Sleep state determination system and sensing system

Abstract

The present disclosure addresses the problem of providing a sleep state determination system that is capable of determining a sleep state in more detail. A sleep state determination system (1) comprises a first extraction part (121), a detection part (122), and a first determination part (123). The first extraction part (121) extracts a respiratory waveform based on respiration of a user (H1) from a sensing result that has been obtained by using a radio wave sensor (10) to sense the movement of the body of the user (H1) who is asleep and acquires a feature amount of the respiratory waveform. The detection part (122) detects, on the basis of the sensing result, the occurrence of an apnea event, which is a continuation of the state in which the respiration of the user (H1) is stopped for a certain period of time or longer, and acquires the frequency of the occurrence of the apnea event. The first determination part (123) determines whether or not the sleep state of the user (H1) is abnormal on the basis of the feature amount that has been acquired by the first extraction part (121) and the occurrence frequency that has been acquired by the detection part (122).

Description

Sleep state determination system and sensing system

[0001] The present disclosure relates to a sleep state determination system and a sensing system, and more particularly to a sleep state determination system that determines the sleep state using sensing results obtained by sensing the body movements of a user while sleeping with a sensor, and a sensing system that includes such a sleep state determination system and a sensor.

[0002] The environmental control system described in Patent Document 1 includes a first acquisition unit, a determination unit, and a control unit. The first acquisition unit acquires input information including physiological index information (e.g., heart rate) indicating a physiological index of a user from a sensor. The determination unit determines a control content for putting the user to sleep from the input information in accordance with a control content determination rule. The control unit controls an environmental control device based on the determined control content.

[0003] The environmental control system of Patent Document 1 performs control such as shortening the sleep induction period and lengthening the sleep period when it determines that the user is highly sleepy based on physiological index information. However, in recent years, there has been a demand for more detailed determination of the sleep state.

[0004] Japanese Patent Application Laid-Open No. 2020-103536

[0005] An object of the present disclosure is to provide a sleep state determination system and a sensing system that can determine the sleep state in more detail.

[0006] A sleep state determination system according to one aspect of the present disclosure includes an extraction unit, a detection unit, and a determination unit. The extraction unit extracts a respiratory waveform based on the user's breathing from a sensing result obtained by sensing the user's body movements while sleeping using a sensor, and acquires feature quantities of the respiratory waveform. The detection unit detects the occurrence of an apnea event, in which the user's breathing stops for a certain period of time or longer, based on the sensing result, and acquires the frequency of the apnea event. The determination unit determines whether the user's sleep state is abnormal based on the feature quantities acquired by the extraction unit and the frequency of the apnea event acquired by the detection unit.

[0007] A sensing system according to one aspect of the present disclosure includes the sleep state determination system and the sensor.

[0008] FIG. 1 is a block diagram of a sensing system including a sleep state determination system according to an embodiment of the present disclosure. FIG. 2 is a signal flow diagram of the sleep state determination system. FIG. 3 is a flowchart of a first determination process by a first determination unit included in the sleep state determination system. FIG. 4 is a flowchart of a second determination process by a second determination unit included in the sleep state determination system. FIG. 5A is a waveform diagram showing a sensing result (Doppler signal) of a radio wave sensor constituting the sensing system. FIG. 5B is a graph showing a temporal change in body movement intensity obtained from the sensing result. FIG. 6 is a waveform diagram showing a first example of the sensing result. FIG. 7 is a waveform diagram showing a respiratory waveform (normal respiratory waveform) extracted from the first example. FIG. 8 is a conceptual diagram showing feature quantities of the respiratory waveform. FIG. 9 is a waveform diagram showing a second example of the sensing result. FIG. 10 is a waveform diagram showing a respiratory waveform (Cheyne-Stokes type respiratory waveform) extracted from the second example. 11 is a block diagram of a sensing system including a modified example of the sleep state determination system of the same, and FIG. 12 is a signal flow diagram of the sleep state determination system according to the modified example.

[0009] This disclosure describes a sleep state determination system that extracts a respiratory waveform and detects apnea events from the results of sensing a user's body movements using a radio wave sensor, and determines whether the user's sleep state is abnormal based on the features of the extracted respiratory waveform and the frequency of the detected apnea events.

[0010] (1) Overview As shown in Fig. 1 , a sleep state determination system 1 according to an embodiment of the present disclosure includes a first extraction unit 121, a detection unit 122, and a first determination unit 123. In the description of the overview, the first extraction unit 121 will be simply referred to as the "extraction unit 121," and the first determination unit 123 will be simply referred to as the "determination unit 123."

[0011] (1-1) Extraction Unit The extraction unit 121 extracts a respiratory waveform based on the breathing of the user H1 from the sensing results obtained by sensing the body movements of the user H1 while he / she is sleeping at home using the radio wave sensor 10, and acquires feature quantities of the respiratory waveform. The body movements are, for example, movements of the body surface.

[0012] The extraction unit 121 includes, for example, a first trained model M1, as shown in Fig. 2. The first trained model M1 is a trained model that has been trained in advance by machine learning using an external learning device, using as training data the results of sensing the body surface movements of a large number of subjects, representing a variety of users, while they are sleeping, using the radio wave sensor 10, and respiratory waveforms obtained at the same time by another contact-type respiratory sensor that can sense the respiratory waveforms. Note that machine learning, the first trained model M1, and a method for generating the first trained model M1 will be described later.

[0013] (1-1-1) Body Surface Movement Body surface movement includes movement due to breathing, movement due to heartbeat, and movement due to bodily movements such as turning over in bed.

[0014] (1-1-2) Radio Wave Sensor In this embodiment, the radio wave sensor 10 measures the body surface movements of the sleeping user H1. The radio wave sensor 10 is preferably a Doppler type sensor, but may be a sensor of a type other than Doppler, such as an FMCW (Frequency Modulated Continuous Wave) type or a pulse type. Note that the body surface movements can also be measured by sensors other than a radio wave sensor, such as a distance sensor, a displacement sensor, an acceleration sensor, or a pressure sensor (see modified examples related to sensors).

[0015] The sensing result is, for example, a Doppler signal from the radio wave sensor 10 using the Doppler method, or time-series data obtained by sampling the Doppler signal at a predetermined period. The time-series data is a set of pairs of signal values ​​and time information, or data in which multiple signal values ​​are arranged in time series.

[0016] The Doppler signal or time-series data includes multiple types of components corresponding to multiple types of movements, such as movements based on breathing, movements based on heartbeats, and movements based on body movement. For example, by plotting multiple points (pairs of signal values ​​and time information) constituting the time-series data with the vertical axis representing the amplitude of the Doppler signal (i.e., values ​​indicating body surface movement) and the horizontal axis representing time, a waveform indicating body surface movement (movement waveform) such as that shown in Fig. 5A can be obtained. The movement waveform is the result of superimposing multiple types of waveforms, such as a waveform based on breathing (respiratory waveform), a waveform based on heartbeats (heartbeat waveform), and a waveform based on body movement (body movement waveform).

[0017] (1-1-3) Respiratory Waveform The respiratory waveform is a movement component based on respiration among the various components of the sensing results (Doppler signal), and is extracted from the sensing results using the first trained model M1.

[0018] The respiratory waveform feature quantity includes a time feature quantity and an amplitude feature quantity. The time feature quantity is, for example, a quantity related to a period or periodic fluctuation. The amplitude feature quantity is, for example, a quantity related to an amplitude or amplitude fluctuation. The feature quantities will be described in detail later.

[0019] (1-2) Detection Unit The detection unit 122 detects the occurrence of an apnea event based on the sensing results of the radio wave sensor 10, and acquires the frequency of occurrence of the apnea event. An apnea event is an abnormal breathing event in which the user H1 stops breathing for a certain period of time or more. The certain period of time is, for example, 10 seconds, but is not limited to 10 seconds. The frequency of occurrence is the number of occurrences per unit time (for example, 1 hour) (times / unit time).

[0020] The detection unit 122 includes, for example, a second trained model M2, as shown in Fig. 2. The second trained model M2 is a trained model that is trained in advance by machine learning using an external learning device as training data the sensing results of the radio wave sensor 10 and time-series data related to the medical staff's determination of the occurrence of an apnea event based on the test results of a PSG (Polysomnography) test. The second trained model M2 and a method for generating the second trained model M2 will be described later.

