images representing intensity variation data will be referred to as wavelet images below.

[0064] exist Figure 6 In the diagram, the vertical axis represents frequency (unit: Hz), and the horizontal axis represents time (unit: min). Figure 6 The density in a wavelet image represents intensity [unit: dB]. That is, a wavelet image is data that shows the intensity distribution of the spectrum in a plane defined by frequency and time.

[0065] In the wavelet image of the present embodiment, the frequency band (intensity) can be represented by a gradation of colors. In the wavelet image, for example, a low frequency band can be represented by a color of a cool color system, and a high frequency band can be represented by a color of a warm color system. Specifically, the frequency band can be represented in order from low to high by a dark blue color, a blue color, a light blue color, a yellowish green color, a light yellowish green color, a yellow color, an orange color, and a red color. The frequency band can also be represented by other colors, and can also be represented by a gray scale, as long as the intensity distribution can be visually recognized in the wavelet image. Figure 6 An example of an image in which the wavelet image represented by the above colors is gray-scaled is shown.

[0066] In the wavelet image of Figure 6 , in a first region AR1 in which the intensity distributed along the time axis in the vicinity of 0.2 Hz is higher than the intensity in the frequency band adjacent to the first region AR1, the intensity represented by a red color is shown. Specifically, in the frequency band of about 0.2 Hz ± about 0.05 Hz of the first region AR1, an intensity band represented by a red color is distributed along the time axis, and intensity regions represented by an orange color, a yellow color, a light yellowish green color, and a yellowish green color are distributed around the intensity band. In Figure 6 , a part of the intensity band represented by a red color is indicated by a symbol 101. Further, a part of the intensity regions represented by an orange color, a yellow color, a light yellowish green color, and a yellowish green color is indicated by a symbol 102. On the other hand, in the adjacent frequency band, intensity regions represented by a light blue color, a blue color, and a dark blue color are mainly distributed, but intensity regions represented by a red color, an orange color, and a yellow color are not distributed. In Figure 6 , a part of the intensity regions represented by a light blue color, a blue color, and a dark blue color is indicated by a symbol 103.

[0067] Further, as shown in Figure 6 , in the wavelet image, an intensity band in the shape of a saw blade is formed along the time axis in a frequency band higher than the frequency band of 0.2 to 0.3 Hz. In Figure 6 , in a second region AR2 (a frequency band of about 0.7 Hz or more), an intensity band in the shape of a saw blade is formed. The second region AR2 has an intensity lower than that of the first region AR1 in a frequency band of about 0.9 to 1.0 Hz, and the intensity gradually decreases toward a frequency band lower than 0.9 Hz and a frequency band of 1.0 Hz or more.

[0068] The intensity band in the shape of a saw blade represents an intensity distribution corresponding to a heart rate. The intensity distribution becomes a band shape along the time axis in a state of being asleep, and the band shape becomes more dispersed as the sleep becomes shallower. The intensity distribution corresponding to the heart rate is an intensity distribution that is not obtained in the Fourier transform processing. By using the wavelet image in the generation of the learned model described below, a learned model in which the heart rate is further added can be generated.

[0069] Figure 6One example of a wavelet image in which the object intensity is emphasized is shown, but it is to be noted that even a wavelet image in which the object intensity is not emphasized shows a higher intensity in the first region AR1 than in the adjacent frequency band.

[0070] Generation of a learned model

[0071] In the determination of the sleep stage of the subject, a learned model (an approximator) for determining the sleep stage of the subject can be used. The learned model is a mathematical model of a neuron that mimics the human brain nervous system (a neural network including an input layer, a hidden layer, and an output layer) that is learned to determine the sleep stage of the user. The mathematical model can be, for example, a convolutional neural network (CNN), a recurrent neural network (RNN), or an LSTM (Long Short Term Memory).

[0072] The learning refers to adjusting the strength and bias of the coupling between units so that the correct operation result is output from the output layer. In the present embodiment, in the case of learning, the learning data is input to the input layer. In the hidden layer, the operation based on the operation data is performed on the learning data, and the operation result in the hidden layer is output as the output data from the output layer. The operation data is adjusted by comparing the teaching data and the output data so that the error becomes small. By repeatedly performing this processing on a plurality of learning data, the learned model in which the operation data is adjusted is generated. That is, in the present embodiment, the learned model can be generated by so-called supervised learning using the learning data and the teaching data. The sleep estimation device 51 described later can determine the sleep stage of the subject by using the learned model thus generated.

