The present invention relates to a sleep-wakefulness determination apparatus configured to determine the sleep and wakefulness of a user, and a program.
It is known that sleep disorders such as insomnia, sleep-disordered breathing, and hypersomnia can be detrimental to health. In order to grasp the sleeping condition of a person, it is necessary to examine the actual condition of how the person actually sleeps, from one night to several days.
The all-night polysomnography (PSG) test, proposed in Patent Application Publication JP2013-99507 and the like, has been developed as a test to examine a person's sleep state. In PSG, a large number of electrodes and sensors are attached to the body of a subject, and each electrode and sensor is connected to a special measurement device to measure basic data such as electroencephalogram, electrocardiogram, electromyogram, and respiratory status, and then the state of sleep and wakefulness is examined based on the basic data.
A sleep test using PSG, such as that described in Patent Application Publication JP2013-99507, requires a large number of measurement devices and the location thereof is limited to hospitals and laboratories. This makes it difficult for many people to easily perform the test for longer than a few days. In addition, wearing a lot of electrodes and sensors on the body in an environment different from home causes stress and makes it difficult to sleep. As a result, it is difficult to test the normal sleep state correctly.
In light of the above circumstances, the present invention provides a sleep-wakefulness determination apparatus and program, which determine sleep-wakefulness with a sufficiently high level of accuracy using a small number of wearing devices.
According to one aspect of the present invention, there is provided a sleep-wakefulness determination apparatus configured to determine sleep and wakefulness of a user. The present sleep-wakefulness determination apparatus comprises a scalar calculation unit, a feature value calculation unit, and a sleep-wakefulness determination unit. The scalar calculation unit is configured to calculate scalar value based on each component of the acceleration vector in a part of a body of the user. The feature value calculation unit is configured to calculate feature value for each epoch defined by a predetermined time based on the scalar value. The sleep-wakefulness determination unit is configured to determine the sleep and wakefulness of the user based on the feature value of a desired epoch and the feature value of peripheral epochs included in a plurality of epochs before and after the desired epoch in a time series.
With the sleep-wakefulness determination apparatus, it is possible to determine sleep and wakefulness by the subject to be examined wearing only a bio-acceleration measuring device.
Hereinafter, embodiments of the present invention will be described with reference to the drawings. Various features described in the embodiment below can be combined with each other.
A program for realizing a software in the present embodiment may be provided as a non-transitory computer readable medium that can be read by a computer, or may be provided for download from an external server, or may be provided so that the program can be activated on an external computer to realize functions thereof on a client terminal (so-called cloud computing).
In the present embodiment, the “unit” may include, for instance, a combination of hardware resources implemented by circuits in a broad sense and information processing of software that can be concretely realized by these hardware resources. Further, although various information is adopted in the present embodiment, this information can be represented, for example, by physical signal values representing voltage and current, by high and low signal values as a bit set of binary numbers composed of 0 or 1, or by quantum superposition (so-called quantum bit). In this way, communication/operation can be executed on a circuit in a broad sense.
Further, the circuit in a broad sense is a circuit realized by combining at least an appropriate number of a circuit, a circuitry, a processor, a memory, and the like. In other words, it is a circuit includes Application Specific Integrated Circuit (ASIC), Programmable Logic Apparatus (e.g., Simple Programmable Logic Apparatus (SPLD), Complex Programmable Logic Apparatus (CPLD), and Field Programmable Gate Array (FPGA)), and the like.
In Section 1, the overall configuration of a sleep-wakefulness determination system 1 will be described.
As shown in
Although wired communication means such as USB, IEEE1394, Thunderbolt, and wired LAN network communication are preferable, the communication unit 21 may include wireless LAN network communication, mobile communication such as 3G/LTE/5G, Bluetooth (registered trademark) communication or the like as necessary. In particular, in the present embodiment, the communication section 21 is preferably configured to write information including time-series three-dimensional (3-D) acceleration vectors v(x, y, z) measured by the acceleration sensor 23 described below to external storage media M. The type and form of the storage media M are not particularly limited, and for example, flash memory, card-type memory, optical disk, etc. may be employed as appropriate.