[0021] Sleep apnea syndrome is broadly divided into two types: obstructive sleep apnea syndrome and central sleep apnea syndrome, and the majority of cases are said to be obstructive. Obstructive sleep apnea syndrome is accompanied by labored breathing (movement of the chest and abdomen), making it difficult to distinguish from normal breathing in the respiratory waveform extracted by the first trained model M1 from the sensing results of the radio wave sensor 10, which senses micromovements on the body surface. In other words, if apnea is detected based on the respiratory waveform output from the first trained model M1, it is difficult to detect apnea caused by obstructive sleep apnea syndrome, making it difficult to improve detection accuracy. Therefore, in this embodiment, a second trained model M2 is provided separately from the first trained model M1 to detect apnea events based on the sensing results input to the first trained model instead of the respiratory waveform output from the first trained model M1.

[0022] (1-3) Determination Unit The determination unit 123 determines whether the sleep state of the user H1 is abnormal based on the feature amounts acquired by the extraction unit 121 and the occurrence frequencies acquired by the detection unit 122. The sleep state refers to a state related to the sleep of the user H1 while asleep. The sleep state may be, for example, a state related to the depth of sleep, such as whether the user is in deep sleep or light sleep. Alternatively, the sleep state may be a state related to the quality of sleep, such as whether deep sleep and light sleep, and REM sleep and non-REM sleep, are present in a balanced manner in a predetermined pattern, or whether they deviate from the predetermined pattern and are lacking in balance.

[0023] An abnormal state is a state that differs from a normal state. A normal state is, for example, the normal state of the individual user H1, but may also be a state that is medically considered healthy. The normal state of the individual user H1 is, for example, the average state of the user H1 over a certain period of time in the past. A state that is medically considered healthy may, for example, be the average of multiple states corresponding to multiple people excluding sick people. An abnormal state is, for example, a state of poor sleep quality or shallow sleep, but may also be a state in which a disease such as sleep apnea syndrome (hereinafter sometimes referred to as "SAS") is suspected.

[0024] The determination unit 123 is realized by, for example, an algorithm as shown in FIG.

[0025] (1-4) Advantages As described above, the sleep state determination system 1 of this embodiment can determine the sleep state in more detail by using the feature amount of the respiratory waveform and the occurrence frequency of apnea events.

[0026] (2) Details The extraction unit 121 described in the overview will be referred to as the first extraction unit 121 in the following description. Similarly, the determination unit 123 described in the overview will be referred to as the first determination unit 123 in the following description.

[0027] (2-1) Sensing System A sensing system 100 according to an embodiment of the present disclosure includes a radio wave sensor 10 and a sleep state determination system 1, as shown in FIG.

[0028] (2-1-1) Radio wave sensor The radio wave sensor 10 senses the movement of the body surface of the user H1 who is sleeping on the bed B1 installed in the room R1, and outputs the sensing result (Doppler signal). The sensing may be performed without the bed or from above the bed.

[0029] The radio wave sensor 10 senses body surface movement using radio waves. Millimeter waves are suitable for use as radio waves because of their high resolution, but radio waves with wavelengths other than millimeter waves may also be used.

[0030] In this way, the sleep state determination system 1 can accurately determine the sleep state of the user H1 in a non-contact manner based on the sensing results of the radio wave sensor 10.

[0031] (2-2) Details of the Sleeping State Determination System Next, we will explain the details of the sleeping state determination system 1. Note that, in the following, explanations of the matters already mentioned will be omitted or simplified.

[0032] The sleep state determination system 1 determines the sleep state of the user H1 (for example, whether the sleep state is abnormal or not) based on the sensing result of the radio wave sensor 10.

[0033] The sleep state determination system 1 is realized by a server connected to a network such as a local area network (LAN). The server that realizes the sleep state determination system 1 includes a processor, a memory, a communication module, etc. The processor operates based on programs and various information stored in the memory and cooperates with the communication module, etc. to realize the functions of the components shown in Fig. 1 (the storage unit 11, the processing unit 12, the reception unit 13, the output unit 14, the reference storage unit 111, the history storage unit 112, the first extraction unit 121, the detection unit 122, the first determination unit 123, the second determination unit 124, the second extraction unit 125, the estimation unit 126, the environment control unit 127, and the notification unit 128).

[0034] 1 , the sleep state determination system 1 includes a storage unit 11, a processing unit 12, a receiving unit 13, and an output unit 14. The storage unit 11 stores various types of information. The various types of information include, for example, information stored in a reference storage unit 111 and information stored in a history storage unit 112. The storage unit 11 also stores timing information indicating the execution timing of the second determination process. The timing information may be, for example, information indicating a predetermined period ΔT (e.g., one week, one month, etc.) described below.

[0035] The processing unit 12 performs various types of processing, such as processing by a first extraction unit 121, processing by a detection unit 122, processing by a first determination unit 123, processing by a second determination unit 124, processing by a second extraction unit 125, processing by an estimation unit 126, processing by an environment control unit 127, and processing by a notification unit 128. The processing unit 12 also performs some of the processing described in the operation example (e.g., determining the timing of execution of the second determination processing).

[0036] The receiving unit 13 receives various types of information, such as the sensing results of the radio wave sensor 10.

[0037] The output unit 14 outputs various types of information. Examples of the various types of information include notifications regarding the determination results of the first determination unit 123. Other information output by the output unit 14 will be explained as appropriate. The output by the output unit 14 may be, for example, a video output to a display, but may also be an audio output from a speaker, a printout by a printer, recording on a recording medium, or transmission to an external device.

[0038] (2-2-1) Storage Unit As shown in FIG. 1, the storage unit 11 includes a reference storage unit 111 and a history storage unit 112.

[0039] (2-2-1a) Criterion Storage Unit The criteria storage unit 111 stores in advance first criteria information related to feature amounts and second criteria information related to occurrence frequencies.

[0040] The first reference information is reference information related to a feature quantity. In this embodiment, the first reference information is first abnormal range information indicating a range of a feature quantity (FV) corresponding to an abnormal state. For example, the first abnormal range information is "FV≦RV1, RV2≦FV," which indicates a range equal to or less than the first reference value RV1 or equal to or greater than the second reference value RV2 (where RV2>RV1), with respect to a pair of reference values, a first reference value RV1 and a second reference value RV2 (where RV2>RV1).

[0041] Alternatively, the first abnormal range information may be, for example, "RV1≦FV≦RV2", which indicates a range greater than or equal to the first reference value RV1 and less than or equal to the second reference value RV2, or "FV≦RV0", which indicates a range less than or equal to the reference value RV0 for a single reference value RV0, or "RV0≦FV", which indicates a range greater than or equal to the reference value RV0.

[0042] Alternatively, the first reference information may be first normal range information (for example, "RV1≦FV≦RV2") that indicates a range of feature amounts corresponding to a normal state.

[0043] Specifically, the first reference information includes, for example, disease reference information related to a disease and environmental reference information related to the environment. The disease here refers to a disease that affects the respiratory waveform, and the disease that affects the respiratory waveform may include central SAS. The disease reference information is reference information related to a disease, and is, for example, a set of feature quantities (including waveform data) of respiratory waveforms specific to various diseases (e.g., the Cheyne-Stokes respiration waveform and the Biot respiration waveform shown in FIG. 10 ). The Cheyne-Stokes respiration waveform will be described later.

[0044] The environmental standard information is standard information related to the environment, and is, for example, a set of respiratory waveform features (including waveform data) that are specific to various environments (such as high temperature and humidity environments and noisy environments).

[0045] In addition, the first reference information in this embodiment may further include feature data indicating feature amounts of the user's (individual's) past respiratory waveform.

[0046] The second reference information is reference information regarding the frequency of occurrence of apnea events. In this embodiment, the second reference information is second abnormal range information indicating a range of occurrence frequency (Fr) corresponding to an abnormal state. The second abnormal range information is, for example, "Fr≧5 [occurrences / hour]" indicating a range equal to or greater than a reference value (e.g., 5 times / hour).

[0047] Alternatively, the second reference information may be second normal range information (for example, "Fr<5 [times / hour]") that indicates a range of occurrence frequencies corresponding to a normal state.

