Biorhythm estimation system and biorhythm estimation method
The biorhythm estimation system uses heart rate data decomposition to simultaneously estimate sleep and daily cycles, with optional lunar cycle estimation, addressing the limitations of separate device measurements and improving accuracy and convenience.
Patent Information
- Application Number
- JP2023084271
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2023-05-23
- Publication Date
- 2024-12-05
AI Technical Summary
Existing methods for measuring human biorhythms, including sleep, daily, and lunar cycles, require separate devices and impose restrictions, leading to inaccurate and cumbersome measurements, especially for lunar cycles due to hormonal influences.
A biorhythm estimation system that utilizes heart rate data decomposition through empirical mode decomposition to simultaneously estimate sleep and daily cycles, with the option to include lunar cycles for women, using a computer device with processing circuits and machine learning models.
Enables easy and nearly simultaneous estimation of sleep and daily cycles, with the capability to also estimate lunar cycles, improving accuracy and convenience by using a single device for comprehensive biorhythm analysis.
Smart Images

Figure 2024167939000001_ABST
Abstract
Description
[Technical field]
[0001] The present disclosure relates to a biological rhythm estimation system and method, and in particular to an apparatus and method for estimating a person's sleep cycle, circadian cycle, and lunar cycle. [Background technology]
[0002] Humans are said to have three main cycles: a 90-minute sleep cycle, a 25-hour circadian cycle, and a monthly lunar cycle.
[0003] The sleep cycle can be influenced by the quality of sleep itself. The circadian cycle can be influenced by lifestyle and exercise habits. Meanwhile, the monthly cycle can be influenced by premenstrual syndrome caused by hormonal fluctuations, especially in women. The monthly cycle can also be influenced by the state of menopause, including in men.
[0004] These cycles are interrelated, and in order for a person to live in a state of physical and mental well-being, it is desirable to know each cycle at the same time.
[0005] Incidentally, the sleep cycle is usually measured by polysomnography. The circadian cycle is obtained based on deep body temperature (for example, rectal temperature) measured continuously for 24 hours, but since an index of the autonomic nervous system is used to indirectly obtain the value of deep body temperature, it is necessary to impose restrictions on movement and daily activities during measurement. Furthermore, the monthly cycle, which is composed of four phases, namely, the luteal phase, the follicular phase, the menstrual phase, and the ovulation phase, is usually measured by measuring body temperature sublingually every morning immediately after waking up. However, since the measured value at this time is affected by the circadian cycle, the accuracy of the measured and calculated monthly cycle is not necessarily high. As described above, since the measuring devices and measuring methods for each cycle are different, the subject must wear a device for each cycle in order to obtain the three cycles. [Prior art documents] [Patent documents]
[0006] [Patent Document 1] JP 2022-025644 A [Patent Document 2] Patent No. 6534601 Summary of the Invention [Problem to be solved by the invention]
[0007] The biological rhythm estimation system and method disclosed herein for estimating a sleep cycle, a circadian cycle, and a lunar cycle aims to easily and substantially simultaneously estimate at least the sleep cycle and the circadian cycle from time-course change data of the heart rate, and further to easily and substantially simultaneously estimate the lunar cycle as necessary. [Means for solving the problem]
[0008] The biological rhythm estimation system of the present disclosure includes an interface device that externally acquires and externally outputs various data including waveform data of an overnight heart rate, which is time-dependent change data of a subject's heart rate, a memory that stores various programs that realize various processes, and a processing circuit that executes the various programs. a time-varying decomposition unit that decomposes the time-varying data of the heart rate acquired via the interface device into a plurality of signal function data from high frequency components to low frequency components and outputs the signal function data; a biological cycle extraction unit that estimates at least a sleep cycle and a circadian cycle from the plurality of signal function data; Equipped with. Effect of the Invention
[0009] By using the biorhythm estimation system and method disclosed herein, at least the sleep cycle and circadian cycle of a single subject can be easily and substantially simultaneously estimated from the time course data of the heart rate of the subject. Furthermore, if necessary, the lunar cycle can also be easily and substantially simultaneously estimated. [Brief description of the drawings]
[0010] [Figure 1A] FIG. 1A is a system configuration diagram of a biological rhythm estimation system according to an embodiment. [Figure 1B] FIG. 1B is a diagram showing an overall block configuration of a biorhythm estimation system according to an embodiment. [Figure 2A] FIG. 2A is a block diagram of a waveform decomposition section in the biorhythm estimation system according to the embodiment. [Figure 2B] FIG. 2B is a system block diagram of the biological rhythm estimation system according to the embodiment in a learning phase. [Figure 2C] FIG. 2C is a system block diagram of the biorhythm estimation system according to the embodiment in the estimation phase. [Diagram 3] FIG. 3 illustrates the decomposition of an overnight heart rate waveform using empirical mode decomposition. [Figure 4A] Fig. 4A(α) and Fig. 4A(β) are process flow diagrams of a sleep cycle learning phase and a sleep cycle estimation phase in a biorhythm estimation system according to an embodiment. [Figure 4B] Fig. 4B (α) is a diagram showing a schematic diagram of how IMF1 to IMF4 are added together to generate one waveform data (IMF1234). Fig. 4B (β) is a diagram showing the differential value, moving average, and variance, which are learning parameters, calculated from the waveform data (IMF1234) generated by adding together IMF1 to IMF4. [Diagram 5] FIG. 5 is a diagram showing an example of a transition of sleep depth over one night. [Figure 6A] FIG. 6A is a process flow diagram for calculating a circadian cycle in the biorhythm estimation system according to the embodiment. [Figure 6B] Fig. 6B(α) is an example of waveform data (original data) of the overnight heart rate. Fig. 6B(β) is one waveform data (IMF456) obtained by adding together IMF4 to IMF6 in the example of the waveform data (original data) of the overnight heart rate shown in Fig. 6B(α). [Figure 7] Fig. 7(α) is a process flow diagram of the monthly cycle learning phase in the biorhythm estimation system according to the embodiment. Fig. 7(β) is a process flow diagram of the monthly cycle estimation phase in the biorhythm estimation system according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0011] Hereinafter, the embodiments will be described in detail with reference to the drawings as appropriate. However, more detailed explanation than necessary may be omitted. For example, detailed explanation of already well-known matters or duplicate explanation of substantially the same configuration may be omitted. This is to avoid the following explanation becoming unnecessarily redundant and to facilitate understanding by those skilled in the art.
