Physiological state estimation system, physiological state estimation method, and physiological state estimation program

Abstract

To provide a physiological state estimation system, a physiological state estimation method, and a physiological state estimation program capable of estimating changes in various physiological states of a living body.SOLUTION: A physiological state estimation system 1 includes: a feature amount calculation unit 201 for dividing a time-series signal indicating a physiological state of a living body into a plurality of frequency bands and calculating a feature amount from a time-series variation of the amplitude and frequency in each frequency band of the time-series signal; a displacement vector calculation unit 202 for calculating a displacement vector indicating a change in the physiological state of the living body, the physiological state of the living body being represented in a multi-dimensional vector space consisting of the calculated feature amount; and a physiological state estimation unit 203 for synthesizing displacement vectors before and after any physiological state change using a displacement vector before and after desired environmental stimulation, and estimating the physiological state of the living body on the basis of the synthesization ratio of the synthesized displacement vectors. The feature amount calculation unit 201, the displacement vector calculation unit 202, and the physiological state estimation unit 203 can be realized using a model learned by machine learning.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a physiological state estimation system, a physiological state estimation method, and a physiological state estimation program for estimating the physiological state of a living body. [Background technology]

[0002] Various techniques for estimating the physiological state of a living organism have been proposed. Patent Document 1 discloses a physiological state estimation device that calculates a time-series waveform of frequency from the time-series waveform of a biological signal and performs frequency analysis on the time-series change in the slope of the frequency obtained for each predetermined time window. The physiological state estimation device then calculates a reference point for each analysis waveform using at least one of a region score based on the slope of each regression line of each analysis waveform obtained by the frequency analysis and a shape score based on the number of breaks in each regression line. The physiological state estimation device estimates the living organism's state based on the time-series change in the reference point for each analysis waveform. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2012-239480 [Patent Document 2] Patent No. 7327363 [Non-patent literature]

[0004] [Non-Patent Document 1] HSTeoh, “Formula for Vector Rotation in Arbitrary Planes in Rn”, [online], April 17, 2005, [Retrieved December 14, 2023], Internet<URL:https: / / www.qfbox.info / 4d / genrot.pdf> Summary of the Invention [Problem to be solved by the invention]

[0005] However, the physiological state estimation device disclosed in Patent Document 1 uses, as a feature, a reference point of each analysis waveform obtained by frequency analysis of time-series changes in the slope of a time-series waveform of frequency based on a biological signal, and therefore the feature that can represent the physiological state of a living organism is limited to the reference point, resulting in a problem that it cannot estimate various changes in the physiological state of a living organism.

[0006] The present disclosure is intended to solve such problems, and aims to provide a physiological state estimation system, a physiological state estimation method, and a physiological state estimation program that are capable of estimating various changes in the physiological state of a living body. [Means for solving the problem]

[0007] A physiological state estimation system for estimating a physiological state of a living body according to an embodiment includes: a feature calculation unit that divides a time-series signal representing a physiological state of a living organism into a plurality of frequency bands and calculates a feature from time-series fluctuations of amplitude and frequency in each frequency band of the time-series signal; a displacement vector calculation unit that represents the physiological state of the living organism in a multidimensional vector space made up of the calculated feature amounts and calculates a displacement vector that indicates a change in the physiological state of the living organism in the multidimensional vector space; a storage unit that stores displacement vectors before and after a plurality of environmental stimuli; The system includes a physiological state estimation unit that uses displacement vectors before and after a desired environmental stimulus stored in the memory unit to synthesize displacement vectors before and after any physiological state change, and estimates the physiological state of the living organism based on the synthesis ratio of the synthesized displacement vectors.

[0008] The physiological state estimation unit has a database of displacement vectors in a multidimensional vector space for multiple environmental stimuli, represents the displacement vector in the multidimensional vector space for a specific time series as a linear combination of the displacement vectors for multiple environmental stimuli in the database, and estimates the combination coefficient as a ratio of the physiological state.