[0073] The learning data is data that becomes an example for generating the learned model. The learning data can be a spectrum generated from the blood flow waveform data. In the present embodiment, a wavelet image is used. As the wavelet image, the object intensity can be emphasized or not. The learning data can be data in which the behavior differs between the time of wakefulness and the time of sleep, or data in which the behavior changes depending on the depth of sleep. As the learning data, various kinds of data (example: a spectrum of blood flow waveform data showing mutually different waveforms) can be used.

[0074] The teaching data is data obtained by associating a correct answer label with learning data. For example, data obtained by associating a person's sleep stage for which blood flow waveform data is acquired with a frequency spectrum as learning data can be used as teaching data. As described above, the sleep stage of a subject can be determined based on brain wave data detected by an electroencephalograph. As the correct answer label, a symbol indicating each sleep stage can be used. Alternatively, as the correct answer label, a symbol indicating a correct answer can be used for a certain specific sleep stage (e.g., stage 2 or stage 3), and a symbol indicating an incorrect answer can be used for sleep stages other than these. In the present embodiment, as an example of teaching data, data obtained by associating a wavelet image corresponding to stage 2 or stage 3 as a correct answer label with a wavelet image determined to be stage 2 or stage 3 based on brain wave data can be used.

[0075] The operation data is data related to operations for generating a learned model, including operation expressions, variables (e.g., biases and weights) of the operation expressions, and activation functions. The biases and weights define the strength of the combination between units. By adjusting the biases and weights, the accuracy of the learned model can be improved. As an adjustment method of the operation data, for example, the error backpropagation method and the gradient descent method can be used.

[0076] [Embodiment 1]

[0077] An example of a sleep estimation system 1 capable of determining a sleep stage of a subject constructed based on the above-described principle will be described below. The sleep estimation system 1 of the present embodiment can be a system capable of determining a sleep stage of a subject using the above-described learned model.

[0078] [Sleep Stage Estimation System]

[0079] Figure 1 is a block diagram showing an example of the schematic configuration of the sleep estimation system 1 of Embodiment 1. As shown in Figure 1 The sleep estimation system 1 includes an accelerometer 2, a blood flow meter 3, and a portable terminal 5. In the portable terminal 5, for example, a sleep estimation device 51 that determines a sleep stage of a subject by executing an application capable of determining the sleep stage is constructed as a part of the functions of a control unit that comprehensively controls each component of the portable terminal 5.

[0080] [Accelerometer]

[0081] The accelerometer 2 is a sensor capable of detecting acceleration generated by the activity of a subject. The accelerometer 2 can transmit the detected acceleration as acceleration data to the sleep estimation device 51 through wireless or wired communication. The accelerometer 2 is attached to a part of the body of a subject, such as the head or a finger, for example. As the accelerometer 2, a publicly known sensor such as a frequency change type, a piezoelectric type, a piezoresistance type, or an electrostatic capacity type can be used.

[0082] <Blood flow meter>

[0083] The blood flow meter 3 can be a blood flow meter explained in the above principle. The blood flow meter 3 can be, for example, a laser Doppler blood flow meter. In the present embodiment, the blood flow meter 3 can transmit the raw waveform data W1 to the sleep estimation device 51 as the blood flow waveform data. The blood flow meter 3 can also transmit the processed waveform data W2 to the sleep estimation device 51 instead of the raw waveform data W1. The blood flow meter 3 can not generate the processed waveform data W2, and the sleep estimation device 51 can generate the processed waveform data W2. The blood flow meter 3 is equipped, for example, to a part of the body of the subject such as an ear, a finger, a wrist, a forearm, a forehead, a nose, or a neck.

[0084] <Portable terminal>

[0085] The portable terminal 5 can be at least a terminal capable of data communication with the accelerometer 2 and the blood flow meter 3. The portable terminal 5 can be, for example, a smartphone or a tablet. The sleep estimation device 51 is built in the portable terminal 5, and has a storage 52 and a notification 53.