[0048] The reference storage unit 111 also stores third reference information in advance. The third reference information is reference information related to the estimation history and is information for determining whether the transition of sleep stages (described later) is a good transition or not. The estimation history is information indicating the history of sleep stages (i.e., time transition) estimated by the estimation unit 126, and is, for example, a set of pairs of sleep stages and time information. The estimation unit 126 and sleep stages will be described later.

[0049] In addition, the reference memory unit 111 further stores in advance at least one piece of reference information corresponding to at least one extraction history stored in the history memory unit 112, from among the fourth reference information regarding the first extraction history and the fifth reference information regarding the second extraction history.

[0050] (2-2-1b) History Storage Unit The history storage unit 112 stores a history related to the processing of the processing unit 12. The history storage unit 112 stores, for example, the estimation history of the estimation unit 126.

[0051] The history storage unit 112 further stores a first extraction history of the first extraction unit 121 , a detection history of the detection unit 122 , and a second extraction history of the second extraction unit 125 .

[0052] The first extraction history is a history of features of the respiratory waveform, and includes, for example, feature information over a predetermined period ΔT. The predetermined period ΔT is a time longer than the time (e.g., 1 second, 10 seconds, etc.) corresponding to the operating cycle of the first extraction unit 121, the detection unit 122, etc., and is, for example, 1 day, 1 week, 1 month, 6 months, 1 year, etc. The feature information is a set of pairs of feature and time information.

[0053] The detection history is a history of the occurrence frequency of apnea events, and includes, for example, detection frequency information over a predetermined period ΔT. The detection frequency information is a set of pairs of a first detection frequency and time information.

[0054] The second extracted history is a history of body movement information related to the body movement of the user H1, and includes, for example, a set of pairs of body movement information and time information over a predetermined period ΔT. Body movement refers to general body movements, such as turning over in bed. Details of body movement and body movement information will be described later.

[0055] The output unit 14 may output a hypnogram (sleep chart) based on the estimated history stored in the history storage unit 112, for example.

[0056] (2-2-2) Processing Unit As shown in FIG. 1, the processing unit 12 includes a first extraction unit 121, a detection unit 122, a first determination unit 123, a second determination unit 124, a second extraction unit 125, an estimation unit 126, an environment control unit 127, and a notification unit 128.

[0057] (2-2-2a) First Determination Unit The first determination unit 123 determines that the sleep state is abnormal if (A) a first comparison result obtained by comparing the feature amount of the respiratory waveform with first reference information satisfies a first condition, or (B) a second comparison result obtained by comparing the frequency of apnea events with second reference information satisfies a second condition. Note that the sleep state is also determined to be abnormal if the first comparison result satisfies the first condition and the second comparison result satisfies the second condition.

[0058] In this embodiment, the first comparison result is information indicating whether the feature value FV is within the range of the first abnormal range information (e.g., "FV≦RV1, RV2≦FV"). However, the first comparison result may be information indicating whether the feature value FV is outside the range of the first normal range information (e.g., "RV1≦FV≦RV2").

[0059] Specifically, the first comparison result includes, for example, a disease comparison result in which the feature amount of the respiratory waveform is compared with disease reference information, and an environment comparison result in which the feature amount of the respiratory waveform is compared with environmental reference information. The disease comparison result is, for example, information indicating the similarity between the feature amount of the respiratory waveform and the disease reference information. The environment comparison result is, for example, information indicating the similarity between the feature amount of the respiratory waveform and the environmental reference information.

[0060] In this embodiment, the first condition is that the feature value FV is within the range of the first abnormal range information (e.g., "FV≦RV1 or RV2≦FV"). However, the first condition may also be that the feature value FV is outside the range of the first normal range information (e.g., "RV1≦FV≦RV2").

[0061] The first condition includes, for example, a disease condition related to the disease criteria information and an environmental condition related to the environmental criteria information. The disease condition is a condition for determining that a sleep abnormality is caused by a disease, such as a condition that the similarity indicated by the disease comparison result is equal to or greater than a first threshold. The environmental condition is a condition for determining that a sleep abnormality is caused by an environment, such as a condition that the similarity indicated by the environment comparison result is equal to or greater than a second threshold.

[0062] In this embodiment, the second comparison result is information indicating whether the feature value FV is within the range of the first abnormal range information (e.g., "FV≦RV1 or RV2≦FV"). However, the second comparison result may be information indicating whether the feature value FV is outside the range of the first normal range information (e.g., "RV1≦FV≦RV2").

[0063] In this embodiment, the second condition is that the occurrence frequency Fr is within the range of the second abnormal range information (e.g., "Fr≧5 times / hour"). However, the second condition may also be that the occurrence frequency Fr is outside the range of the second normal range information (e.g., "Fr<5 times / hour").

[0064] In this way, the first judgment unit 123 compares the feature amount with the first reference information and the occurrence frequency with the second reference information, and makes a judgment based on the two types of comparison results, so that the sleep state judgment system 1 can judge the sleep state in more detail and with greater accuracy.

[0065] Furthermore, if the second comparison result satisfies the second condition, the first determination unit 123 determines that SAS has been detected. The SAS here includes not only central SAS but also obstructive SAS. This allows the first determination unit 123 to accurately determine whether SAS has been detected.

[0066] Specifically, for example, if the disease comparison result satisfies the disease condition or the environment comparison result does not satisfy the environmental condition, the first determination unit 123 determines that the abnormal condition is caused by a disease. The disease here may include not only diseases other than SAS but also central SAS. This improves the accuracy of determining whether a disease has been detected. Furthermore, if the disease comparison result does not satisfy the disease condition or the environmental comparison result satisfies the environmental condition, the first determination unit 123 determines that the abnormal condition is caused by the environment of the space in which the user H1 is sleeping.

[0067] This allows the first determination unit 123 to further determine whether the SAS is caused by a disease or the environment when it is determined that the SAS has not been detected.

[0068] In addition, when determining whether the user's sleeping state is abnormal or not, the first determination unit 123 may determine that the state is abnormal if the difference between the feature acquired by the first extraction unit 121 and the feature data included in the first reference information (i.e., the feature of the user's own breathing in the past) is greater than or equal to a predetermined value.

[0069] Furthermore, the first determination unit 123 may determine that an abnormal state exists when a state in which the acquired feature quantity matches the disease criteria information included in the first reference information or a state in which the acquired feature quantity deviates from the feature data included in the first reference information (i.e., the difference between the feature quantity and the feature data is equal to or greater than a predetermined value) continues for a predetermined time or more. The predetermined time is, for example, but not limited to, one hour, 30 minutes, or the like.

[0070] (2-2-2b) Second Extraction Unit The second extraction unit 125 acquires body movement information based on the sensing results.

[0071] The body movement information is information that quantitatively indicates the body movement, and in particular, information regarding the frequency components of the body movement and the time change in the strength of the body movement. The body movement information is acquired by the second extraction unit 125 based on the sensing result of the radio wave sensor 10.

[0072] The body movement information includes, for example, a power spectrum. The power spectrum is information about the frequency components of body movement and indicates the frequency distribution of the power (received signal strength) of the sensing results. The power spectrum is obtained by performing an FFT (Fast Fourier Transform) on the sensing results. Specifically, the power spectrum is, for example, a function that represents a curve on a graph with frequency on the horizontal axis and power on the vertical axis, or a data table that includes information equivalent to such a function.

[0073] Furthermore, the body movement information may include, for example, body movement intensity (see FIG. 5B , described below) in addition to or instead of the power spectrum. Body movement intensity is the strength of body movement and is the result of integrating the power spectrum over the frequency range of the evaluation target. The frequency range of the evaluation target is, for example, a range of several tens of Hz ±10 Hz, but is not limited to this.

[0074] (2-2-2c) Estimation Unit The estimation unit 126 estimates a sleep stage based on the feature amount acquired by the first extraction unit 121 and the body movement information acquired by the second extraction unit 125. A sleep stage is an evaluation value that indicates a stage of sleep.

[0075] The sleep stage is a value that indicates, for example, whether REM sleep or non-REM sleep and the depth of non-REM sleep. Specifically, the sleep stage is composed of four stages: stage R indicating REM sleep, stage N1 indicating light non-REM sleep, stage N2 indicating medium-depth non-REM sleep, and stage N3 indicating deep non-REM sleep.