[0009] Specifically, the feature calculation unit calculates feature quantities representing a person's physiological state from their morning and evening biological state time-series data. Then, the displacement vector calculation unit calculates a displacement vector in a multidimensional vector space. When the displacement vector can be synthesized using, for example, three vectors in a database representing mental fatigue, physical fatigue, and rest in a ratio of A:B:C, the physiological state estimation unit estimates that the changes in the person's physiological state from morning to evening are mental fatigue A, physical fatigue B, and rest C.

[0010] The feature calculation unit performs a Hilbert transform on waveform information of biological information for each predetermined frequency band, and calculates the instantaneous amplitude and instantaneous frequency of the Hilbert-transformed waveform information as features. Next, the feature calculation unit calculates the distributions of the instantaneous amplitude and instantaneous frequency in a section that can be considered a physiological steady state, in other words, in a section where the distribution is unimodal. The calculated distributions of the instantaneous amplitude and instantaneous frequency are each Gaussian-like. The feature calculation unit then approximates the calculated distributions of the instantaneous amplitude and instantaneous frequency with normal distributions, and calculates the mean and standard deviation for each of the distributions of the instantaneous amplitude and instantaneous frequency. In this way, the feature calculation unit calculates the mean and standard deviation of the distribution of the instantaneous amplitude and the mean and standard deviation of the distribution of the instantaneous frequency as features representing the physiological state (see Patent Document 2).

[0011] The frequency bands of the autonomic nervous system are known to be HF (High Frequency) (0.15-0.4 Hz), which represents the frequency of respiratory fluctuations in heart rate variability, and LF (Low Frequency) (0.04-0.15 Hz), which represents the frequency of blood pressure fluctuations. We have confirmed the frequency bands of VLF (Very Low Frequency) 1 (15-40 mHz) and VLF 2 (4-15 mHz), which represent autonomic nervous fluctuations at lower frequencies than these, and have also confirmed that there are frequency bands at the same ratio (i.e., 1 / 10 times) for even lower frequency bands.

[0012] To calculate features from autonomic nervous fluctuations at frequencies lower than the VLF2 band, time-series physiological signals with three or more wavelengths are required to calculate the standard deviation of the instantaneous amplitude distribution and the standard deviation of the instantaneous frequency distribution in that band. For example, in the ULF (ultra-low frequency)1 (1.5-4 mHz) band, 2000 seconds (= 3 / 1.5 mHz) of time-series data is required. Maintaining a physiological steady state over a time interval of more than 30 minutes is difficult because subjects may be exposed to various environmental stimuli. Therefore, when features are calculated in the four frequency bands of HF, LF, VLF1, and VLF2, 16 features (= 4 bands × 2 (instantaneous amplitude and instantaneous frequency) × 2 (mean and standard deviation)) are obtained, forming a 16-dimensional vector space. These features are mutually independent.

[0013] However, when physiological state time series data is obtained during a relatively long period of time that can be considered a physiologically steady state, such as during sleep, frequency bands lower than VLF2, such as ULF1 (1.5 to 4 mHz), ULF2 (0.4 to 1.5 mHz), UULF (Ultra ULF)1 (0.15 to 0.4 mHz), UULF2 (0.04 to 0.15 mHz), etc., may be added to calculate features and form a multidimensional vector space.

[0014] Whether or not a physiological steady state is present can be determined by applying the Hilbert transform to a certain frequency band waveform of a physiological state time series signal and checking whether the instantaneous amplitude and instantaneous frequency are distributed like a Gaussian. If the time series signal contains multiple physiological states, the distribution of the instantaneous amplitude and instantaneous frequency will be multi-peaked.

[0015] Here, the physiological state time-series signal may be any biomeasurement data that provides physiological state information, such as electrocardiogram waveforms, pulse waveforms, cerebral blood flow waveforms, fMRI (functional Magnetic Resonance Imaging) waveforms, and far-infrared images.