[0086] The storage 52 can store programs and data used by the control section (particularly, the sleep estimation device 51). The storage 52 stores, for example, the learned model generated as described above, and a threshold value for determining whether the subject is at rest.

[0087] The notification 53 can notify various information to the surroundings (example: the subject) of the portable terminal 5. In the present embodiment, the notification 53 can notify various information following the notification instruction from the sleep estimation device 51. The notification 53 can be at least any one of a sound output device that outputs sound, a vibration device that vibrates the portable terminal 5, and a display device that displays an image.

[0088] (Sleep estimation device)

[0089] The sleep estimation device 51 can determine the sleep stage of the subject equipped with the accelerometer 2 and the blood flow meter 3. The sleep estimation device 51 can have a 2nd acquisition section 11, a 2nd determination section 12, a 1st acquisition section 13 (acquisition section), a generation section 14, a 1st determination section 15 (determination section), and a notification section 16.

[0090] The 2nd acquisition section 11 can acquire acceleration data from the accelerometer 2. The 2nd determination section 12 can determine whether the subject is at rest based on the acceleration data acquired by the 2nd acquisition section 11. The 2nd determination section 12 determines, for example, that the subject is at rest in a case where the acceleration indicated by the acceleration data is less than the threshold value stored in the storage 52. The 2nd determination section 12 can transmit determination result data to the generation section 14.

[0091] The first acquisition unit 13 can acquire the blood flow waveform data (raw waveform data W1) from the blood flow meter 3. The generation unit 14 can generate a frequency spectrum of the blood flow waveform data by performing a frequency analysis process on the blood flow waveform data acquired by the first acquisition unit 13. The first determination unit 15 can determine the sleep stage of the subject based on the frequency spectrum generated by the generation unit 14. In other words, the first determination unit 15 can determine the transition of the depth of sleep of the subject based on the frequency spectrum.

[0092] In the present embodiment, the generation unit 14 can generate a wavelet image by performing a wavelet transform process in which the target frequency band is relatively emphasized more than other frequency bands as the frequency analysis process. The wavelet image is process data indicating the result of the time-frequency analysis process of the blood flow waveform data. The first determination unit 15 can determine the sleep stage of the subject based on the wavelet image generated by the generation unit 14.

[0093] In the present embodiment, the first determination unit 15 can determine that the sleep stage of the subject is the period 2 or the period 3 in a case where it is determined that the characteristic waveform Sh is contained in the frequency band of 0.2 to 0.3 Hz of the frequency spectrum. In this case, the first determination unit 15 can determine that the sleep stage of the subject is in transition from the period 1 to the period 2 or the period 3. On the other hand, the first determination unit 15 can determine that the sleep stage of the subject is a sleep stage other than the period 2 or the period 3 in a case where it is determined that the characteristic waveform Sh is not contained in the frequency band of 0.2 to 0.3 Hz. The first determination unit 15 can determine that the sleep stage of the subject is in transition from the period 2 or the period 3 to the period 1 in a case where it is determined that the characteristic waveform Sh is not contained in the frequency band of 0.2 to 0.3 Hz after determining that the subject is in the sleep stage of the period 2 or the period 3. The determination of whether or not the above-described characteristic waveform Sh is contained can be performed, for example, by a determination of whether or not the first intensity is larger than the second intensity by a given value or more.

[0094] In the present embodiment, the first determination unit 15 can perform the determination of the sleep stage of the subject by using the learned model. In this case, the first determination unit 15 can output the determination result of the sleep stage of the subject from the output layer of the learned model by assigning the wavelet image generated by the generation unit 14 as input data to the input layer of the learned model.

[0095] As described above, the learning completed model is generated using, as an example of the teaching data, data associated with the wavelet image whose correct answer label and the case corresponding to period 2 or period 3 are known. For this reason, the first determination section 15 is able to determine that the sleep stage of the subject is period 2 or period 3, and is also able to determine that the transition from period 1 to period 2 or period 3 is made, by giving the wavelet image generated by the generation section 14 to the learning completed model. The first determination section 15 particularly has a possibility of being able to determine that the sleep stage of the subject is period 2 or period 3 with good accuracy, in a case where the spectrum including the characteristic waveform Sh in the frequency band of 0.2 to 0.3 Hz is given to the learning completed model.