[0076] However, the number of sleep stages is not limited to 4, and may be 3 or less or 5 or more. For example, the number of sleep stages may be 5 by dividing non-REM sleep into stages N1 to N4 in order of lightness, or the number of stages may be 6 by adding stage W, which indicates wakefulness.

[0077] In this way, by using the feature amount of the respiratory waveform during sleep and the estimation result regarding the sleep stage, the sleep state determination system 1 can estimate the sleep stage with high accuracy.

[0078] (2-2-2d) Second Determination Unit The second determination unit 124 further determines whether the sleep stage transition is of good quality based on a third comparison result obtained by comparing the estimation history with third reference information. Specifically, the second determination unit 124 determines that the sleep stage transition is of good quality when the third comparison result satisfies a third condition. The third condition is, for example, that the similarity between the estimation history and the third reference information does not satisfy a third reference value.

[0079] The second determination unit 124 further determines whether the sleep state is continuously abnormal based on one or more of the following comparison results: a fourth comparison result obtained by comparing the first extraction history with fourth reference information, a fifth comparison result obtained by comparing the detection history with fifth reference information, and a sixth comparison result obtained by comparing the second extraction history with sixth reference information. The sixth reference information is reference information related to the second extraction history, i.e., the extraction history of body movement information.

[0080] In this way, by using the estimation history, the second determination unit 124 can determine whether the transition between sleep stages is of good quality.

[0081] (2-2-2e) Environmental Control Unit The environmental control unit 127 controls the environment of the space where the user H1 is sleeping, based on the determination result of the first determination unit 123. The space is, for example, the space of the room R1 as shown in FIG. 1 , but is not limited to the space of the room R1.

[0082] The environment includes, for example, the state of the air filling the space, the brightness in the space, and the sound generated in or propagating through the space. The state of the air includes, for example, the temperature, humidity, airflow, and cleanliness. The sound is, for example, background music, but may also be environmental sounds such as the sound of waves.

[0083] Examples of environmental control include air condition control, lighting control, and sound control. Examples of air condition control include temperature control, humidity control, airflow control, and air purification.

[0084] For example, when it is determined that the sleeping state is abnormal, the environment control unit 127 controls the environment via the spatial device 200. When it is determined that the sleeping state is not abnormal, the environment control unit 127 does not control the environment.

[0085] Alternatively, the environment control unit 127 may control the environment if the first comparison result satisfies the first condition, and may not control the environment if the first comparison result does not satisfy the first condition.

[0086] Alternatively, the environment control unit 127 may control the environment when, for example, the first determination unit 123 determines that the sleep state of the user H1 is an abnormal state caused by the environment (i.e., No in step S106 in FIG. 3 → step S108).If it is determined that the abnormal state is caused by something other than the environment (for example, a disease), the environment control unit 127 does not control the environment.

[0087] (2-2-2f) Notification Unit The notification unit 128 provides notification regarding the determination result of the first determination unit 123. In this embodiment, the notification unit 128 provides notification of the acquired determination result in response to the first determination unit 123 acquiring the determination result. In other words, whether the determination result indicates an abnormality or a normality, the notification unit 128 provides notification of the determination result. However, the notification unit 128 may provide notification only when the determination result indicates an abnormality, and may not provide notification when the determination result indicates a normality.

[0088] Alternatively, the notification unit 128 may issue a notification, for example, in response to a change in the sleep state to an abnormal state, that is, when there is a change in the determination result of the first determination unit 123. In this case, it is more preferable that the notification include, for example, information urging the user to visit a medical institution.

[0089] In this way, the notification unit 128 notifies the user H1 of the determination result of the first determination unit 123, and the sleep state determination system 1 allows the user H1 to recognize the sleep state. As a result, it becomes possible to provide support for appropriate measures against SAS and other diseases, for example.

[0090] (2-2-3) Spatial Device and Notification Device The spatial device 200 adjusts at least one of the temperature, humidity, air volume, air speed, air direction, brightness, color temperature, color deviation, etc. of the space in the room R1. The spatial device 200 is, for example, an air conditioner, an air quality device, a lighting fixture, or an audio device.

[0091] The notification device 300 notifies the user H1 by outputting at least one of video and audio. The notification device 300 is, for example, a communication device such as a smartphone or a tablet terminal, but may also be an AV device such as a television or radio connected to a network such as a LAN.

[0092] (2-2-4) Doppler Signal FIG. 5A shows a Doppler signal, which is an example of the sensing result of the radio wave sensor 10. In FIG. 5A, the horizontal axis corresponds to time and the vertical axis corresponds to amplitude. The Doppler signal includes multiple components, such as a component of movement based on breathing, a component of movement based on heartbeat, and a component of movement based on body movement. Note that the sensing result in FIG. 5A includes two Doppler signals (I signal and Q signal) indicated by thin and thick lines, but a single signal (only one of the two Doppler signals) may also be used.

[0093] (2-2-5) Body Movement Intensity A graph showing the change in body movement intensity over time is shown in FIG. 5B. The change in body movement intensity over time is expressed, for example, as time-series data in which the body movement intensity (integration results) at each time are arranged in time series. In FIG. 5B, the horizontal axis corresponds to time, and the vertical axis corresponds to body movement intensity. The body movement intensity is acquired based on the Doppler signal. The graph in FIG. 5B is the result of arranging the body movement intensity at each time in time series.

[0094] (2-2-6) Respiratory Waveform For example, a respiratory waveform such as that shown in Fig. 7 is extracted based on a Doppler signal such as that shown in Fig. 6. In Fig. 6, the horizontal axis corresponds to time and the vertical axis corresponds to the amplitude of the Doppler signal, while in Fig. 7, the horizontal axis corresponds to time and the vertical axis corresponds to the amplitude of respiratory movement. The amplitude of respiratory movement corresponds to the depth of breathing.

[0095] The first extraction unit 121 extracts a respiratory waveform of the person H1 based on the Doppler signal received by the reception unit 13 from the radio wave sensor 10. More specifically, the first extraction unit 121 includes a first trained model M1, which is a trained model generated by machine learning in an external learning device. The first trained model M1 receives the Doppler signal as input data and the respiratory waveform of the person H1 as output data. That is, the first trained model M1 receives the input Doppler signal and outputs the respiratory waveform of the person H1. The respiratory waveform output by the first trained model M1 is respiratory waveform data (estimated waveform data) estimated based on the Doppler signal.

[0096] (2-2-6a) Feature Amounts of Respiratory Waveform Feature amounts of respiratory waveforms include one or more types of feature amounts among feature amounts related to time and feature amounts related to amplitude. Various feature amounts used in this embodiment are shown in FIG. 8 . In FIG. 8 , the horizontal axis corresponds to time, and the vertical axis corresponds to the amplitude of respiratory movement. The feature amount related to time includes, for example, at least one of breath time (BT), inspiratory time (IT), expiratory time (ET), and post-expiratory pause time (PT). The feature amount related to amplitude includes at least one of an amount corresponding to inspiratory volume (IV), an amount corresponding to expiratory volume (EV), an amount corresponding to minute ventilatory volume, and respiratory rate.

[0097] As shown in Figure 8, the respiratory waveform has repeated peaks and valleys. The respiratory time BT is the time from when the lung volume (i.e., the amount of air in the lungs) reaches a minimum value VIST at a trough of the respiratory waveform to when the lung volume reaches a minimum value VIST+1 at the next trough and then begins to increase. The respiratory time BT is the sum of the inhalation time IT, the exhalation time ET, and the pause time PT.

[0098] The inspiration time IT is the time from when the lung volume reaches a minimum value VIST at the trough of the respiratory waveform to when the lung volume reaches a maximum value VIEO at the next peak.

[0099] The expiratory time ET is the time from when the lung volume reaches the maximum value VIEO at the peak of the respiratory waveform to when the lung volume reaches a value near the minimum value VIST+1 at the next valley. The value near the minimum value VIST+1 is the value of the lung volume at the point where a line L1 passing through the maximum point (peak) corresponding to "lung volume = VIEO" and an inflection point P1 on the decreasing curve from the maximum point (peak) to the minimum point (trough) corresponding to "lung volume = VIST+1" intersects with the decreasing curve, as shown in Figure 8, for example.

[0100] The pause time PT is the time during which the fluctuation of the lung volume is below a certain value at the trough of the respiratory waveform. After the pause time PT ends, the lung volume begins to increase.