[0016] When the physiological state time-series data is a pulse wave, pulse interval data can be extracted from the pulse wave, resulting in two time-series data: a pulse wave series and a pulse interval series. The instantaneous phase difference can be defined from these two time-series data. When the distribution of this instantaneous phase difference is calculated, it resembles a Von Mises distribution in a physiological steady state, allowing the mean value and concentration of the instantaneous phase difference to be defined. This instantaneous phase difference is an important feature because it contains information about vasoconstriction and relaxation in the autonomic nervous system. Therefore, when a multidimensional vector space is formed using the pulse wave, a 24-dimensional (= 4 frequency bands × 3 (instantaneous amplitude, instantaneous frequency, and instantaneous phase difference) × (mean and variance)) vector space can be formed (see Patent Document 2).

[0017] In addition, using time series data of physiological states such as electroencephalograms and cerebral blood flow, feature values ​​can be calculated for the frequency bands δ (0.4 to 4 Hz), θ (4 to 8 Hz), α (8 to 13 Hz), β (13 to 30 Hz), and γ (30 Hz or higher) to form a multidimensional vector space.

[0018] This section describes a method for calculating a database of displacement vectors in a multidimensional vector space in response to multiple environmental stimuli. Using parameter data that are statistically significant for a certain environmental stimulus, the feature values ​​before and after the environmental stimulus are calculated and mapped to multidimensional spatial coordinates. The feature values ​​are standardized using data from before the multiple environmental stimuli, with an average value of 0 and a standard deviation of 1. This is to ensure accuracy when estimating the physiological state. Next, a displacement vector from the multidimensional spatial coordinates before the environmental stimulus to the multidimensional spatial coordinates after the environmental stimulus is calculated. The calculated displacement vectors are then averaged with respect to the parameter values ​​to calculate an average vector, and the calculated average vector is registered in a database as the displacement vector for that environmental stimulus.

[0019] The above process is repeated for a statistically significant number of subjects, calculating the features before and after stimulation for, for example, mental fatigue, physical fatigue, rest, sleep, massage, heat, cold, noise, vibration, etc., and creating a database of displacement vectors and information distances based on these features.

[0020] For example, the standardized feature quantity representing the physiological state is calculated from the time series data of the biological state of a person in the morning and evening. Next, a displacement vector ab is calculated based on the standardized feature quantity in a multidimensional vector space. The calculated displacement vector is then used to calculate the displacement vector p of multiple environmental stimuli in the database. i q i When the displacement vector p i q i The combined ratio of these two physiological states can be used to express changes in the two states. [Effects of the Invention]

[0021] The present invention can provide a physiological state estimation system, a physiological state estimation method, and a physiological state estimation program that are capable of estimating various changes in the physiological state of a living body. [Brief explanation of the drawings]

[0022] [Figure 1] FIG. 1 is a diagram illustrating a configuration of a physiological state estimation device according to an embodiment. [Figure 2] FIG. 10 is a diagram showing an example of a database generation program for calculating a physiological state displacement vector. [Figure 3] FIG. 10 is a diagram illustrating an example of a physiological state estimation program. [Figure 4] FIG. 10 is a conceptual diagram showing an example of a displacement vector. [Figure 5] 10 is a flowchart showing an example of a database generation program for calculating a physiological state displacement vector. [Figure 6] 10 is a flowchart illustrating an example of a physiological state estimation program. [Figure 7] FIG. 10 is a diagram showing an example of a displacement vector due to a mental fatigue task. [Figure 8] FIG. 10 is a diagram showing an example of a displacement vector due to a physical fatigue task. [Figure 9] FIG. 10 is a diagram illustrating an example of a displacement vector obtained through a verification experiment. [Figure 10] FIG. 10 is a diagram showing the relationship between displacement vector groups obtained through a verification experiment. [Figure 11] FIG. 1 is a conceptual diagram showing information obtained by a Fourier transform and a Hilbert transform. [Figure 12] 1 is a conceptual diagram of a method for estimating a physiological state according to the present disclosure. [Figure 13] 12. FIG. 13 is a diagram showing the relationship between the displacement vector groups shown in FIG. 10, with the vector arrangement being the same as that shown in FIG. [Figure 14] FIG. 10 is a diagram showing the distribution of logarithmic instantaneous amplitudes for each frequency band of electroencephalograms. [Figure 15] FIG. 10 is a diagram showing the distribution of instantaneous frequencies for each frequency band of electroencephalograms. [Figure 16] FIG. 12 is an explanatory diagram of Equation 12 for calculating the x and y vectors of Equation 11. DETAILED DESCRIPTION OF THE INVENTION