[0012] It should be noted that the inventor provides the accompanying drawings and the following description so that those skilled in the art can fully understand the present disclosure, and they are not intended to limit the subject matter described in the claims.
[0013] 1. [Background to this disclosure] Among human biological rhythms, the three most representative cycles, namely, the sleep cycle, the circadian cycle, and the lunar cycle, are said to be interrelated and affect each other.
[0014] For example, if exercise habits become extremely poor, sleep quality declines, causing disruption of the circadian cycle. Disruption of the circadian cycle also aggravates menstrual symptoms, causing disruption of the lunar cycle. Furthermore, if you are unable to get a good night's sleep (sleep phase delay), disruption of the circadian cycle occurs, which worsens your sleep cycle the following day and beyond. Severe menstrual symptoms can cause sleep disruption, leading to a significant shift in the circadian cycle and social jet lag.
[0015] In such cases, even if you want to improve your sleep, regulate your circadian rhythm, or regulate your monthly cycle, it is difficult to come up with solutions to improve your condition unless you can grasp the three most common cycles simply and roughly simultaneously.
[0016] The present disclosure has been made with the above-mentioned problem in mind. In other words, the biological rhythm estimation system of the present disclosure realizes easy and substantially simultaneous estimation of a sleep cycle, a circadian cycle, and a lunar cycle of a subject from waveform data of the subject's overnight heart rate.
[0017] 2. [Embodiment] Preferred embodiments of the present disclosure will now be described with reference to the accompanying drawings.
[0018] 2.1. [System Configuration] The biological rhythm estimation system according to the embodiment is a device that estimates biological rhythms including sleep cycles, circadian cycles, and lunar cycles from waveform data of overnight heart rate. Fig. 1A is a system configuration diagram of a biological rhythm estimation system 2 according to the present embodiment.
[0019] As will be described in detail later, the lunar cycle, in particular, estimated by the biological rhythm estimation system 2 according to the embodiment, is related to women. Note that the sleep cycle and circadian cycle estimated by the biological rhythm estimation system 2 according to the embodiment can be applied to both men and women. Therefore, the biological rhythm estimated by the biological rhythm estimation system 2 includes at least the sleep cycle and circadian cycle.
[0020] The biorhythm estimation system 2 includes a computer device 4 and a storage device 12. The computer device 4 and the storage device 12 are connected by a wired or wireless communication line, and can transmit and receive data to and from each other. The biorhythm estimation system 2 is preferably further connected to an external network 16, and preferably transmits and receives data to and from other computer systems connected to the external network 16.
[0021] In particular, the biorhythm estimation system 2 is connected to one or more heart rate data recording devices 14 directly or via an external network 16. Furthermore, the biorhythm estimation system 2 is preferably connected to a learning server 18 via the external network 16.
[0022] The computer device 4 is a server machine, a personal computer, a workstation computer, etc., equipped with one or more processors. As will be described later, the computer device 4 may be a processing circuit or a control unit including a processor in a smartwatch.
[0023] The storage device 12 is a storage device provided outside the computer device 4, such as a disk drive or a flash memory, and stores various data sets and various computer programs used by the computer device 4. The storage device 12 records, for example, waveform data of the overnight heart rate transmitted from a heart rate data recording device 14, which will be described later, or from an external device.
[0024] The heart rate data recording device 14 is a recording device such as a disk drive or flash memory, and in particular records a large number of waveform data of the overnight heart rate, which is data on changes in the subject's heart rate over time obtained by a pulse meter or an electrocardiograph. The heart rate data recording device 14 may acquire a large number of waveform data of the overnight heart rate via an external network 16. The heart rate data recording device 14 transmits the waveform data of the overnight heart rate to the computer device 4.
[0025] The learning server 18 uses a large amount of waveform data of overnight heart rates recorded in the heart rate data recording device 14, and ground truth data, etc. to train a learning device and a sleep estimation model, which are machine learning models, as described later. By configuring the computer device 4 to perform the operations of the learning server 18, the learning server 18 does not need to be provided separately and physically.
[0026] The external network 16 is, for example, the Internet, and is connected to the computer device 4 via an interface device 6 such as a network terminal.
[0027] The computing device 4 further includes an interface device 6 , a processing circuit 8 , and a memory 10 .
[0028] The interface device 6 is an interface unit that can acquire data from outside and output data to outside, including a network terminal, a video input terminal, a USB terminal, a keyboard, a mouse, etc. Various data are acquired from outside via the interface device 6. The data is, for example, waveform data of the overnight heart rate, which is data on changes over time in the heart rate of the person being measured, which will be described later. After acquisition, these data can be recorded in the storage device 12. The data recorded in the storage device 12 can be appropriately acquired in the computer device 4 via the interface device 6.
[0029] Various data generated by the biorhythm estimation system 2 is appropriately recorded in the storage device 12. The various data are, for example, short-period IMF data (IMF1-4), long-period IMF data (IMF4-6), lowest point data, etc., which will be described later. The various data generated by the biorhythm estimation system 2 and appropriately recorded in the storage device 12 can be acquired again in the computer device 4 via the interface device 6.