[0101] The inspiration volume IV is the difference between the minimum value VIST of the lung volume at the trough of the respiratory waveform or a value near the minimum value VIST+1 and the maximum value VIEO of the lung volume at the next peak.

[0102] The expiratory volume EV is the difference between the maximum value VIEO of the lung volume at the next peak of the respiratory waveform and the minimum value VIST+1 of the lung volume at the next trough.

[0103] Minute ventilation is the total volume of gas ventilated in one minute. It is the product of the tidal volume (inspiration volume IV or expiration volume EV) and the number of breaths (number of breaths) per minute.

[0104] The feature quantity of the respiratory waveform may be the respiratory waveform itself, that is, the pattern or rhythm of the waveform curve.

[0105] (2-2-6b) Example of Determining Whether or Not a Disease Occurs Based on a Respiratory Waveform For example, a respiratory waveform such as that shown in Fig. 10 is extracted based on a Doppler signal such as that shown in Fig. 9. In Fig. 9, the horizontal axis corresponds to time, and the vertical axis corresponds to the amplitude of the Doppler signal, while in Fig. 10, the horizontal axis corresponds to time, and the vertical axis corresponds to the amplitude of respiratory movement.

[0106] The first extraction unit 121 extracts the respiratory waveform of Fig. 10 based on the Doppler signal of Fig. 9. More specifically, the first trained model M1 receives the Doppler signal of Fig. 9 as input and outputs the respiratory waveform of Fig. 10.

[0107] The waveform in Figure 10 shows the waveform pattern of Cheyne-Stokes respiration, a type of abnormal respiration specific to congestive heart failure or central nervous system disease. As shown in Figure 10, the Cheyne-Stokes respiration waveform is characterized by a repeated change pattern in which the depth and rate of breathing gradually increase, then gradually decrease, and then an apneic state occurs.

[0108] 10 are associated with a determination result indicating a sleep abnormality that may be a cause of a disease and are stored in advance in the reference storage unit 111. The first determination unit 123 obtains a determination result indicating a sleep abnormality that may be a cause of a disease when the similarity between the feature of the respiratory waveform extracted by the first extraction unit 121 and the feature stored in the reference storage unit 111 is equal to or greater than a threshold.

[0109] (3) Operational Example Next, the operation of the sleep state determination system 1 will be described with reference to Figures 1 to 4. Note that the following description is an example, and the method of classifying abnormal states shown in Figure 2 and the execution order of each step shown in Figures 3 and 4 may be changed as appropriate. Furthermore, in the following, descriptions of previously mentioned matters will be omitted or simplified.

[0110] Of the components of the sleep state determination system 1, the reference storage unit 111, history storage unit 112, first extraction unit 121, detection unit 122, first determination unit 123, second determination unit 124, second extraction unit 125, and estimation unit 126 operate according to the signal flow diagram of Fig. 2. In the signal flow diagram, thick-lined blocks indicate components of the sleep state determination system 1, and thin-lined blocks indicate information processed by the sleep state determination system 1. Note that the operation of Fig. 2 starts when the sleep state determination system 1 is started and ends when the operation is stopped.

[0111] 1 and 2, various types of reference information are stored in advance in a reference storage unit 111 included in the sleep state determination system 1. The various types of reference information include, for example, first to seventh reference information.

[0112] When the sleep state determination system 1 is activated, first, the reception unit 13 receives the sensing results of the radio wave sensor 10. As shown in Fig. 2, the received sensing results are input to each of the first extraction unit 121, the detection unit 122, and the second extraction unit 125. Each of the first extraction unit 121, the detection unit 122, and the second extraction unit 125 performs the following processing using the input sensing results.

[0113] The processing of the first extraction unit 121 (hereinafter referred to as "first extraction processing"), the processing of the detection unit 122 (hereinafter referred to as "detection processing"), and the processing of the second extraction unit 125 (hereinafter referred to as "second extraction processing") are executed in parallel. Processing executed in parallel refers to parallel processing by the multiple processors when the sleep state determination system 1 has multiple processors, for example, but may also be time-shared processing by a single processor. The cycles at which the first extraction processing, the detection processing, and the second extraction processing are executed are different from one another.

[0114] (3-1) First Extraction Process The first extraction unit 121 executes a first extraction process, which is a process of extracting a respiratory waveform from the sensing result and acquiring feature quantities of the extracted respiratory waveform. More specifically, the first extraction unit 121 includes a first trained model M1 generated in advance.

[0115] (3-1-1) First Trained Model The first trained model M1 that realizes the first extraction unit 121 is a model that receives sensing results as input and outputs a respiratory waveform. The first extraction unit 121 acquires features from the respiratory waveform output by the first trained model M1.

[0116] The history storage unit 112 stores the first extraction history including the features acquired by the first extraction unit 121 in association with time information indicating the current time (hereinafter simply referred to as "time information") acquired from the processor's built-in clock or an NTP (Network Time Protocol) server, etc.

[0117] (3-2) Detection Process The detection unit 122 executes the detection process, which is a process of detecting apnea events based on the sensing results and acquiring the frequency of occurrence of apnea events. More specifically, the detection unit 122 includes a second trained model M2 that has been generated in advance through machine learning by an external learner.

[0118] (3-2-1) Second Trained Model The second trained model M2 uses a Doppler signal as input data and time-series data related to the occurrence of apnea events of the person H1 as output data. The time-series data related to the occurrence of apnea events is, for example, time-series data that indicates a first value (e.g., "0") when the person H1 is breathing (i.e., a period when no apnea events are occurring) and a second value (e.g., "1") when the person H1 is not breathing (i.e., a period when an apnea event is occurring).

[0119] That is, the second trained model M2 receives the input of the Doppler signal and outputs, for example, "0" during periods when no apnea events occur and "1" during periods when apnea events occur. However, the second trained model M2 may receive the input of the Doppler signal and output data indicating that the person H1 is in an apneic state during periods when apnea events occur, but may not output any data other than during periods when apnea events occur.

[0120] The time-series data regarding the occurrence of apnea events is preferably data indicating the frequency of occurrence of apnea events. The frequency of occurrence of apnea events may be, for example, second data obtained by calculating the average value of the first data per unit time (e.g., one hour) at each time point based on first data that indicates "0" during periods when no apnea events occur and "1" during periods when an apnea event occurs, and arranging the calculation results in time series.

[0121] The history storage unit 112 stores the detection history including the occurrence frequency acquired by the detection unit 122 in association with time information.

[0122] (3-3) Second Extraction Process (Body Movement Extraction Process) The second extraction unit 125 repeatedly performs the following second extraction process based on the sensing results to acquire body movement information related to the user's body movements. Note that the second extraction process may also be referred to as body movement extraction process.

[0123] The second extraction unit 125 performs FFT on the sensing result to obtain a power spectrum, then integrates the obtained power spectrum over the frequency range of the evaluation target, and sets the integration result as the body movement intensity.

[0124] The history storage unit 112 stores the acquired body movement information in association with time information.

[0125] (3-4) First Determination Process The feature amount acquired by the first extraction unit 121 and the occurrence frequency acquired by the detection unit 122 are input to the first determination unit 123 as shown in FIG.

[0126] The first determination unit 123 executes a first determination process using the feature amount input from the first extraction unit 121 and the occurrence frequency input from the alignment detection unit 122 .

[0127] The first determination process is executed, for example, according to the flowchart of Fig. 3. The process of Fig. 3 is executed every time a set of a respiratory waveform, a feature amount, and an occurrence frequency is input.

[0128] First, the first judgment unit 123 compares the feature acquired by the first extraction unit with the first reference information stored in the reference memory unit 111, and the occurrence frequency acquired by the detection unit 122 with the second reference information stored in the reference memory unit 111 (step S101).

[0129] Next, the first determination unit 123 determines whether at least one of the first condition and the second condition is satisfied (step S102). If it is determined that neither the first condition nor the second condition is satisfied (No in step S102), the first determination unit 123 determines that the sleeping state of the user H1 is normal (step S103). Then, the process ends.

[0130] If it is determined in step S102 that at least one of the first condition and the second condition is satisfied (Yes), the first determination unit 123 further determines whether the second condition is satisfied (step S104).If it is determined that the second condition is not satisfied (No), the process proceeds to step S106 (described below).