[0023] FIG. 1 is a diagram showing the configuration of a physiological state estimation device 1 according to an embodiment. The physiological state estimation device 1 is one aspect of a physiological state estimation system. Specific examples of the physiological state estimation device 1 include a PC (Personal Computer), a server, a Raspberry Pi, a smartphone, a wearable device, a tablet device, and an in-vehicle device. These devices correspond to computers.

[0024] The physiological state estimation device 1 includes a calculation device 10, a communication interface (I / F) 11, a storage device 12, and a display device 13.

[0025] The arithmetic device 10 is a device that performs overall control of the physiological state estimation device 1. Specific examples of the arithmetic device 10 include a CPU (Central Processing Unit), an MPU (Micro Processing Unit), an ECU (Electronic Control Unit), etc. The arithmetic device 10 corresponds to a computer.

[0026] The arithmetic device 10 executes a database generation program 100 and a physiological state estimation program 200, which will be described later. The database generation program 100 and the physiological state estimation program 200 will be described in detail later. Note that the physiological state estimation program may be executed by a semiconductor device such as an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). These semiconductor devices also correspond to computers.

[0027] The communication interface (I / F) 11 is a device that exchanges data with an external device. The communication interface 11 can acquire biological information from a sensor that detects biological information that represents the physiological state of a living organism. The communication interface 11 may also acquire biological information via a network configured of a LAN (Local Area Network) and / or WAN (Wide Area Network). The biological information may also be input to the computing device via a wired connection with the biological information sensor via an analog-to-digital conversion device. Specific examples of sensors that detect biological information include pulse wave sensors, heart rate sensors, cerebral blood flow sensors such as near-infrared spectroscopy sensors, and fMRI that can observe blood flow in the brain. In other embodiments, the physiological state estimation device 1 may also include a sensor that detects biological information.

[0028] The storage device 12 is a storage device that stores various information processed by the arithmetic device 10, such as a database generation program, a physiological state estimation program, a database representing feature quantities for each physiological state, physiological measurement data, etc. The display device 13 is a device that displays various information such as information indicating the physiological state calculated by the arithmetic device 10. In other embodiments, the display device 13 may be a device separate from the physiological state estimation device 1.

[0029] 2 is a diagram showing an example of a database generation program 100 for generating a database including feature quantities indicating physiological states due to environmental stimuli and displacement vectors indicating changes in the physiological states. The database generation program 100 includes a feature quantity calculation unit 101 and a displacement vector calculation unit 102.

[0030] The feature calculation unit 101 calculates a feature indicating a physiological state for each environmental stimulus. Heart rate variability data is measured before and after a certain environmental stimulus, for example, a mental fatigue task (a 20-minute 3-Back task), for a time period considered to be a physiological steady state, for example, 10 minutes. Other environmental stimuli may include physical fatigue tasks such as climbing stairs or using a treadmill, a nap, a massage, staying in a space, or hot and cold stimulation. The feature calculation unit 101 performs bandpass filtering on the heart rate variability waveform in four frequency bands, for example, HF (0.15-0.4 Hz), LF (0.04-0.15 Hz), VLF1 (15-40 mHz), and VLF2 (4-15 mHz). Next, the feature calculation unit 101 performs a Hilbert transform on the waveforms in each frequency band obtained by the bandpass filtering process to convert them into complex numbers, thereby calculating the instantaneous amplitude and instantaneous frequency (see Equation 4 and Equation 6). When the distributions of instantaneous amplitude and instantaneous frequency are calculated in a section that can be considered physiologically steady, i.e., a section where the distribution is unimodal, the instantaneous amplitude and instantaneous frequency are distributed in a Gaussian-like manner. The feature calculation unit 101 approximates the distributions of instantaneous amplitude and instantaneous frequency with a normal distribution function to calculate the mean and standard deviation, resulting in 16 mathematically independent feature quantities (= 4 (frequency band) × 2 (instantaneous amplitude and instantaneous frequency) × 2 (mean and standard deviation)). Because the physiological state before multiple environmental stimuli can be considered the same, the data before and after the environmental stimuli are standardized by setting the mean value to 0 and the standard deviation to 1, calculated using the multiple feature quantities before the environmental stimuli stored in the database, and then registered in the database. This standardization is performed to ensure the accuracy of vector synthesis in the 16-dimensional data space.