[0030] The processing circuit 8 is composed of a processor. The processor here includes a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit). The various processes of the biological rhythm estimation system 2 according to this embodiment are realized by the processing circuit 8 executing various programs. The various processes may be realized by an ASIC (Application Specific Integrated Circuit) or a combination of these.
[0031] The processing circuit 8 in the present disclosure may be composed of a plurality of signal processing circuits. Each signal processing circuit may be, for example, a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and may be called a "processor." Some of the various processes in the biological rhythm estimation system 2 according to the present embodiment may be executed by one processor (for example, a GPU), and other processes may be executed by another processor (for example, a CPU).
[0032] The memory 10 is a rewritable storage unit within the computer device 4, and is configured, for example, by a RAM (Random Access Memory) including a large number of semiconductor memory elements. The memory 10 temporarily stores specific computer programs, variable values, parameter values, etc., when the processing circuit 8 executes various processes. The memory 10 may include a so-called ROM (Read Only Memory). The ROM stores in advance a computer program that realizes the processes of the biorhythm estimation system 2 described below. The processing circuit 8 reads out the computer program from the ROM and loads it into the RAM, thereby enabling the processing circuit 8 to execute the computer program.
[0033] The biorhythm estimation system 2 according to the present embodiment is constructed using a computer language such as Python. Computer languages that can be used to construct the biorhythm estimation system 2 according to the present disclosure are not limited to these, and other computer languages may of course be used.
[0034] Furthermore, in the biological rhythm estimation system 2 according to the present embodiment, a trained machine learning model is constructed. As will be described later, the machine learning model (learner, sleep estimation model) according to the present embodiment is constructed using a decision tree structure such as a random forest or an extra tree. Of course, the machine learning model according to the present embodiment may be constructed using another structure, for example, a network structure.
[0035] 2.2. [Overall block configuration of the biological rhythm estimation system] FIG. 1B is a diagram showing an overall block configuration of a biological rhythm estimation system 2 according to an embodiment.
[0036] The overall block of the biological rhythm estimation system 2 is made up of at least a time-course change decomposition section 21 and a biological cycle extraction section 30. The overall block of the biological rhythm estimation system 2 may also include a biological rhythm presentation section 62.
[0037] The time-course change decomposition section 21 receives as input data time-course change data 23 of the subject's heart rate. The time-course change data 23 of the subject's heart rate includes waveform data 24 of the subject's overnight heart rate.
[0038] The time-course change decomposition unit 21 decomposes the time-course change data of the heart rate into a plurality of signal function data 25 ranging from high frequency components to low frequency components and outputs the data. The decomposition into signal function data here includes fast Fourier transform (FFT), wavelet transform, empirical mode decomposition (EMD), and the like. The method for decomposition into signal function data may be any method for analyzing non-stationary signals. However, in the following, "waveform decomposition using empirical mode decomposition (EMD)" will be discussed. The waveform decomposition unit 22 included in the time-course change decomposition unit 21 executes waveform decomposition using empirical mode decomposition (EMD).
[0039] The biological cycle extraction unit 30 estimates (extracts) the biological cycle (biorhythm) of the subject from a plurality of signal function data 25 related to the time-dependent change data 23 of the heart rate. The subject's biological cycle (biorhythm) includes at least a sleep cycle and a circadian cycle. In particular, when the subject is a woman, the biological cycle extraction unit 30 estimates (extracts) the monthly cycle.
[0040] 1B, the biological cycle extraction section 30 may include a sleep cycle learning section 32a, a sleep cycle extraction section 32b, a circadian (daily) cycle extraction section 42, a lunar cycle learning section 52a, and a lunar cycle extraction section 52b, which will be described below. These blocks operate when the waveform decomposition section 22 in the time-course change decomposition section 21 operates. Note that if the person being measured is not a woman, the lunar cycle learning section 52a and the lunar cycle extraction section 52b do not need to operate.
[0041] These blocks are realized by loading appropriate computer programs into the memory 10 of the computer device 4 and executing them in the processing circuit 8 in the biological rhythm estimation system 2 according to the embodiment shown in FIG. 1A.
[0042] The estimated (extracted) biological cycle (biorhythm) can be presented by the biological rhythm presentation unit 62.
[0043] 2.3. [Block structure of biological rhythm estimation system] 2A to 2C are block diagrams showing the configuration of each phase of the biorhythm estimation system 2 according to the embodiment. That is, FIG. 2A is a block diagram of the waveform decomposition unit 22 in the biorhythm estimation system 2 according to the embodiment, FIG. 2B is a system block diagram of the biorhythm estimation system 2 according to the embodiment in the learning phase, and FIG. 2C is a system block diagram of the biorhythm estimation system according to the embodiment in the estimation phase. These blocks (diagrams) are also realized by developing appropriate computer programs in the memory 10 of the computer device 4 in the biorhythm estimation system 2 according to the embodiment shown in FIG. 1A and executing them in the processing circuit 8. The block diagrams shown in FIGS. 2A to 2C will be described below.
[0044] 2.3.1. [Block structure centered on waveform decomposition section] FIG. 2A is a block diagram mainly showing the waveform decomposition unit 22 in the biological rhythm estimation system 2. Here, a summary will be given of the well-known "waveform decomposition by empirical mode decomposition (EMD)" performed in the waveform decomposition unit 22. Empirical mode decomposition is one of the methods for analyzing non-stationary signals, and is a technique for converting a time series signal into a non-stationary, non-linear time-frequency space. In other words, performing empirical mode decomposition roughly means decomposing the original signal x(t) into an intrinsic mode function IMF (Intrinsic Mode Function) (sum) and a residue (noise). While FFT and wavelet transform are used to convert into the frequency domain by assuming a basis such as a sine function, empirical mode decomposition is used to perform a so-called heuristic (empirical) transformation without assuming a basis.