[0096] Here, in a case where the sleep stage is estimated using the blood flow waveform data, there is a possibility that the sleep estimation device is not able to determine whether or not the subject is asleep, with only the blood flow waveform data. For example, in a case of a subject whose blood flow is stable even when awake, it can be that no significant difference is seen between the blood flow waveform data when awake and the blood flow waveform data when asleep, and in this case, there is a possibility that the sleep estimation device is not able to determine whether or not the subject is asleep. In the present embodiment, the generation section 14 is able to perform the determination processing based on the first determination section 15 in a state where the possibility that the subject is asleep is high, by performing the frequency analysis processing on the blood flow waveform data in a case where it is determined by the second determination section 12 that the subject is at rest.

[0097] The notification section 16 is able to perform the notification processing based on the determination result of the first determination section 15. The notification section 16 can transmit a notification instruction according to the notification processing to the notification section 53. Thereby, the notification section 53 is able to notify the information according to the notification processing to the surroundings of the portable terminal 5. The notification section 16 can have a first notification section 161 and a second notification section 162.

[0098] As the above-described notification processing, the first notification section 161 is able to perform the first notification processing after a given time elapses from when the transition from period 1 to period 2 or period 3 is detected by the first determination section 15. The first notification processing is the notification processing accompanying the detection of the transition, and for example, can be an alarm processing for prompting the subject to wake up, or a processing that notifies of the detection of the transition. The given time can be appropriately set according to the purpose of the notification, for example, by experiment. In the present embodiment, for example, when the time is counted from the time when the transition from period 1 to period 2 or period 3 is made, the given time is set to a time at which it is estimated that the subject is likely to wake up. By the first notification processing, for example, the subject is able to wake up at a good wake-up timing after falling asleep. The first notification section 161 can also perform the first notification processing when the transition from period 1 to period 2 or period 3 is detected by the first determination section 15.

[0099] The second notification section 162 can execute the second notification process as the notification process described above, after a given time elapses from the detection of the transition to the period 2 or the period 3 by the first determination section 15. The second notification process is a notification process accompanying the detection of the transition, and can be, for example, an alarm process for prompting the subject to wake up, or a process of notifying the detection of the transition. The given time can be appropriately set according to the purpose of the notification, for example, by experiment. In the present embodiment, for example, when the time is counted from the time of the transition from the period 2 or the period 3 to the period 1, the given time can be set to a time at which it is presumed that the subject is likely to wake up. Through the second notification process, the subject can wake up at a good wake-up timing after falling asleep, for example. The second notification section 162 can also execute the second notification process when the transition from the period 2 or the period 3 to the period 1 is detected by the first determination section 15.

[0100] 〔Flow of processing〕

[0101] Figure 7 is a flowchart showing an example of the flow of processing based on the sleep estimation device 51 (sleep estimation method). In a case where the sleep stage of the subject is determined by the sleep estimation device 51, the blood flow meter 3 can start the detection of the blood flow waveform data after the blood flow meter 3 is attached to the subject.

[0102] As shown in Figure 7 , in the sleep estimation device 51, the first acquisition section 13 can acquire the blood flow waveform data from the blood flow meter 3 (S1: first acquisition process, acquisition process). The generation section 14 can generate a frequency spectrum of the blood flow waveform data by performing a frequency analysis process on the blood flow waveform data. In the present embodiment, the generation section 14 can generate a wavelet image in which the object intensity is emphasized by performing a wavelet transform process on the blood flow waveform data (S2: generation process). The first determination section 15 can determine the sleep stage of the subject based on the frequency spectrum. In the present embodiment, the first determination section 15 can determine the sleep stage of the subject by inputting the wavelet image in which the object intensity is emphasized to the learned model (S3) (S4: first determination process, determination process).

[0103] The first determination section 15 can determine whether the transition from the period 1 to the period 2 or the period 3 is made based on the determination result of the sleep stage (S5). In a case where the first determination section 15 determines that the transition from the period 1 to the period 2 or the period 3 is made (YES in S5), the first notification section 161 can determine whether a given time elapses from the determination (S6). The first notification section 161 can execute, as the first notification process, for example, an alarm process that notifies the alarm sound from the notification section 53 in a case where it is determined that the given time elapses (YES in S6) (S7). The notification section 53 can receive the notification instruction from the first notification section 161 and notify the alarm sound.