The storage unit 22 stores various information defined by the aforementioned description. The storage unit 22 can be implemented, for instance, as a storage device such as a solid state drive (SSD), or as a memory such as a random access memory (RAM) that stores temporarily necessary information (argument, array, or the like) regarding program operation, or any combination thereof. In particular, the storage unit 22 can store information including time-series 3-D acceleration vectors v(x, y, z) measured by the acceleration sensor 23 described below. It may be implemented to store the information directly in the aforementioned storage media M without going through the storage unit 22.
The acceleration sensor 23 is configured to measure the acceleration of a part of a body (e.g., an arm) of the user U as 3-D vector information. In other words, information including time-series 3-D acceleration vectors v(x, y, z) can be acquired from the user U.
As shown in
Although wired communication means such as USB, IEEE1394, Thunderbolt, and wired LAN network communication are preferable, the communication unit 21 may include wireless LAN network communication, mobile communication such as 3G/LTE/5G, Bluetooth (registered trademark) communication or the like as necessary. In particular, in the present embodiment, it is preferable to implement the communication unit 31 as a storage media reading unit configured to read information stored in external storage media M. The storage media M stores information including time-series 3-D acceleration vectors v(x, y, z) acquired from the user U by the wearable device 2. As a result, the communication unit 31, which is a storage media reading unit, can read the 3-D acceleration vector v(x, y, z) stored in the storage media M.
The storage unit 32 stores various information defined by the aforementioned description. The storage unit 32 can be implemented, for instance, as a storage device such as a solid state drive (SSD), or as a memory such as a random access memory (RAM) that stores temporarily necessary information (argument, array, or the like) regarding program operation, or any combination thereof.
In particular, the storage unit 32 stores a scalar value calculation program for calculating a scalar value a based on each component (x, y, z) of the 3-D acceleration vector v(x, y, z) at a part of the body of the user U. The storage unit 32 also stores a feature value calculation program for calculating a feature value f(N) for each epoch defined by a predetermined time based on the scalar value a. Further, the storage unit 32 stores a sleep-wakefulness determination program for determining the sleep and wakefulness of the user U based on the feature value f(N) of the desired epoch and the feature value f(N±δ) of peripheral epochs included in the plurality of epochs before and after the desired epoch in a time series. Further, the storage unit 32 stores various programs with respect to the sleep-wakefulness determination apparatus 3 executed by the controller 33, etc. in addition to the above.
Further, the storage unit 32 stores a machine learning model allowed to learn correlation of the feature value f(N) of the desired epoch, the feature value f(N±δ) of the peripheral epoch and the sleep and wakefulness of the user U. Preferably, conventional algorithms can be employed for the algorithm for such machine learning as appropriate. For example, logistic regression, random forest, XGBoost, multilayer perceptron (MLP), or the like can be adopted. In addition, each time the sleep-wakefulness determination apparatus 3 is used, machine learning using the results thereof as training data can be further performed to update such machine learning model.
The controller 33 processes and controls overall operation regarding the sleep-wakefulness determination apparatus 3. The controller 33 is implemented as, for instance, an unshown central processing unit (CPU). The controller 33 realizes various functions with respect to the sleep-wakefulness determination apparatus 3 by reading out a predetermined program stored in the storage unit 32. Specifically, the scalar value calculation function for calculating a scalar value a based on each component (x, y, z) of the 3-D acceleration vector v(x, y, z) in a part of the body of the user U, the feature value calculation function for calculating the feature value f(N) for each epoch defined by a predetermined time based on the scalar value a, the sleep and wakefulness determination function for determining the sleep and wakefulness of the user U based on the feature value f(N) of the desired epoch and the feature value f(N±δ) of peripheral epochs included in the plurality of epochs before and after the desired epoch in a time series, in such epochs or the like are included.