[0131] If it is determined in step S104 that the second condition is satisfied (Yes), the first determination unit 123 determines that the sleep state of the user H1 is an abnormal state caused by SAS (i.e., SAS has been detected) (step S105).Then, the process ends.

[0132] In step S106, the first determination unit 123 determines whether the disease condition is satisfied or whether the environmental condition is not satisfied. If it is determined that the disease condition is satisfied or the environmental condition is not satisfied (Yes in step S106), the first determination unit 123 determines that the sleep state of the user H1 is an abnormal state caused by a disease (i.e., a disease has been detected) (step S107). Then, the processing ends.

[0133] If it is determined in step S106 that the disease condition is not satisfied or the environmental condition is satisfied (No), the first determination unit 123 determines that the sleep state of the user H1 is an abnormal state caused by the environment (step S108).

[0134] (3-5) First notification process The judgment result of the first judgment process, that is, the judgment result regarding whether the user's sleep state is abnormal or not, whether SAS has been detected or not, and whether the cause of the abnormality is a disease or the environment, is input to the notification device 300 shown in Figure 1.

[0135] The notification unit 128 executes a first notification process for notifying, via the notification device 300, the determination result of the first determination process input from the first determination unit 123.

[0136] (3-6) Environmental Control Processing The determination result of the first determination processing is also input to the notification device 300 shown in Fig. 1. The environmental control unit 127 executes environmental control processing, which is processing for performing environmental control based on the determination result of the first determination processing input from the first determination unit 123.

[0137] (3-7) Estimation Processing The features acquired by the first extraction unit 121 (result of the first extraction processing) and the body movement information acquired by the second extraction unit 125 (result of the second extraction processing) are input to the estimation unit 126, as shown in Figure 2.

[0138] The estimation unit 126 executes an estimation process to estimate a sleep stage based on the input feature amount and body movement information.

[0139] The history storage unit 112 stores the estimation history including the sleep stages acquired by the estimation unit 126 in association with time information.

[0140] (3-8) Second Determination Process The processing unit 12 repeatedly executes a determination process to determine whether the current time is the timing to execute the second determination process based on time information acquired from a processor or the like and timing information stored in the storage unit 11. Note that the determination process may be executed by the second determination unit 124. If it is determined that the current time is the timing to execute the second determination process, the second determination unit 124 executes the second determination process using various histories stored in the history storage unit 112. The various histories are history (first extraction history) regarding the results of the first extraction process (output of the first trained model M1), history (detection history) regarding the results of the detection process (output of the second trained model M2), and history (estimation history) regarding the results of the estimation process.

[0141] The second determination process is executed, for example, according to the flowchart of Fig. 4. The process of the flowchart of Fig. 4 is executed repeatedly (for example, once a week) at a period based on the timing information.

[0142] First, the second determination unit 124 compares the estimated history stored in the history storage unit 112 with the third reference information stored in the reference storage unit 111 (step S201). Then, the second determination unit 124 determines whether the third condition is satisfied based on the comparison result in step S201 (step S202). If the second determination unit 124 determines that the third condition is not satisfied (No in step S202), it determines that the sleep stage transition is of good quality (step S203).

[0143] If it is determined in step S202 that the third condition is satisfied (Yes), the second determination unit 124 determines that the transition between sleep stages is of good quality (step S204).

[0144] Next, the second determination unit 124 determines whether the abnormal state continues based on the detection history stored in the history storage unit 112 and the first and second extraction histories stored in the history storage unit 112 (step S205). Then, the process ends.

[0145] (3-9) Second Notification Process The notification unit 128 executes a second notification process for notifying the determination result of the second determination process executed by the second determination unit 124.

[0146] (4) Method for Generating Trained Models The first trained model and the second trained models M1 and M2 are generated in advance by machine learning using an external learner. The machine learning used in this embodiment is, for example, a regression neural network suitable for handling time-series data. However, the machine learning may be a neural network other than a regression neural network, or may be various machine learning algorithms (such as deep learning) that apply a neural network.

[0147] (4-1) Method for generating a first trained model For a large number of subjects, assuming a variety of users, sensing is performed in parallel using a radio wave sensor equivalent to the radio wave sensor 10 described above and a contact-type respiratory sensor capable of sensing respiratory waveforms (for example, a pressure sensor wrapped around the abdomen of the subject). The sensing results from the radio wave sensor and the sensing results from the contact-type respiratory sensor correspond to a common time axis.

[0148] A set of sensing results (Doppler signals) collected from a large number of subjects by a radio wave sensor and waveform data (measured waveform data) of respiratory waveforms actually measured by a contact respiratory sensor are repeatedly input to a learning device as training data. The learning device generates a first trained model M1 by repeatedly executing a machine learning learning process using the input training data. The first trained model M1 thus generated in advance by an external learning device is stored in the memory of the sleep state determination system 1, thereby realizing the function of the first extraction unit 121.

[0149] (4-2) Method for Generating the Second Trained Model The subject at the time of generating the second trained model is a patient with symptoms corresponding to sleep apnea syndrome. Sensing using a radio wave sensor equivalent to the radio wave sensor 10 described above and a PSG test are performed in parallel on the subject while the subject is sleeping in a hospital room. When a medical professional such as a doctor or technician determines that the subject is in an apneic state while observing the PSG test results, the medical professional performs an operation to set a flag indicating the occurrence period of an apneic event on a common time axis. In response to this operation, a flag indicating the occurrence period of an apneic event is set on the time axis.

[0150] A set of the sensing result of the radio wave sensor and flag information related to one or more flags based on the judgment of a medical professional is input to the learning device as training data, and the learning device generates the second trained model M2 by executing a machine learning learning process using the input training data. The second trained model M2 thus generated in advance by the external learning device is stored in the memory of the sleep state determination system 1, thereby realizing the function of the second extraction unit 125.

[0151] (5) Modification of Sleep State Determination System In this modification, the explanation of the previously mentioned matters will be omitted, and the differences from the embodiment will be explained in detail.

[0152] In the sleep state determination system 1 of this modification, the processing unit 12 in the sleep state determination system 1 of the embodiment (see FIG. 1) further includes a third extraction unit 129, as shown in FIG.

[0153] The third extraction unit 129 includes a third trained model (M3) that receives the sensing results of the radio wave sensor 10 as input and outputs a heartbeat interval or a heart rate. The third trained model M3 is a trained model that has been trained in advance by machine learning using an external learning device, using as training data the results of sensing the body surface movements of a large number of subjects, representing a variety of users, using the radio wave sensor 10 while they sleep, and the heartbeat intervals obtained at the same time by a contact-type heart rate meter that can detect the heartbeat intervals. The third extraction unit 129 may also be referred to as the "heartbeat extraction unit 129."

[0154] The history storage unit 112 further stores a third extraction history. The third extraction history is an extraction history of the third extraction unit 129. The third extraction history is a history related to the heart rate, and includes, for example, heart rate fluctuations over a predetermined period ΔT and a set of pairs of heart rate and time information.

[0155] The first determination unit 123 determines whether the sleep state is abnormal or not, further based on the heart rate acquired by the third extraction unit 129 .

[0156] In this way, the third extraction unit 129 acquires the heart rate, and the first judgment unit 123 takes into account the heart rate or heart rate fluctuations when judging whether the sleep state is abnormal or not, thereby enabling the sleep state judgment system 1 to further improve the judgment accuracy.

[0157] The acquired heart rate or heart rate variability is also used by the estimation unit 126 to estimate the sleep stage.

[0158] 12 , the sensing results are input to the first extraction unit 121, the detection unit 122, and the second extraction unit 125, and further to the third extraction unit 129. The third extraction unit 129 extracts the heartbeat interval from the sensing results and acquires the heart rate. The history storage unit 112 stores the heart rate acquired by the third extraction unit 129 in association with time information.

[0159] In this modification, the first determination process (see FIG. 3 ) is performed in consideration of the heart rate acquired by the third extraction unit 129. That is, the first determination unit 123 determines whether the sleep state is normal or not based on the heart rate or heart rate fluctuations from the third extraction unit 129 in addition to the feature amount from the first extraction unit 121 and the occurrence frequency from the detection unit 122.