[0104] In the case where "No" in S5, the processing can return to S1. In the case where "No" in S6, the processing of S6 can be repeated. Further, in S5, in the case where the first determination section 15 determines that the period 2 or the period 3 is transitioned to the period 1, the second notification section 162 can perform the second notification processing, for example, after a given time elapses.

[0105] <Problems in the Related Art and Effects of the Sleep Estimation Device>

[0106] A method of detecting non-rapid eye movement sleep including steps 1 to 4 described below is disclosed in Patent Literature 1.

[0107] • Step 1: a step of generating time series data of an interbeat interval of a heart of a subject.

[0108] • Step 2: a step of setting a window of a given time length moving along a time axis of the time series data, and performing spectral analysis on the time series data in the window with respect to each of a plurality of determination time points on the time axis.

[0109] • Step 3: a step of calculating a concentration degree of a power of a heart rate variability high frequency component from the spectrum of each window.

[0110] • a step of determining whether or not it is non-rapid eye movement sleep based on the calculated concentration degree.

[0111] In the above method, as a detection device of the time series data of step 1, a pulse wave meter or an electrocardiograph is used, for example.

[0112] However, a high level of expertise is required in the handling of the electroencephalograph and the acquisition of the brain waves. Further, the installation of the electroencephalograph is cumbersome. Therefore, it is difficult for the subject to acquire the brain waves simply and easily, and it is difficult to grasp the sleep stage of oneself simply and easily.

[0113] Further, as described above, the inventors found that, in the spectrum of the blood flow waveform data, in the case where a significant intensity change is seen in the frequency band of 0.2 to 0.3 Hz, the subject is highly likely to be in the sleep stage of the period 2 or the period 3.

[0114] The sleep estimation device 51 of the present disclosure can determine the sleep stage of the subject using the blood flow waveform data. A high level of expertise more than the electroencephalograph is not required in the handling of the blood flow meter and the acquisition of the blood flow waveform data. Further, the installation of the blood flow meter is easier than the electroencephalograph. That is, the sleep estimation device 51 can determine the sleep stage of the subject more simply and easily because it can acquire the blood flow waveform data of the subject more simply and easily. Further, according to the sleep estimation device 51, the subject can grasp the sleep stage of oneself simply and easily.

[0115] Further, the sleep estimation device 51 of the present disclosure is able to determine that the sleep stage of the subject is period 2 or period 3 when the frequency spectrum having the above-described significant intensity change that is not seen in the electrocardiogram waveform data is obtained. For this reason, the sleep estimation device 51 is able to increase the likelihood that the sleep stage of the subject is period 2 or period 3 can be accurately estimated.

[0116] [Embodiment 2]

[0117] The following describes other embodiments of the present disclosure. In addition, for the convenience of explanation, the same symbol is attached to a member having the same function as that described in the above-described embodiments, and the description thereof is not repeated. Figure 8 is a block diagram showing an example of the schematic configuration of the sleep estimation system 1A of Embodiment 2.

[0118] In the sleep estimation system 1 of Embodiment 1, the portable terminal 5 is able to acquire the blood flow waveform data from the blood flow meter 3 equipped in the subject, for example, by wireless communication. Further, the sleep estimation device 51 constructed in the portable terminal 5 is able to determine the sleep stage of the subject on the basis of the blood flow waveform data.

[0119] On the other hand, as shown in Figure 8 In the sleep estimation system 1A of Embodiment 2, the blood flow meter 3 can be provided in the wearable device 20. In addition thereto, the sleep estimation device 51 can be constructed in the wearable device 20 as a part of the function of the control section that comprehensively controls each member of the wearable device 20. That is, in the sleep estimation system 1A, the sleep estimation device 51 can be installed in the wearable device 20 together with the blood flow meter 3. For this reason, the acquisition processing of the blood flow waveform data and the determination processing of the sleep stage on the basis of the blood flow waveform data can be performed with one device. Further, various devices or components and the like that are required when wireless or wired communication is performed between two devices are not required. The wearable device 20 can be equipped in the same position as the position in which the blood flow meter 3 is equipped in the subject.