[0031] The displacement vector calculation unit 102 uses the standardized feature values ​​calculated by the feature calculation unit 101 to calculate, for example, displacement vectors in a vector space before and after a mental fatigue task. For example, the heart rate interval measurement data of 30 people is measured before and after a mental fatigue task, and the standardization is performed to calculate a displacement vector. The average vector of the displacement vectors of the 30 people is then calculated and defined as a displacement vector representing a change in physiological state due to the mental fatigue task. This is then registered in a displacement vector database 30, which corresponds to a storage unit. If the coordinate change before and after the mental fatigue task is not statistically significant, the parameter can be increased until a statistically significant displacement vector is obtained. Alternatively, heart rate interval measurement data of a specific person before and after an environmental stimulus may be obtained multiple times, and the displacement vectors may be similarly registered in the database.

[0032] The above example shows a case where the change in physiological state due to environmental stimuli is not very large. However, if the change in physiological state due to environmental stimuli is very large, accurate vector synthesis becomes impossible. In other words, the variance of the standardized features becomes large, and the contribution rate of features with large variance to the vector synthesis becomes large, making accurate physiological state estimation impossible. This process is performed to ensure orthogonality between vectors representing the features. In such a case, an eigenvalue vector matrix is ​​calculated using the standardized features before the environmental stimulus in the database, and a Cartesian coordinate system is calculated by calculating the matrix product of the eigenvalue vector matrix and the standardized features before and after the environmental stimulus, thereby expressing the original feature data in the Cartesian coordinate system. Next, the converted displacement vectors are calculated, and the average displacement vectors are calculated for each environmental stimulus and registered in the displacement vector database 30. Methods for this conversion to a Cartesian coordinate system include PCA and sweep-out algorithms.

[0033] Figure 4 shows a three-dimensional simulated representation of the 16-dimensional physiological state vector space. The distributions of the features before and after the environmental stimulus in the 16-dimensional vector space are designated P and Q, respectively. The 16-dimensional features of P are used to perform the standardization or orthogonalization operation described above to determine the Cartesian coordinate axes C1, C2, C3, ..., C16. Next, displacement vectors from coordinates before the environmental stimulus to coordinates after the environmental stimulus are calculated in the Cartesian coordinate space, and an average displacement vector is calculated for each environmental stimulus. This is defined as the displacement vector from the P group before the environmental stimulus to the Q group after the environmental stimulus and registered in the displacement vector database 30. By repeating this operation for each environmental stimulus, the displacement vector database 30 can be constructed.

[0034] Using 10 minutes of heart rate interval measurement data before and after a mental fatigue task (20 minutes of 3-Back task) and a physical fatigue task (climbing stairs 5 times twice), standardized features were calculated and displacement vectors were calculated. The results are shown in Figures 7 and 8, respectively. Here, the horizontal axes mh, sh, ml, sl, mv, sv, mw, sw, mhf, shf, mlf, slf, mvf, svf, mwf, and swf represent displacement vector components of the standardized features. The letter m represents the mean, s represents the standard deviation, h represents the HF band, l represents the LF band, v represents the VLF1 band, and w represents the VLF2 band. The two letters represent amplitude-related features. The three letters followed by f represent frequency-related features.