In other words, the information processing by software (stored in the storage unit 32) is specifically realized by hardware (controller 33), in such a manner that the controller 33 may be executed as a scalar value calculation unit 331, a feature value calculation unit 332, and a sleep-wakefulness determination unit 333 as shown in
The scalar value calculation unit 331 is configured to execute the information processing by software (stored in the storage unit 32) specifically realized by hardware (controller 33). The scalar value calculation unit 331 calculates a scalar value a based on each component (x, y, z) of the 3-D acceleration vector v(x, y, z) at a part of the body of the user U. For example, the scalar value a is the L2 norm (so-called magnitude) of v(x, y, z). Of course, the scalar value a can also be the L1 norm.
Further, the scalar value calculation unit 331 may calculate the scalar value a (e.g. the L2 norm) based on each component of the time difference vector Δv(x, y, z) acquired from the 3-D acceleration vector v(x, y, z). Here, the time difference vector Δv(x, y, z) is a difference vector between two 3-D acceleration vectors v_1(x, y, z) and v_2 (x, y, z) in a time series. The two 3D acceleration vectors v(x, y, z) at adjacent times are more preferably employed. Due to this process, the effect of gravitational acceleration can be suppressed compared to the case in which the 3-D acceleration vector v(x, y, z) is used directly.
Alternatively, the scalar value calculation unit 331 calculates the scalar value a (e.g., L2 norm) based on each component of the n-th-order time derivative vector v{circumflex over ( )}n(x, y, z) acquired from the 3-D acceleration vector v(x, y, z). Here, the n-th-order time derivative vector v{circumflex over ( )}n(x, y, z) is the n-th-order time derivative of the 3-D acceleration vector where n is a natural number (n≥1). Due to this process, the effect of gravitational acceleration can be suppressed compared to the case in which the 3-D acceleration vector v(x, y, z) is used directly.
The values of n are, for example, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, and may be in the range between any two of the numbers indicated above.
The feature value calculation unit 332 is configured to execute the information processing by software (stored in the storage unit 32) specifically realized by hardware (controller 33). The feature value calculation unit 332 calculates the feature value f(N) for each epoch specified by the predetermined time based on the scalar value a calculated by the scalar value calculation unit 331. These will be described in detail in Section 2.
The sleep-wakefulness determination unit 333 is configured to execute the information processing by software (stored in the storage unit 32) specifically realized by hardware (controller 33). The sleep-wakefulness determination unit 333 determines the sleep and wakefulness of the user U based on the feature value f(N) of the desired epoch and the feature value f(N±δ) of the peripheral epochs included in the plurality of epochs before and after the desired epoch in a time series, in the epochs. At this time, the sleep-wakefulness determination unit 333 can determine such sleep and wakefulness based on the above-described machine learning model stored in the storage unit 32.
That is, the number of epochs can be, for example, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, and may be in the range between any two of the numbers indicated above.
The determination result conversion unit 334 is configured to execute the information processing by software (stored in the storage unit 32) specifically realized by hardware (controller 33). It is preferable that the controller 33 further comprises the determination result conversion unit 334. The determination result conversion unit 334 converts the results determined by the sleep-wakefulness determination unit 333. These will be described in detail in Section 4.
Section 2 describes the details of the sleep-wakefulness determination system 1 with reference to the experimental data. In the experiment, the predetermined time to define the epoch was set to 30 seconds, but it should be noted that this is not the limit of the experiment.
As mentioned above, the present sleep-wakefulness determination system 1 extracts the feature value f(N) from the 3-D acceleration vector v (x, y, z) etc. from the L2 norm, and uses the feature value f(N) to determine sleep and wakefulness. Specifically, for example, the feature value f(N) can be a histogram generated by dividing the scalar value a or the logarithm thereof into classes with multiple thresholds, or a power spectrum based on the product of the scalar value a multiplied by a window function. For example,
As described above, the sleep-wakefulness determination unit 333 determines the sleep and wakefulness of the user U based on the feature value f(N) of the desired epoch and the feature value f(N±δ) of the peripheral epochs included in the plurality of epochs before and after the desired epoch in a time series, in the epochs. The number of epochs (the sum of one desired epoch and the peripheral epochs) should be selected appropriately. For reference, a comparison between a case where the number of epochs is 1 and a case where the number of epochs is 9 is shown in
In Section 3, a method of determining the sleep-wakefulness using the sleep-wakefulness determination system 1 is described according to a flowchart shown in
Using the wearable device 2 worn by the user U, information including the time-series 3D acceleration vector v(x, y, z) at a part of the body of the user U is acquired. The information acquired thereby is read into the sleep-wakefulness determination apparatus 3 via the storage media M.