[0120] [Modified Example]

[0121] The above describes the invention related to the present disclosure on the basis of each drawing and the embodiments. However, the invention related to the present disclosure is not limited to each of the above-described embodiments. That is, the invention related to the present disclosure can be variously changed within the scope shown in the present disclosure, and an embodiment obtained by appropriately combining the disclosed technical means for different embodiments is also included in the technical scope of the invention related to the present disclosure. That is, it is to be noted that various modifications or corrections can be easily made on the basis of the present disclosure by those skilled in the art. Further, it is to be noted that these modifications or corrections are included in the scope of the present disclosure.

[0122] (Modified Example of Frequency Resolution Processing)

[0123] For example, the generation section 14 can generate a wavelet transform image by performing wavelet transform processing on the blood flow waveform data, and does not necessarily have to perform wavelet transform processing to emphasize the object intensity.

[0124] Further, the generation section 14 can perform a time-frequency analysis processing other than wavelet transform processing on the blood flow waveform data. The generation section 14 can perform, for example, short-time Fourier transform processing on the blood flow waveform data. The short-time Fourier transform processing performs Fourier transform processing on a plurality of waveform data cut out along a time axis using a window function, respectively. In this case, the generation section 14 can generate the same image (intensity change data) as the wavelet image by performing short-time Fourier transform processing that relatively emphasizes the object intensity more than other frequency bands, as the processing data.

[0125] Further, the generation section 14 can perform a processing other than the time-frequency analysis processing as the frequency analysis processing. The generation section 14 can perform, for example, Fourier transform processing. In a case where the generation section 14 performs Fourier transform processing, the generation section 14 can generate, for example, a spectrum Fw1 or Fw2 shown in FIG. 9. Figure 3 Further, the generation section 14 can generate a spectrum (waveform) similar to the spectrum Fw1 or Fw2 by performing short-time Fourier transform processing. Figure 3

[0126] As such, the generation section 14 can generate various spectra. Therefore, the first determination section 15 can determine the sleep stage of the subject by inputting the various spectra to the learned model. The learned model is generated by using a spectrum similar to the spectrum generated by the generation section 14 as learning data and teaching data.

[0127] (Modified example of determination processing of the first determination section)

[0128] The learned model can be stored in the storage section 52 of the portable terminal 5 or the wearable device 20. That is, the sleep estimation system 1 or 1A can determine the sleep stage of the subject without using the learned model.

[0129] For example, the first determination section 15 can determine that the sleep stage of the subject is the period 2 or the period 3 in a case where it is determined that the first intensity is larger than the second intensity by a given value or more in the spectrum generated by the generation section 14. The given value can be stored in the storage section 52 instead of the learned model.

[0130] ​Further, in a case where the first determination section 15 determines the sleep stage of the subject using the spectrum FW1, it can determine that the sleep stage of the subject is period 2 or period 3 when a characteristic waveform Sh can be extracted in the frequency band of 0.2 to 0.3 Hz. In this case, in the storage section 52, instead of the learned model, a reference waveform of the characteristic waveform Sh that can be extracted in the spectrum FW1 can be stored. The first determination section 15 can determine that the characteristic waveform Sh can be extracted in the frequency band of 0.2 to 0.3 Hz of the spectrum FW1 in a case where it is determined that a waveform that coincides with the reference waveform exists in the frequency band. The first determination section 15 can determine that a waveform that coincides with the reference waveform exists in the frequency band of 0.2 to 0.3 Hz in a case where the degree of coincidence of the waveform contained in the frequency band of 0.2 to 0.3 Hz with the reference waveform is equal to or higher than a threshold value set through experiments, for example. Further, the first determination section 15 can determine whether or not the characteristic waveform Sh is contained in the frequency band of 0.2 to 0.3 Hz by other indexes (example: the degree of change in the inclination of the waveform). Furthermore, as for the spectrum FW2, the same determination as the spectrum FW1 can be performed.