[0045] The well-known empirical mode decomposition (EMD) algorithm is as follows (1) to (7). (1) Detect all extreme values of the input signal x_0(t). (2) Interpolate the maximum and minimum points to obtain the upper envelope emax(t) and the lower envelope emin(t). (3) Calculate the local average m(t) = (emin(t) + emax(t)) / 2 of the upper and lower envelopes. (4) The difference between the input and the local average, y(t) = x(t) - m(t), is treated as the input and (1) to (4) are repeated. (5) When the standard deviation of the local average becomes equal to or smaller than a threshold, y(t) is regarded as the IMF and the loop ((1)-(4)) is terminated. (6) The residual between the input signal and the IMF, x_1(t) = x_0(t) - y(t), is used as the new input signal, and steps (1) to (6) are repeated. (7) When all IMFs have been extracted (i.e., when there is only one extreme value of x_n(t)), stop.
[0046] FIG. 3 is a diagram showing an example of decomposition of waveform data of a certain all-night heart rate by empirical mode decomposition (EMD). The first (top) shows the original waveform data of the all-night heart rate. Below that, "IMF1", "IMF2", "IMF3", "IMF4", "IMF5", and "IMF6" are intrinsic mode function data (IMF1 to 6) from six high frequency components to low frequency components, i.e., six short cycles to long cycles. In all data, the horizontal axis is the time (minutes) in all-night sleep. The inventor has empirically understood that one waveform data of the all-night heart rate as shown in FIG. 3 is decomposed into IMF data 1 to 6. Among these IMF data, IMF1, IMF2, IMF3, and IMF4, which are short cycles (i.e., high frequency components), are the four IMF data from the shortest short cycle, and in this embodiment, as will be described later, they are added together to form one waveform data and used for calculating the sleep cycle (see FIG. 4B(α)). It is noted that IMF4 data does not necessarily have to be used.
[0047] In addition, IMF6, IMF5, and IMF4, which are long-period (i.e., low-frequency components), are the three IMF data with the longest periods, and in this embodiment, as will be described later, they are added together to form one waveform data, which is then used to calculate the daily and monthly periods (see FIG. 6B(β)). Note that IMF4 data does not necessarily have to be used.
[0048] 2A, a waveform decomposition unit 22 that performs empirical mode decomposition (EMD) inputs waveform data 24 of the all-night heart rate, and decomposes and outputs short-cycle IMF data (IMF1-4) 26 and long-cycle IMF data (IMF4-6) 28. As described below, the waveform data 24 of the all-night heart rate may be for learning or may be for estimation.
[0049] 2.3.2. [Block structure of the biological rhythm estimation system in the learning phase] Next, Fig. 2B is a system block diagram of the biorhythm estimation system 2 according to the embodiment in the learning phase. First, the sleep cycle learning unit 32a adds and synthesizes the short-period IMF data (IMF1-4) 26a for learning for IMF1-IMF4 to form one waveform data (see Fig. 4B(α)), and uses it as input data. Furthermore, the sleep cycle learning unit 32a includes a learning parameter calculation unit 34a, a sleep estimation model 36a, and a correct data (sleep depth data) holding unit 38a. The short-period IMF data (IMF1-4) 26a for learning, which are added and synthesized to form one waveform data, are a plurality (sets) of short-period IMF data (IMF1-4), and the sleep depth data for each of them is clear as correct data, and the correct data (sleep depth data) is held in the correct data (sleep depth data) holding unit 38a.
[0050] Each piece of sleep depth data represents a transition of sleep depth over one night, as shown in Fig. 5. As described above, such a transition of sleep depth over one night is usually measured using polysomnography.
[0051] The learning parameter calculation unit 34a calculates learning parameters (see FIG. 4B(β)) such as differential values, moving averages, and variances for each of the short-period learning IMF data (IMF1-4) 26a that have been added together and synthesized into one waveform data. The sleep estimation model 36a is trained using the learning parameters calculated by the learning parameter calculation unit 34a as explanatory variables and the sleep depth data stored in the correct data (sleep depth data) storage unit 38a corresponding to each of the short-period learning IMF data (IMF1-4) 26a as objective variables.
[0052] The sleep estimation model 36a is assumed to be constructed using a random forest or an extra tree, but may be constructed using other structures, for example, a network structure.
[0053] In the learning phase, the circadian (daily) cycle extraction unit 42 adds and synthesizes the long-period IMF data (IMF4-6) 28a for learning for IMF4-IMF6 to generate one waveform data (see FIG. 6B (β)), which it then uses as input data. Furthermore, the circadian (daily) cycle extraction unit 42 includes a lowest point calculation unit 44 and a lowest point time calculation unit 46, and outputs lowest point data 48 (for learning). The lunar cycle learning unit 52a uses the lowest point data 48 (for learning) as input data, and includes a learner 54a and a correct data (label data) holding unit 56a.
[0054] In the learning phase of the biorhythm estimation system 2, the circadian (daily) cycle extraction unit 42 itself does not perform learning. In the learning phase of the biorhythm estimation system 2, the circadian (daily) cycle extraction unit 42 is for generating (outputting) the lowest point data 48, which is input data for the subsequent monthly cycle learning unit 52a, i.e., learning data. Since the lowest point data 48 (for learning) is learning data for the learning device 54a for estimating the monthly cycle, the lowest point data 48 for learning requires at least one month's worth of daily "lowest point data" for one woman. Therefore, in the learning phase of the biorhythm estimation system 2, the long-period IMF data (IMF4-6) 28a for learning, which is input data for the circadian (daily) cycle extraction unit 42, also requires at least one month's worth of daily data for one woman.