[0035] 3 is a diagram showing an example of the physiological state estimation program 200. The physiological state estimation program 200 includes a feature amount calculation unit 201, a displacement vector calculation unit 202, and a physiological state estimation unit 203.

[0036] The feature calculation unit 201 is a program that calculates the standardized feature from biological information based on any different physiological state, similar to the above-mentioned feature calculation unit 101. The sensor that acquires the biological information may be any sensor that can acquire a signal containing information about the physiological state, such as a pulse wave sensor, an electrocardiogram sensor, an electroencephalogram sensor, a cerebral blood flow sensor, fMRI, or an infrared image sensor.

[0037] The displacement vector calculation unit 202 is a program that calculates a displacement vector based on the feature amounts calculated by the feature amount calculation unit 201. Based on the standardized first feature amount and second feature amount, mapping to a 16-dimensional physiological state space is performed, and a displacement vector from the first space coordinates to the second space coordinates can be calculated.

[0038] The physiological state estimation unit 203 is a program that estimates a physiological state using the displacement vector calculated by the displacement vector calculation unit 202 and the displacement vector stored in the displacement vector database 30. Specifically, when the displacement vector calculated by the displacement vector calculation unit 202 indicates a displacement from coordinate a to coordinate b of the displacement vector ab in a 16-dimensional vector space, the physiological state estimation unit 203 estimates a physiological state using the displacement vectors D ij (i=1 to 16 (number of features), j=1 to n (number of types of environmental stimuli)) is used to calculate the vector ab. The vector ab can be expressed by a linear combination formula in Equation 1.

number

[0039] The physiological state estimation unit 203 substitutes Equation 2 into Equation 3 and solves the n-dimensional linear simultaneous equations based on the partial differentials, thereby obtaining the optimal K j In this way, any physiological state can be calculated by the coefficient K j Then, by solving the minimization problem of Equation 2, K j For example, by solving Equation 3, K j can be obtained.

number

number

[0040] In the above-described embodiment, a 16-dimensional vector space was constructed for the four frequency bands HF, LF, VLF1, and VLF2 to determine displacement vectors. However, in other embodiments, the multidimensional vector space may be expanded by adopting frequency bands lower than VLF2, such as ULF1 (1.5-4 mHz), ULF2 (0.4-1.5 mHz), UULF1 (0.15-0.4 mHz), UULF2 (0.04-0.15 mHz), etc. Furthermore, for physiological measurement data such as carotid pulse waves, cerebral blood flow waves, and fMRI signals, the multidimensional vector space may be expanded by adopting frequency bands of electroencephalograms, such as δ (0.4-4 Hz), θ (4-8 Hz), α (8-13 Hz), β (13-30 Hz), and γ (30 Hz or higher). Furthermore, the frequency band of electroencephalograms may be divided into multiple sections.

[0041] When EEG signals are divided into the above-mentioned δ, θ, α, β, and γ frequency bands and the waveforms of each band are subjected to a Hilbert transform to calculate the instantaneous amplitude and instantaneous frequency, their distributions are as shown in Figures 14 and 15. Both the logarithmic instantaneous amplitude and instantaneous frequency in the δ, θ, α, β, and γ frequency bands are distributed in a Gaussian manner, and when these distributions are approximated with a Gaussian function and the mean and standard deviation are calculated, 5 (frequency bands) x 2 (logarithmic instantaneous amplitude, instantaneous frequency) x 2 (mean, standard deviation) = 20 feature quantities are obtained, making it possible to estimate emotions and mental states.

[0042] Furthermore, by using a pulse wave signal, it is possible to define the phase difference feature from two types of waveform series: the pulse interval waveform and the pulse waveform. Specifically, a vector space equivalent to the heartbeat interval data is formed from the pulse interval data. On the other hand, the pulse waveform data reveals changes in blood output and the state of tension and relaxation of the muscles around the blood vessels. These changes are observed as the phase difference between the pulse interval and the pulse waveform (see Patent Document 2).