[0131] (Modified example of sleep estimation system 1 or 1A)

[0132] It can not be that the acceleration generated by the activity of the subject is detected by the accelerometer 2 and the sleep estimation device 51 determines the sleep state of the subject based on the acceleration. In this case, the sleep estimation system 1 or 1A can not be provided with the accelerometer 2, and the sleep estimation device 51 can not be provided with the second acquisition section 11 and the second determination section 12.

[0133] (Example of software-based implementation)

[0134] The control block of the sleep estimation device 51 can be implemented by a logic circuit (hardware) formed as an integrated circuit (IC chip) or the like, or can be implemented by software.

[0135] In the latter case, the sleep estimation device 51 has a computer that executes a command of a program that is software that realizes each function. The computer has, for example, at least one processor (control device), and has at least one recording medium that is readable by the computer and that stores the above program. Also, in the above computer, the above program is read from the above recording medium by the above processor and executed, and the purpose of the present disclosure is achieved. As the above processor, for example, a CPU (Central Processing Unit) can be used. As the above recording medium, in addition to a "non-temporary tangible medium" such as a ROM (Read Only Memory) or the like, a tape, a disk, a card, a semiconductor memory, a programmable logic circuit, or the like can be used. Furthermore, a RAM (Random Access Memory) or the like that expands the above program can also be provided. Furthermore, the above program can be provided to the above computer via any transmission medium (communication network, carrier wave, or the like) that can transmit the program. In addition, one embodiment of the present disclosure can also be realized in the form of a data signal embedded in a carrier wave by electronically transmitting the above program.

[0136] Symbol explanation

[0137] 1, 1A sleep estimation system

[0138] 3 blood flow meter

[0139] 11 second acquisition unit

[0140] 12 second determination unit

[0141] 13 first acquisition unit (acquisition unit)

[0142] 14 generation unit

[0143] 15 first determination unit (determination unit)

[0144] 20 wearable device

[0145] 51 sleep estimation device

[0146] 161 first notification unit

[0147] 162 second notification unit

Claims

1. A sleep estimation apparatus comprising: a first acquisition unit configured to acquire blood flow data representing blood flow of a subject; a generation unit configured to generate a frequency spectrum of the blood flow data by performing a frequency analysis process on the blood flow data; and a first determination unit configured to determine a sleep stage of the subject based on the frequency spectrum, wherein, when a sleep stage in non-rapid eye movement sleep is set in order from a light sleep stage as a period 1, a period 2, and a period 3, in a case where the frequency spectrum in which a maximum intensity in a first range, which is a part of a region including a frequency band of 0.2 Hz or more and less than 0.3 Hz, is larger than a second intensity in a second range, which is a part of the region other than the first range, is generated as the frequency spectrum, the first determination unit determines that the sleep stage of the subject transitions from the period 1 to the period 2 or the period 3, and the frequency spectrum generated by the generation unit in a case where the sleep stage of the subject is determined by the first determination unit to transition from the period 1 to the period 2 or the period 3 includes a wide waveform that is convex upward in the frequency band with the first intensity as a center. 2.The sleep estimation apparatus according to claim 1, wherein the sleep estimation apparatus comprises: a first notification unit configured to perform a first notification process when the transition to the period 2 or the period 3 from the period 1 is determined by the first determination unit or after a lapse of a given time after the transition to the period 2 or the period 3 from the period 1 is determined. 3.The sleep estimation apparatus according to claim 1, wherein the generation unit generates, as the frequency spectrum, intensity variation data representing variation of intensity in each frequency band with time by performing a time-frequency analysis process as the frequency analysis process, the first determination unit determines that the sleep stage of the subject transitions from the period 1 to the period 2 or the period 3 by giving the intensity variation data generated by the generation unit to a learned model that is learned using teaching data in which a correct answer label and the intensity variation data in which it is known that the period 2 or the period 3 corresponds are associated. 4.The sleep estimation apparatus according to any one of claims 1 to 3, wherein the sleep estimation apparatus comprises: a second acquisition unit configured to acquire acceleration data representing acceleration generated by activity of the subject; and a second determination unit configured to determine whether the subject is at rest based on the acceleration data, and the generation unit performs the frequency analysis process in a case where the subject is determined to be at rest by the second determination unit. 5.A sleep estimation apparatus comprising: an acquisition unit configured to acquire blood flow data representing blood flow of a subject; a generation unit configured to generate a frequency spectrum representing a result of a time-frequency analysis process of the blood flow data by performing a wavelet transform process or a short-time Fourier transform process in which intensity in a given frequency band is more relatively emphasized than in other frequency bands on the blood flow data; and a determination unit configured to determine a sleep stage of the subject based on the frequency spectrum. ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ ​ In a case where, as the frequency spectrum, a frequency spectrum in which a maximum intensity, that is, a first intensity in a first range including a part of a region of a frequency band of 0.2 Hz or more and less than 0.3 Hz is greater than a second intensity in a second range including a part of the region other than the first range is generated, In a case where, as the frequency spectrum, a frequency spectrum in which a maximum intensity, that is, a first intensity in a first range including a part of a region of a frequency band of 0.2 Hz or more and less than 0.3 Hz is greater than a second intensity in a second range including a part of the region other than the first range is generated, The frequency spectrum generated by the generation unit in a case where it is determined by the determination unit that the sleep stage of the subject transitions from the period 1 to the period 2 or the period 3 includes a wide waveform that is convex upward in the frequency band with the first intensity as a center.