[0055] As described above, the long-period IMF data (IMF4-6) 28a for learning is a plurality of long-period IMF data (IMF4-6), for which at least one month's worth of daily data is required for one woman, and for each of them, the label data is made clear as the correct answer data, and this correct answer data (label data) is stored in the correct answer data (label data) storage unit 56a.
[0056] The label data as the correct answer data here is, for example, data consisting of four phases: luteal phase, follicular phase, menstrual phase, and ovulation phase. The label data here may be composed of less than four phases, or may be composed of four or more phases, as long as it is composed of a predetermined number of phases.
[0057] The learning device 54a in the lunar cycle learning unit 52a learns using, as explanatory variables, the "lowest points" of the lowest point data extracted by the circadian (daily) cycle extraction unit 42 from the long-period IMF data for learning (IMF4-6) 28a that have been added together and synthesized to form a single waveform data, and using, as objective variables, the label data held in the correct data (label data) holding unit 56a corresponding to the long-period IMF data for learning (IMF4-6) 28a from which the "lowest points" were extracted.
[0058] The learner 54a is also assumed to be constructed using a random forest or an extra tree, although the learner 54a may be constructed using other structures, for example, a network structure.
[0059] 2.3.3. [Block configuration of the estimation phase of the biological rhythm estimation system] Next, Fig. 2C is a system block diagram in the estimation phase of the biorhythm estimation system 2 according to the embodiment. First, the sleep cycle extraction unit 32b adds and synthesizes the short cycle IMF data (IMF1-4) 26b to be estimated for IMF1-IMF4 to generate one waveform data (see Fig. 4B(α)), and then uses it as input data. Furthermore, the sleep cycle extraction unit 32b includes an extraction parameter calculation unit 34b and a trained sleep estimation model 36b.
[0060] The extraction parameter calculation unit 34b calculates extraction parameters such as a differential value, a moving average, and a variance for each of the short cycle IMF data (IMF1-4) 26b to be estimated that have been added together and synthesized into one waveform data. Using the extraction parameters calculated by the extraction parameter calculation unit 34b as explanatory variables, sleep depth data that should correspond to each of the short cycle IMF data (IMF1-4) 26b to be estimated is extracted (i.e., estimated) as a response variable by the trained sleep estimation model 36b.
[0061] The biorhythm presentation unit 62 presents the extracted (i.e., estimated) sleep depth data and the sleep cycle calculated from the sleep depth data. The biorhythm presentation unit 62 is configured with a presentation device such as a display and a speaker, and presents the sleep depth data and the sleep cycle.
[0062] The circadian (daily) cycle extraction unit 42 receives input of the long cycle IMF data (IMF4-6) to be estimated, which have been added together and synthesized to form a single waveform data (see FIG. 6B (β)), and generates lowest point data 48 (of the estimation target) by a lowest point calculation unit 44 and a lowest point time calculation unit 46. In order to calculate the lowest point data more accurately, it is desirable to have a plurality of long cycle IMF data (IMF4-6) to be estimated, such as one month's worth of data for one person. The lunar cycle extraction unit 52b receives the lowest point data 48 (to be estimated) as input data, and includes a trained learner 54b.
[0063] In the estimation phase of the biorhythm estimation system 2, the circadian (daily) cycle extraction unit 42 generates (outputs) lowest point data 48 (of the estimation target) as input data for the subsequent monthly cycle extraction unit 52b. Since the lowest point data 48 (of the estimation target) is input data to the trained learner 54b for estimating the monthly cycle, the lowest point data 48 of the estimation target requires at least one month's worth of daily "lowest point data" for one woman. Therefore, in the estimation phase of the biorhythm estimation system 2, it is desirable to prepare at least one month's worth of daily data for one woman for the long-period IMF data (IMF4-6) 28b of the estimation target, which is the input data for the circadian (daily) cycle extraction unit 42.
[0064] The trained learner 54b in the lunar cycle extraction unit 52b uses the “lowest point” of the lowest point data 48 (of the estimation target) as an explanatory variable, and extracts, as a dependent variable, label data corresponding to each of the lowest point data 48, and further, the long-period IMF data (IMF4-6) 28b of the estimation target which is the basis for extracting the “lowest point”.
[0065] The biorhythm presentation unit 62 presents a diurnal cycle and a transition of the diurnal cycle based on the transition of the extracted (i.e., estimated) lowest point data 48, as well as a monthly cycle and a transition of the monthly cycle based on the transition of the label data consisting of a predetermined number of phases including the four phases of the luteal phase, the follicular phase, the menstrual phase, and the ovulation phase.
[0066] 2.4.[System Operation] 4A, 6A, and 7 are process flow diagrams of the learning phase and estimation (extraction) phase of the sleep cycle, circadian cycle, and monthly cycle in the biorhythm estimation system 2 according to this embodiment. The details of the processing of the learning phase and estimation (extraction) phase of the sleep cycle, circadian cycle, and monthly cycle in the biorhythm estimation system 2 will be described with reference to these Figs. 4A, 6A, and 7.
[0067] 2.4.1. [Learning and Estimation Phases of the Sleep Cycle] FIG. 4A(α) is a process flow diagram of the sleep cycle learning phase in the biological rhythm estimation system 2 according to the embodiment. In the learning phase, first, a plurality of short-period intrinsic mode function data for learning (IMF1 to IMF4) are prepared (step S02). At this time, the data IMF1 to IMF4 are added together and synthesized to form one waveform data. FIG. 4B(α) is a diagram showing a schematic diagram of how IMF1 to IMF4 are added together and synthesized to form one waveform data (IMF1234). In FIG. 4B(α), the horizontal axis represents the time (minutes) during a full night's sleep.