[0160] More specifically, the reference storage unit 111 further stores seventh reference information. The seventh reference information is reference information related to the heart rate or heart rate variability. In step S101 shown in FIG. 3 , the first determination unit 123 further compares the heart rate with the seventh reference information and determines whether the comparison result satisfies a fourth condition. The fourth condition is, for example, that the similarity between the heart rate or heart rate variability and the fourth reference information does not satisfy a fourth reference value.

[0161] In step S102, the first determination unit 123 determines whether at least one of the first, second, and fourth conditions is satisfied. If it is determined that none of the first, second, and fourth conditions is satisfied (No in step S102), the process proceeds to step S103. If it is determined that at least one of the first, second, and fourth conditions is satisfied (Yes in step S102), the process proceeds to step S104.

[0162] The heart rate or heart rate variability may also be taken into consideration in the second determination process (see FIG. 4). That is, the heart rate or heart rate variability may also be used to estimate the sleep stage.

[0163] According to this modification, the first determination unit 123 takes the heart rate or heart rate fluctuation into consideration when making a determination, thereby further improving the accuracy of determining whether the sleep state is abnormal. Also, the second determination unit 124 takes the heart rate into consideration when making a determination, thereby further improving the accuracy of determining whether the sleep stage transition is good.

[0164] (6) Other Modifications (6-1) Modifications Related to the Sensor The sensor in the embodiment is the Doppler-type radio wave sensor 10, and the movement detected by this is the speed of movement of the body surface. However, the sensor in this modification may be, for example, a distance measurement sensor such as a laser distance measurement sensor, or a displacement sensor. This may detect a change in the position of the body surface (displacement: for example, a change in the distance from the radio wave sensor 10 to the body surface), and the speed of the body surface may be calculated from the detected displacement.

[0165] Alternatively, the sensor may be an acceleration sensor attached to the body surface, which detects the acceleration of the body surface, and the velocity of the body surface may be calculated from the detected acceleration.

[0166] Alternatively, the sensor may be a pressure sensor attached to the upper surface of the bed B1 (for example, a mattress), which detects changes in pressure, and the velocity of the body surface may be calculated from the detected changes in pressure.

[0167] (6-2) Modified Examples of Sensing Target The body movement that is the sensing target is not limited to the movement of the body surface, but may be, for example, the movement of the center of gravity of the body, etc. This type of movement can be acquired, for example, by an acceleration sensor.

[0168] (6-3) Modified Examples of Machine Learning Machine learning may be machine learning other than neural networks. Examples of machine learning in modified examples include decision trees, random forests, and support vector machines (SVMs), but the type is not limited thereto.

[0169] (6-4) Modifications of the Storage Unit The storage unit 11 may be located outside the sleep state determination system 1. The storage unit 11 may be, for example, a storage device connected to a LAN or a cloud server connected to the Internet that is separate from the sleep state determination system 1. The location of the storage unit 11 does not matter as long as it is accessible by the processor of the sleep state determination system 1.

[0170] (6-5) Modifications of the First Extraction Unit and the Detection Unit The first extraction unit 121 may be realized by an algorithm other than machine learning.

[0171] The detection unit 122 may be realized by an algorithm other than machine learning.

[0172] The third extraction unit 129 may be realized by an algorithm other than machine learning.

[0173] (7) Summary The sleep state determination system (1) according to the first aspect includes an extraction unit (first extraction unit 121), a detection unit (122), and a determination unit (123). The extraction unit (first extraction unit 121) extracts a respiratory waveform based on the breathing of the user (H1) from a sensing result obtained by sensing the body movement of the user (H1) while sleeping using a sensor (10), and acquires feature quantities of the respiratory waveform. The detection unit (122) detects the occurrence of an apnea event, in which the user (H1) stops breathing for a certain period of time or longer, based on the sensing result, and acquires the frequency of occurrence of the apnea event. The determination unit (123) determines whether the sleep state of the user (H1) is abnormal based on the feature quantities acquired by the extraction unit (first extraction unit 121) and the frequency of occurrence acquired by the detection unit (122).

[0174] According to this aspect, the sleep state determination system (1) can determine the sleep state in more detail by using the feature amount of the respiratory waveform and the occurrence frequency of apnea events.

[0175] The sleep state determination system (1) according to the second aspect is the same as the first aspect, but further includes a reference storage unit (111). The reference storage unit (111) pre-stores first reference information related to feature amounts and second reference information related to occurrence frequencies. The determination unit (123) determines whether the sleep state is abnormal based on the first comparison result and the second comparison result. The first comparison result is a result of comparing the feature amounts with the first reference information. The second comparison result is a result of comparing the occurrence frequencies with the second reference information.

[0176] According to this aspect, the sleep state determination system (1) can determine the sleep state in more detail and with higher accuracy based on the first comparison result of the respiratory waveform feature with the first reference information and the second comparison result of the occurrence frequency of apnea events with the second reference information.

[0177] In the sleep state determination system (1) according to the third aspect, in the second aspect, the determination unit (123) determines that the sleep state is abnormal when the first comparison result satisfies a first condition or when the second comparison result satisfies a second condition.

[0178] According to this aspect, the sleep state determination system (1) can determine the sleep state in more detail and with higher accuracy.

[0179] In the sleep state determination system (1) according to the fourth aspect, in the third aspect, the determination unit (123) determines that sleep apnea syndrome (SAS) has been detected when the second comparison result satisfies the second condition.

[0180] According to this aspect, the sleep state determination system (1) can accurately determine whether or not SAS has been detected.

[0181] In a sleep state determination system (1) according to a fifth aspect, in the fourth aspect, the first reference information includes disease reference information. The disease reference information is information relating to one or more types of disease. The first condition includes a disease condition related to the disease reference information. The first comparison result includes a disease comparison result obtained by comparing the feature amount with the disease condition. If the first comparison result satisfies the first condition, the determination unit determines that the abnormal state is caused by a disease.

[0182] According to this aspect, when the sleep state determination system (1) determines that SAS has not been detected, it can further determine whether the abnormal state is caused by a disease (in other words, whether a disease has been detected).

[0183] In the sleep state determination system (1) according to the sixth aspect, in the fifth aspect, the first reference information further includes environmental standard information related to the environment of the space in which the user is sleeping. The first condition further includes environmental conditions related to the environmental standard information. The first comparison result further includes an environmental comparison result obtained by comparing the feature amount with the environmental conditions. The determination unit (123) determines that the abnormal condition is caused by a disease when the disease comparison result satisfies the disease condition or when the environmental comparison result does not satisfy the environmental condition. Furthermore, the determination unit (123) determines that the abnormal condition is caused by the environment when the disease comparison result does not satisfy the disease condition or when the environmental comparison result satisfies the environmental condition.

[0184] According to this aspect, when the sleep state determination system (1) determines that SAS has not been detected, it can further determine whether the SAS is caused by a disease or an environment.

[0185] In a sleep state determination system (1) according to a seventh aspect, in any of the third to sixth aspects, the first reference information includes feature data indicating feature amounts of the user's past respiratory waveform. The determination unit (123) determines that the user's sleep state is abnormal when the first comparison result satisfies a first condition or the difference between the feature amount and the feature data is equal to or greater than a predetermined value for a predetermined period of time or when the second comparison result satisfies a second condition.

[0186] According to this aspect, it is possible to accurately determine whether the sleep state of the user (H1) is abnormal or not.

[0187] In a sleep state determination system (1) according to an eighth aspect, in any of the second to seventh aspects, the extraction unit (first extraction unit 121) is the first extraction unit (121). The sleep state determination system (1) further includes a second extraction unit (125) and an estimation unit (126). The second extraction unit (125) acquires body movement information related to the body movement of the user (H1) based on the sensing result. The estimation unit (126) estimates which of a plurality of predetermined sleep stages the sleep state corresponds to based on the feature acquired by the first extraction unit (121) and the body movement information acquired by the second extraction unit (125).

[0188] According to this aspect, the sleep state determination system (1) can estimate the sleep stage with high accuracy by using the feature amount of the respiratory waveform and the body movement information related to the body movement.

[0189] In a sleep state determination system (1) according to a ninth aspect, in the eighth aspect, the determination unit (123) is a first determination unit (123). The sleep state determination system (1) further includes a history storage unit (112) and a second determination unit (124) different from the first determination unit (123). The history storage unit (112) stores the estimation history of the estimation unit (126). The reference storage unit (111) further stores third reference information related to the estimation history in advance. The second determination unit (124) determines whether the sleep stage transition is a good transition or not based on a third comparison result obtained by comparing the estimation history with the third reference information.