Following the step S1, the scalar value calculation unit 331 in the sleep-wakefulness determination apparatus 3 calculates the scalar value a based on each component (x, y, z) of the 3-D acceleration vector v(x, y, z) at a part of the body of the user U.
Following the step S2, the feature value calculation unit 332 in the sleep-wakefulness determination apparatus 3 calculates the feature value f(N) for each epoch defined by a predetermined time based on the scalar value a calculated by the scalar value calculation unit 331.
Following the step S3, the sleep-wakefulness determination unit 333 in the sleep-wakefulness determination device 3 determines the sleep and wakefulness of the user U by employing the feature value f(N) of the desired epoch and the feature value f(N±δ) of the peripheral epochs included in a plurality of epochs before and after the desired epoch in a time series, in the epochs, as inputs to the machine learning model stored in the storage unit 32
Due to the present method of determining sleep and wakefulness, the sleep and wakefulness with a sufficiently high degree of accuracy using only a few devices can be determined.
In Section 4, a method of converting the results of sleep-wakefulness determination using the sleep-wakefulness determination system 1 is explained according to flowcharts shown in
In the result determined by the sleep-wakefulness determination unit 333, as shown in
Following the step S11, the period of the sleep-wakefulness is calculated using the Chi-square periodogram method.
Following the step S11, the coefficient of variation (standard deviation divided by the mean) is calculated to determine the amplitude of sleep-wakefulness.
According to the method of converting the determination results, the sleep and wakefulness of the user U can be examined from various perspectives.
In Section 5, variations of the sleep-wakefulness determination system 1 according to the present embodiment will be described. That is, the sleep-wakefulness determination system 1 according to the present embodiment may be further devised in the following manner.
First, instead of the storage media M, the communication unit 21 in the wearable device 2 may transmit information including a time-series 3-D acceleration vector v(x, y, z) in a part of the body of the user U to the communication unit 31 in the sleep-wakefulness determination apparatus 31, via wireless communication. In other words, the communication unit 31 is configured to communicate with the wearable device 2 (including the acceleration sensor 23) worn by the user U on a part of the body thereof so as to receive the 3-D acceleration vector v(x, y, z) measured by the acceleration sensor 23.
Secondly, the wearable device 2 and the sleep-wakefulness determination apparatus 3 may be configured as a single unit. In other words, the sleep-wakefulness determination apparatus 3 may be a wearable device 2 to be worn by the user U on a part of the body, and may further comprise an acceleration sensor 23. The acceleration sensor 23 may be configured to measure a 3-D acceleration vector v(x, y, z).
As described above, the present embodiment makes it possible to implement a sleep-wakefulness determination apparatus 3 configured to determine sleep and wakefulness with a sufficiently high degree of accuracy using a small number of wearing devices.
There is provided a sleep-wakefulness determination apparatus configured to determine sleep and wakefulness of a user, comprising a scalar calculation unit configured to calculate a scalar value based on each component of an acceleration vector in a part of a body of the user; a feature value calculation unit configured to calculate a feature value for each epoch defined by a predetermined time based on the scalar value; and a sleep-wakefulness determination unit configured to determine the sleep and wakefulness of the user based on the feature value of a desired epoch and the feature value of peripheral epochs included in a plurality of epochs before and after the desired epoch in a time series, in the epoch.