[0068] Next, learning parameters such as differential value, moving average, and variance are calculated (step S04). Fig. 4B (β) is a diagram showing learning parameters, differential value, moving average, and variance, calculated from waveform data (IMF1234) generated by adding together IMF1 to IMF4. In Fig. 4B (β), the horizontal axis also represents the time (minutes) during a full night's sleep.
[0069] Next, a sleep estimation model is trained using the calculated training parameters and the corresponding correct data (sleep depth data) (step S06), and a trained sleep estimation model is generated (step S08).
[0070] Next, Fig. 4A (β) is a process flow diagram of the sleep cycle estimation phase in the biological rhythm estimation system 2 according to the embodiment. In the estimation phase, first, short cycle intrinsic mode function data (IMF1 to IMF4) to be estimated for estimating sleep depth is input (step S12). At this time, the data IMF1 to IMF4 are also added together and synthesized to form one waveform data (see Fig. 4B (α)).
[0071] Next, the sleep depth transition is estimated using the trained sleep estimation model (step S14). Fig. 5 is a diagram showing an example of the sleep depth transition over one night. The horizontal axis represents the sleep time (hours). The sleep cycle is calculated from the estimated sleep depth transition, and the sleep depth transition and the sleep cycle are presented by the biorhythm presentation unit 62 (step S16).
[0072] 2.4.2. [Calculation of Circadian Cycle] FIG. 6A is a process flow diagram for calculating a circadian cycle in a biorhythm estimation system 2 according to an embodiment. First, a plurality of long-period intrinsic mode function data (IMF4 to IMF6) for calculating a circadian cycle are prepared (step S22). At this time, the data IMF4 to IMF6 are added together to form one waveform data. FIG. 6B (β) shows one waveform data (IMF456) in which IMF4 to IMF6 are added together. FIG. 6B (α) shows the original data. In addition, in order to calculate the lowest point data more accurately, it is preferable that there are a plurality of long-period IMF data (IMF4 to 6) to be estimated, such as one month's worth for one person. In both of FIG. 6B (α) and (β), the horizontal axis indicates the time (minutes) during a full night's sleep.
[0073] Next, for the plurality of IMF4 to IMF6 (IMF456) that are treated as one waveform data, the lowest point of the heart rate is calculated for each night (step S24). Furthermore, for the plurality of IMF4 to IMF6 (IMF456) that are treated as one waveform data, the time of the lowest point of the heart rate is calculated for each night (step S26). FIG. 6B (β) shows the lowest point and the lowest point time in one waveform data (IMF456) that is synthesized by adding together IMF4 to IMF6. From the data on the plurality of lowest points and the lowest point times (i.e., lowest point data), the daily cycle (circadian cycle) and the transition of the daily cycle are calculated and presented by the biorhythm presentation unit 62 (step S28).
[0074] 2.4.3. [Learning and estimation phases of the lunar cycle] 7(α) is a process flow diagram of the monthly cycle learning phase in the biological rhythm estimation system 2 according to the embodiment. First, minimum point data 48 of the circadian cycle (daily cycle) for learning is prepared (step S42). The minimum point data 48 (for learning) is learning data for the learning device 54a for estimating the monthly cycle, so the minimum point data 48 for learning requires at least one month's worth of daily "minimum point data" for one woman.
[0075] Next, the learning device 54a is trained using the lowest score data 48 for learning and the corresponding correct answer data (label data) (step S44), and a trained learning device 54b is generated (step S08).
[0076] 7(β) is a process flow diagram of the monthly cycle estimation phase in the biological rhythm estimation system 2 according to the embodiment. In the estimation phase, first, lowest point data 48 of the circadian cycle (daily cycle) is input to estimate the monthly cycle (step S52). The lowest point data 48 to be estimated is input data to the trained learner 54b for estimating the monthly cycle, so the lowest point data 48 to be estimated requires at least one month's worth of daily "lowest point data" for one woman.
[0077] Next, the trained learning device estimates the transition of the label data consisting of a predetermined number of periods in the monthly cycle (step S54). The lunar cycle is calculated from the estimated transition of the predetermined number of periods, and the lunar cycle and the transition of the lunar cycle are presented by the biorhythm presentation unit 62 (step S56).
[0078] 2.5. [Summary of the embodiment] The biological rhythm estimation system 2 according to the embodiment includes an interface device 6 that acquires from outside and outputs to outside various data including waveform data 24 of the overnight heart rate, which is time-varying data 23 of the subject's heart rate, a memory 10 that stores various programs for implementing various processes, and a processing circuit 8 that executes the various programs. The processing circuit 8 includes a time-varying decomposition section 21 that decomposes the time-varying data 23 of the heart rate acquired via the interface device 6 into multiple signal function data 25 from high frequency components to low frequency components and outputs the multiple signal function data 25, and a biological cycle extraction section 30 that estimates at least a sleep cycle and a circadian cycle from the multiple signal function data 25.
[0079] In the biological rhythm estimation system of this embodiment, at least the sleep cycle and circadian cycle of a single person can be easily and substantially simultaneously estimated from only the time-dependent change data of the person's heart rate.
[0080] 3. [Other embodiments] As described above, the embodiments have been described as examples of the technology disclosed in the present application. However, the technology in the present disclosure is not limited to these, and may be applied to embodiments in which modifications, substitutions, additions, omissions, etc. are made as appropriate.