6. The sleep estimation device according to claim 5, wherein The generation unit generates intensity change data indicating a change in intensity in each frequency band over time within a given time as the frequency spectrum, The determination unit determines that the sleep stage of the subject transitions from the period 1 to the period 2 or the period 3 by giving the intensity change data generated by the generation unit to a learning completed model that has learned using teaching data in which a correct answer label and the intensity change data in which it is known that the period 2 or the period 3 corresponds are associated.

7. A sleep estimation system comprising: the sleep estimation device according to any one of claims 1 to 6; and a blood flow meter that detects the blood flow data by receiving scattered light generated by irradiating light to a blood vessel of the subject.

8. A wearable device comprising: the sleep estimation device according to any one of claims 1 to 6; and a blood flow meter that detects the blood flow data by receiving scattered light generated by irradiating light to a blood vessel of the subject.

9. A sleep estimation method comprising: a first acquisition step of acquiring blood flow data indicating blood flow of a subject; a generation step of generating a frequency spectrum of the blood flow data by performing frequency analysis processing on the blood flow data; and a first determination step of determining a sleep stage of the subject based on the frequency spectrum, In a case where, as the frequency spectrum, a frequency spectrum in which a maximum intensity, that is, a first intensity in a first range including a part of a region of a frequency band of 0.2 Hz or more and less than 0.3 Hz is greater than a second intensity in a second range including a part of the region other than the first range is generated, In a case where, as the frequency spectrum, a frequency spectrum in which a maximum intensity, that is, a first intensity in a first range including a part of a region of a frequency band of 0.2 Hz or more and less than 0.3 Hz is greater than a second intensity in a second range including a part of the region other than the first range is generated, The frequency spectrum generated by the generation unit in a case where it is determined by the determination unit that the sleep stage of the subject transitions from the period 1 to the period 2 or the period 3 includes a wide waveform that is convex upward in the frequency band with the first intensity as a center.

10. A sleep estimation method comprising: a blood flow data acquisition step of acquiring blood flow data representing blood flow of a subject; a generation step of generating a spectrum representing a result of time-frequency resolution processing of the blood flow data by performing wavelet transform processing or short-time Fourier transform processing in which intensity in a given frequency band is more relatively emphasized than in other frequency bands, on the blood flow data; and a determination step of determining a sleep stage of the subject based on the spectrum, when the sleep stage in non-rapid eye movement sleep is set in order from a light sleep stage as a period 1, a period 2, and a period 3, in a case where, as the spectrum, a spectrum in which a first intensity that is a maximum intensity in a first range that is a part of a region including a frequency band of 0.2 Hz or more and less than 0.3 Hz is greater than a second intensity in a second range that is a part of the region other than the first range is generated, in the determination step, it is determined that the sleep stage of the subject transitions from the period 1 to the period 2 or the period 3, the spectrum generated by the generation step in a case where it is determined by the determination step that the sleep stage of the subject transitions from the period 1 to the period 2 or the period 3 includes a wide waveform that is convex upward in the frequency band with the first intensity as a center.

Images