Software for implementing the sleep-wakefulness determination apparatus 3 as hardware so as to determine the sleep and wakefulness with sufficiently high accuracy using a small number of wearing devices can also be implemented as a program. Such a program may be provided as a non-transitory computer readable medium that can be read by a computer, or may be provided for download from an external server, or may be provided so that the program can be activated on an external computer to realize functions thereof on a client terminal (so-called cloud computing).
There is provided a program which allows a computer to function as a sleep-wakefulness determination apparatus configured to determine sleep and wakefulness of a user, the apparatus comprising: a scalar calculation unit configured to calculate a scalar value based on each component of an acceleration vector in a part of a body of the user; a feature value calculation unit configured to calculate a feature value for each epoch defined by a predetermined time based on the scalar value; and a sleep-wakefulness determination unit configured to determine the sleep and wakefulness of the user based on the feature value of a desired epoch and the feature value of peripheral epochs included in a plurality of epochs before and after the desired epoch in a time series.
Finally, various embodiments of the present invention have been described, but these are presented as examples and are not intended to limit the scope of the invention. The novel embodiment can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the abstract of the invention. The embodiment and its modifications are included in the scope and abstract of the invention and are included in the scope of the invention described in the claims and the equivalent scope thereof.
It may be provided in each of the following forms.
The sleep-wakefulness determination apparatus, wherein the scalar calculation unit is configured to calculate the scalar value based on each component of a time difference vector, the time difference vector being a difference vector of two acceleration vectors in a time series.
The sleep-wakefulness determination apparatus, wherein the scalar calculation unit is configured to calculate the scalar value based on each component of an n-th-order time derivative vector, the n-th-order time derivative vector being a vector in which the acceleration vector is differentiated by n-th-order time derivative, and n being a natural number.
The sleep-wakefulness determination apparatus, wherein the scalar value is an L2 norm or an L1 norm.
The sleep-wakefulness determination apparatus, wherein the feature value is a histogram generated by dividing the scalar value or logarithm thereof into classes with a plurality of threshold values.
The sleep-wakefulness determination apparatus, wherein the feature value is a power spectrum based on a product of the scalar value multiplied by a window function.
The sleep-wakefulness determination apparatus, further comprising a storage unit which storages a machine learning model allowed to learn correlation of the feature value of the desired epoch, the feature value of the peripheral epochs and the sleep and wakefulness of the user, wherein the sleep-wakefulness determination unit is configured to determine the sleep and wakefulness based on the machine learning model.
The sleep-wakefulness determination apparatus, further comprising a storage media reading unit configured to read the acceleration vector stored in storage media.
The sleep-wakefulness determination apparatus, further comprising a communication unit configured to communicate with an acceleration sensor worn on a part of a body of the user, and to receive the acceleration vector measured by the acceleration sensor.
The sleep-wakefulness determination apparatus, the apparatus being a wearable device worn on a part of a body of the user, and further comprising an acceleration sensor configured to measure the acceleration vector.
The sleep-wakefulness determination apparatus, further comprising a determination result conversion unit configured to convert a result determined by the sleep-wakefulness determination unit.
A program which allows a computer to function as a sleep-wakefulness determination apparatus configured to determine sleep and wakefulness of a user, the apparatus comprising: a scalar calculation unit configured to calculate a scalar value based on each component of an acceleration vector in a part of a body of the user; a feature value calculation unit configured to calculate a feature value for each epoch defined by a predetermined time based on the scalar value; and a sleep-wakefulness determination unit configured to determine the sleep and wakefulness of the user based on the feature value of a desired epoch and the feature value of peripheral epochs included in a plurality of epochs before and after the desired epoch in a time series.
Of course, the above embodiments are not limited thereto.
Number | Date | Country | Kind |
---|---|---|---|
2019-125950 | Jul 2019 | JP | national |
This application is a U.S. National Phase application under 35 U.S.C. 371 of International Application No. PCT/JP2020/026345, filed on Jul. 6, 2020, which claims priority to Japanese Patent Application No. 2019-125950, filed on Jul. 5, 2019. The entire disclosures of the above applications are expressly incorporated by reference herein.
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/JP2020/026345 | 7/6/2020 | WO |