[0081] In recent years, smartwatches have become capable of measuring and storing heart rate throughout the night. Thus, by loading and deploying a computer program that realizes various operations of the biological rhythm estimation system according to the embodiment and a trained machine learning model into a processing circuit including a processor in the smartwatch, it becomes possible to realize the various operations of the biological rhythm estimation system according to the embodiment, i.e., the operations of estimating sleep cycles, circadian cycles, and lunar cycles, in the smartwatch.
[0082] The heart rate data recorder 14 may record waveform data of the heart rate throughout the night obtained by a mat-type heart rate measuring device or a wireless millimeter wave type heart rate measuring device. This may contribute to the establishment of a system for automatically monitoring the biological rhythms, including the sleep cycle, circadian cycle, and lunar cycle, of the elderly and children.
[0083] In addition, the accompanying drawings and detailed description are provided to explain the embodiments. Therefore, among the components described in the accompanying drawings and detailed description, not only essential components for solving the problem but also components that are not essential for solving the problem in order to illustrate the above technology may be included. Therefore, the fact that the non-essential components are described in the accompanying drawings or detailed description should not be interpreted as immediately indicating that the non-essential components are essential.
[0084] Furthermore, since the above-described embodiments are intended to illustrate the technology in the present disclosure, various modifications, substitutions, additions, omissions, and the like can be made within the scope of the claims or their equivalents. [Explanation of symbols]
[0085] 2 biorhythm estimation system, 4 computer device, 6 interface device, 8 processing circuit, 10 memory, 12 storage device, 14 heart rate data recording device, 22 waveform decomposition section, 23 time-dependent change data of heart rate, 24 waveform data of overnight heart rate, 25 signal function data, 26 short-period IMF data (IMF1-4), 26a short-period IMF data for learning (IMF1-4), 26b short-period IMF data to be estimated (IMF1-4), 28 long-period IMF data (IMF4-6), 28a long-period IMF data for learning data (IMF4 to 6), 28b...long cycle IMF data to be estimated (IMF4 to 6), 30...biological cycle extraction unit, 32a...sleep cycle learning unit, 32b...sleep cycle extraction unit, 34a...learning parameter calculation unit, 34b...extraction parameter calculation unit, 36a...sleep estimation model, 36b...trained sleep estimation model, 38a...correct data (sleep depth data) holding unit, 48...lowest point data, 52a...lunar cycle learning unit, 52b...lunar cycle extraction unit, 54a...learner, 54b...trained learner, 56b...correct data (label data) holding unit, 62...biorhythm presentation unit.
Claims
1. an interface device that externally acquires various data including waveform data of the overnight heart rate, which is data on changes in the heart rate of the subject over time, and outputs the data to the outside; A memory in which various programs for implementing various processes are stored; A processing circuit for executing the various programs; A biological rhythm estimation system comprising: The processing circuitry comprises: a time-varying decomposition unit that decomposes the time-varying data of the heart rate acquired via the interface device into a plurality of signal function data from high frequency components to low frequency components and outputs the signal function data; a biological cycle extraction unit that estimates at least a sleep cycle and a circadian cycle from the plurality of signal function data; A biological rhythm estimation system comprising:
2. The time-dependent change decomposition unit a waveform decomposition unit that performs empirical mode decomposition, which inputs waveform data of an overnight heart rate, which is data on changes in the heart rate over time acquired via the interface device, and decomposes the data into short-period intrinsic mode function data, which are high-frequency components, and long-period intrinsic mode function data, which are low-frequency components, and outputs the data; The biorhythm estimation system according to claim 1 .
3. The biological cycle extraction unit further includes a sleep cycle learning unit, The sleep cycle learning unit includes: a learning parameter calculation unit that calculates predetermined learning parameters for learning short-period intrinsic mode function data output by the waveform decomposition unit in response to input of multiple learning waveform data of overnight heart rates; a sleep depth data storage unit that stores sleep depth data that is correct answer data corresponding to the short-period intrinsic mode function data for learning; Including, the sleep cycle learning unit learns a sleep estimation model using the predetermined parameters for learning and the sleep depth data as teacher data, to obtain a learned sleep estimation model. The biorhythm estimation system according to claim 2 .
4. The biological cycle extraction unit further includes a sleep cycle extraction unit including the trained sleep estimation model, The sleep cycle extraction unit an extraction parameter calculation unit that calculates predetermined parameters for extraction from short-period intrinsic mode function data of the estimation target outputted from the waveform decomposition unit in response to input of waveform data of all-night heart rates of the estimation targets; the sleep cycle extraction unit extracts, as an explanatory variable, sleep depth data corresponding to each of the short cycle intrinsic mode function data to be estimated, using the trained sleep estimation model, the predetermined parameters for extraction calculated by the parameter calculation unit for extraction, and as an objective variable, thereby calculating a sleep cycle; The biorhythm estimation system according to claim 3 .
5. The biological cycle extraction unit further includes a circadian cycle extraction unit, The circadian cycle extraction unit is Calculating minimum point data of the plurality of estimation targets for long-period intrinsic mode function data of the estimation targets output by the waveform decomposition unit in response to input of waveform data of the night-time heart rates of the plurality of estimation targets, and further calculating a diurnal cycle and a transition of the diurnal cycle from the minimum point data of the plurality of estimation targets. The biorhythm estimation system according to claim 4 .
6. The biological cycle extraction unit further includes a lunar cycle learning unit, The circadian rhythm extraction unit calculates a plurality of learning minimum point data for the learning long-period intrinsic mode function data output by the waveform decomposition unit in response to input of a plurality of learning overnight heart rate waveform data; The lunar cycle learning unit includes: A label data storage unit stores label data, which is correct answer data corresponding to the long-period intrinsic mode function data for learning and is composed of a predetermined number of periods; the lunar cycle learning unit learns a learning device using the plurality of lowest point data for learning and label data consisting of the predetermined number of periods as teacher data, to obtain a learned learning device; The biorhythm estimation system according to claim 5 .