[0190] According to this aspect, by using the estimation history, the sleep state determination system (1) can determine whether the transition between sleep stages is of good quality.

[0191] A sleep state determination system (1) according to a tenth aspect is the sleep state determination system (1) of any one of the first to ninth aspects, further including a third extraction unit (129). The third extraction unit (129) acquires the heart rate of the user (H1) from the sensing result. The determination unit (123) determines whether the sleep state is abnormal based on the feature amount, the occurrence frequency, and the heart rate.

[0192] In the tenth aspect, the third extraction unit (129) may extract a heartbeat interval based on the heartbeat of the user (H1) from the sensing result and acquire the heart rate based on the heartbeat interval. Furthermore, the determination unit (123) may determine whether the sleep state is abnormal based on the feature amount, the occurrence frequency, and the heart rate or heart rate variability.

[0193] According to this aspect, by taking the heart rate into consideration when determining whether the sleep state is abnormal, the sleep state determination system (1) can further improve the determination accuracy.

[0194] Furthermore, when the estimation unit (126) estimates the sleep stage, it may use heart rate or heart rate variability information in addition to the feature acquired by the first extraction unit (121) and the body movement information acquired by the second extraction unit (125), thereby enabling accurate estimation of the sleep stage.

[0195] The sleep state determination system (1) according to an eleventh aspect is any one of the first to tenth aspects, and further includes an environment control unit (127). The environment control unit (127) controls the environment of the space in which the user (H1) is sleeping, based on the determination result of the determination unit (123).

[0196] According to this aspect, the sleep state determination system (1) can perform appropriate environmental control by using the determination result.

[0197] The sleep state determination system (1) according to a twelfth aspect is the sleep state determination system (1) of any one of the first to eleventh aspects, further comprising a notification unit (128). The notification unit (128) notifies the user of the determination result of the determination unit (123).

[0198] According to this aspect, by notifying the user (H1) of the result of the determination, the sleep state determination system (1) can allow the user (H1) to recognize the sleep state, thereby providing support for appropriate treatment of, for example, sleep apnea syndrome (SAS) and other diseases.

[0199] In a sleep state determination system (1) according to a thirteenth aspect, in any one of the first to twelfth aspects, the extraction unit (first extraction unit 121) includes a first trained model (M1) of machine learning that receives sensing results as input and outputs a respiratory waveform. The detection unit (122) includes a second trained model (M2) of machine learning that receives sensing results as input and outputs time-series data related to the occurrence of apnea events.

[0200] According to this aspect, the sleep state determination system (1) can determine the sleep state in more detail and with higher accuracy by using a trained model of machine learning.

[0201] A sensing system (100) according to a fourteenth aspect includes the sleep state determination system (1) according to any one of the first to thirteenth aspects and a sensor (10).

[0202] According to this aspect, the sleep state determination system (1) can determine the sleep state in more detail by using the feature amount of the respiratory waveform and the occurrence frequency of apnea events.

[0203] In a sensing system (100) according to a fifteenth aspect, in the fourteenth aspect, the sensor (10) is a radio wave sensor (10). The radio wave sensor (10) senses movement of the body surface of the user (H1) using radio waves.

[0204] According to this aspect, by using the radio wave sensor (10), the sleep state determination system (1) can make a contactless and accurate determination regarding the sleep state.

[0205] REFERENCE SIGNS LIST 100 Sensing system 10 Sensor (radio wave sensor) 1 Sleep state determination system 111 Reference memory unit 112 History memory unit 121 Extraction unit (first extraction unit) 122 Detection unit 123 First determination unit 124 Second determination unit 125 Second extraction unit 126 Estimation unit 127 Environment control unit 128 Notification unit 129 Third extraction unit H1 User M1 First trained model M2 Second trained model

Claims

1. A sleep state determination system comprising: an extraction unit that extracts a respiratory waveform based on the user's breathing from the sensing results obtained by sensing the body movements of a user while sleeping using a sensor, and acquires feature quantities of the respiratory waveform; a detection unit that detects the occurrence of an apnea event, in which the user's breathing stops for a certain period of time or longer, based on the sensing results, and acquires the frequency of occurrence of the apnea event; and a determination unit that determines whether the user's sleep state is abnormal based on the feature quantities acquired by the extraction unit and the frequency of occurrence acquired by the detection unit.

2. The sleep state determination system of claim 1, further comprising a reference memory unit that pre-stores first reference information regarding the feature amount and second reference information regarding the occurrence frequency, and the determination unit determines whether the sleep state is the abnormal state based on a first comparison result obtained by comparing the feature amount with the first reference information and a second comparison result obtained by comparing the occurrence frequency with the second reference information.

3. The sleep state determination system according to claim 2, wherein the determination unit determines that the sleep state is the abnormal state when the first comparison result satisfies a first condition or when the second comparison result satisfies a second condition.

4. The sleep state determination system according to claim 3, wherein the determination unit determines that sleep apnea syndrome has been detected when the second comparison result satisfies the second condition.

5. The sleep state determination system of claim 4, wherein the first reference information includes disease reference information relating to one or more types of disease, the first condition includes a disease condition relating to the disease reference information, the first comparison result includes a disease comparison result obtained by comparing the feature amount with the disease condition, and the determination unit determines that the abnormal state is caused by a disease when the first comparison result satisfies the first condition.

6. The sleep state determination system of claim 5, wherein the first reference information further includes environmental standard information regarding the environment of the space in which the user is sleeping, the first conditions further include environmental conditions regarding the environmental standard information, the first comparison result further includes an environmental comparison result in which the feature is compared with the environmental conditions, and the determination unit determines that the abnormal state is caused by a disease if the disease comparison result satisfies the disease condition or if the environmental comparison result does not satisfy the environmental condition, and determines that the abnormal state is caused by the environment if the disease comparison result does not satisfy the disease condition or if the environmental comparison result satisfies the environmental condition.

7. The sleep state determination system of claim 3, wherein the first reference information includes feature data indicating feature amounts of the user's past respiratory waveform, and the determination unit determines that the user's sleep state is abnormal if the first comparison result satisfies the first condition, or if the difference between the feature amount and the feature data is equal to or greater than a predetermined value, continues for a predetermined time or longer, or if the second comparison result satisfies the second condition.

8. A sleep state determination system as described in any one of claims 2 to 7, wherein the extraction unit is a first extraction unit, and further comprising: a second extraction unit that acquires body movement information regarding the body movement of the user based on the sensing results; and an estimation unit that estimates which of a plurality of predetermined sleep stages the sleep state corresponds to based on the feature acquired by the first extraction unit and the body movement information acquired by the second extraction unit.

9. The sleep state determination system of claim 8, wherein the determination unit is a first determination unit, and further comprises: a history storage unit that stores the estimation history of the estimation unit; and a second determination unit different from the first determination unit, wherein the reference storage unit further stores in advance third reference information regarding the estimation history, and the second determination unit determines whether the transition of the sleep stage is a good transition or not based on a third comparison result obtained by comparing the estimation history with the third reference information.

10. A sleep state determination system as described in any one of claims 1 to 7, further comprising a third extraction unit that acquires the user's heart rate from the sensing results, and the determination unit determines whether the sleep state is the abnormal state based on the feature amount, the occurrence frequency, and the heart rate.

11. A sleep state determination system according to any one of claims 1 to 7, further comprising an environmental control unit that controls the environment of the space in which the user is sleeping based on the determination result of the determination unit.

12. The sleep state determination system according to any one of claims 1 to 7, further comprising a notification unit that notifies the user of the determination result of the determination unit.

13. A sleep state determination system as described in any one of claims 1 to 7, wherein the extraction unit includes a first trained machine learning model that receives the sensing result as an input and outputs the respiratory waveform, and the detection unit includes a second trained machine learning model that receives the sensing result as an input and outputs time series data regarding the occurrence of the apnea event.

14. A sensing system comprising: a sleep state determination system according to any one of claims 1 to 7; and the sensor.

15. The sensing system according to claim 14, wherein the sensor is a radio wave sensor that senses the movement of the user's body surface using radio waves.

Images