7. The biological cycle extraction unit further includes a lunar cycle extraction unit including the trained learning device, The lunar cycle extraction unit, by the trained learning device, using the lowest point data of the plurality of estimation objects calculated by the daily cycle extraction unit as an explanatory variable, extracting label data corresponding to long-period intrinsic mode function data of the plurality of estimation objects, which is a basis for calculating the lowest point data of each of the estimation objects, as a dependent variable, and calculating a monthly cycle and a transition of the monthly cycle based on the extracted label data; The biorhythm estimation system according to claim 6.
8. The biological rhythm estimation system according to claim 7 , further comprising a display unit configured to display at least one of the calculated sleep cycle, circadian cycle, and lunar cycle.
9. The short-period intrinsic mode function data is a predetermined number of intrinsic mode function data from the shortest short period in empirical mode decomposition performed on the waveform data of each overnight heart rate, The long-period intrinsic mode function data is a predetermined number of intrinsic mode function data from the longest long period in empirical mode decomposition performed on the waveform data of each overnight heart rate, The predetermined parameters include a differential value, a moving average, and a variance; The predetermined phases constituting the label data are composed of four phases: a luteal phase, a follicular phase, a menstrual phase, and an ovulation phase; The presentation unit is configured by a presentation device including a display or a speaker. The biorhythm estimation system according to claim 8.
10. a time-varying decomposition step of decomposing the time-varying data of the heart rate into a plurality of signal function data from high frequency components to low frequency components and outputting the signal function data; a biological cycle extraction step of estimating at least a sleep cycle and a circadian cycle from the plurality of signal function data by the processing circuit; A biological rhythm estimation method comprising:
11. The time-dependent decomposition step includes: An empirical mode decomposition execution step of inputting waveform data of an overnight heart rate, which is time-dependent change data of the heart rate, and performing empirical mode decomposition to decompose the data into short-period intrinsic mode function data, which is a high-frequency component, and long-period intrinsic mode function data, which is a low-frequency component, and outputting the data; The biological rhythm estimation method according to claim 10.
12. The biological cycle extraction step includes: A first input step of inputting a plurality of learning waveform data of overnight heart rates; A step of calculating predetermined parameters for learning with respect to the short-period intrinsic mode function data for learning outputted in the empirical mode decomposition performing step based on the plurality of learning waveform data of overnight heart rates in the first input step; a step of learning a sleep estimation model using the predetermined learning parameters and sleep depth data, which is correct data stored in a memory and corresponds to the learning short-period intrinsic mode function data, as teacher data to generate a learned sleep estimation model; The method for estimating a biological rhythm according to claim 11 , comprising:
13. The biological cycle extraction step further includes: A second input step of inputting waveform data of multiple overnight heart rates to be estimated; A step of calculating predetermined parameters for extraction for short-period intrinsic mode function data of the estimation target outputted in the empirical mode decomposition executing step based on waveform data of the multiple estimation targets in the second input step; extracting sleep depth data corresponding to the short cycle intrinsic mode function data of the estimation target as a response variable using the calculated predetermined parameters for extraction as an explanatory variable by the trained sleep estimation model, thereby calculating a sleep cycle; The method for estimating a biological rhythm according to claim 12, comprising:
14. The biological cycle extraction step further includes: A step of calculating minimum point data of the plurality of estimation objects for long-period intrinsic mode function data of the estimation objects output in the empirical mode decomposition executing step based on waveform data of the overnight heart rates of the plurality of estimation objects in the second input step, and calculating a diurnal cycle and a transition of the diurnal cycle based on the minimum point data of the plurality of estimation objects. The method for estimating a biological rhythm according to claim 13, comprising:
15. The biological cycle extraction step further includes: Calculating a plurality of learning minimum point data for the long-period learning intrinsic mode function data outputted in the empirical mode decomposition executing step based on the plurality of learning overnight heart rate waveform data in the first input step; a step of training a learning device using the plurality of lowest point data for learning and label data consisting of a predetermined number of periods, which is correct answer data stored in a memory and corresponds to the long-period intrinsic mode function data for learning, as training data to generate a trained learning device; The method for estimating a biological rhythm according to claim 14, comprising:
16. The biological cycle extraction step further includes: a step of using the lowest point data of the plurality of estimation objects calculated in the step of calculating the daily cycle and the transition of the daily cycle by the trained learning device as explanatory variables, extracting label data corresponding to long-period intrinsic mode function data of the plurality of estimation objects, which is the basis for calculating the lowest point data of each of the estimation objects, as a dependent variable, and calculating a lunar cycle and the transition of the lunar cycle using the extracted label data; The method for estimating a biological rhythm according to claim 15, comprising:
17. Furthermore, A step of displaying at least one of the calculated sleep cycle, circadian cycle, and monthly cycle by the display unit. The method for estimating a biological rhythm according to claim 16, comprising:
18. The short-period intrinsic mode function data is a predetermined number of intrinsic mode function data from the shortest short period in empirical mode decomposition performed on the waveform data of each overnight heart rate, The long-period intrinsic mode function data is a predetermined number of intrinsic mode function data from the longest long period in empirical mode decomposition performed on the waveform data of each overnight heart rate, The predetermined parameters include a differential value, a moving average, and a variance; The predetermined phases constituting the label data are composed of four phases: a luteal phase, a follicular phase, a menstrual phase, and an ovulation phase; The presentation unit is configured by a presentation device including a display or a speaker. The biorhythm estimation method according to claim 17.
Citation Information
Patent Citations
Biological condition management device, biological condition management system, biological condition management method, and computer program
JP2022025644A
Life rhythm recommendation device
JP6534601B2