[0043] The pulse interval waveform is Φ(t), and the instantaneous amplitude is A k、 Let the instantaneous phase be Ψ k (t) Instantaneous frequency in Ω k(t), while the pulse waveform is φ(t) and the instantaneous amplitude is a k , the instantaneous phase is ψ k (t), and the instantaneous frequency is ω k (t), instantaneous phase difference θ k , k=HF, LF, VLF1, VLF2, the pulse interval waveform, pulse waveform, instantaneous frequency, and instantaneous phase difference can be expressed by Equations 4 to 8.

number

number

number

number

number

[0044] In the physiological steady state, the instantaneous amplitude A k ,a k , instantaneous frequency Ω k (t),ω k (t) is distributed like a Gaussian, and the instantaneous phase difference θ k (t) is distributed like a Von Mizese distribution as an angular distribution. Feature quantities corresponding to the mean and variance can be calculated from these distributions, resulting in 24 feature quantities (= 4 (frequency band) × 3 (instantaneous amplitude, instantaneous frequency, and instantaneous phase difference) × 2 (mean and variance)), allowing the construction of a 24-dimensional vector space. The 8-dimensional feature quantities added by Equation 8 represent information on changes in cardiac output and the state of tension and relaxation of muscles around blood vessels, and can be important physiological feature quantities. Therefore, by expressing the physiological state in a 24-dimensional vector space including the instantaneous phase difference mentioned above, it is possible to capture changes in the physiological state in more detail than when using only information on the heartbeat interval.

[0045] Furthermore, similar data processing may be performed using physiological measurement data of electroencephalograms and cerebral blood flow (NIRS, etc.), and similar feature quantities may be calculated from their amplitudes, frequencies, and phase differences to form a multidimensional vector space.

[0046] The significance of applying the Hilbert transform to physiological state signals is described below. LF / HF is usually used to evaluate the autonomic nervous system. When a person is relaxed, respiratory heart rate variability becomes dominant. Therefore, by Fourier transforming the time series data of heart rate intervals, it is possible to determine whether the state of tension or relaxation is high or low based on the magnitude of the power spectrum ratio between respiratory heart rate variability HF (0.15-0.4 Hz) and blood pressure heart rate variability LF (0.04-0.15 Hz). The heart rate interval waveform for each frequency band is expressed as φ k (t), k=HF, LF, VLF1, VLF2..., the signal F after the Fourier transform is expressed by Equation 9. On the other hand, the signal H after the Hilbert transform and complex numbering is expressed by Equation 10.

number

number

[0047] Φ k The Hilbert transform of Φ gives us a complex time function that is analyzed. k When the results of the Hilbert transform are plotted on a graph with the real part on the horizontal axis and the time derivative of the imaginary part on the vertical axis, the result is as shown in Figure 11. When the frequency band waveform of the heartbeat interval k (k=HF, LF, VLE1, VLF2) is subjected to a Hilbert transform, it becomes a complex time function, and in the physiological steady state, the real term corresponding to the instantaneous amplitude and the instantaneous frequency (time derivative of the imaginary term) are both distributed in a Gaussian manner, and are distributed in an elliptical shape on the complex plane.

[0048] Since instantaneous amplitude and instantaneous frequency are distributed in a Gaussian manner in physiological steady states, these distributions are approximated by normal distributions to determine the mean and variance. The amplitude and frequency of the distribution of instantaneous amplitude and instantaneous frequency, and the mean and variance are mathematically independent. By employing these features, a wide range of physiological states can be expressed.

[0049] On the other hand, in the case of a Fourier transform, the time integral of the instantaneous phase term loses its physical meaning, and only the time average value of the amplitude can be extracted as a physical quantity. However, the centroid frequency can be obtained by time integrating the product of the amplitude term and frequency and dividing by the time integral value of the amplitude term. It is also possible to estimate a similar physiological state using the 8-dimensional (= 4 frequency bands × 2) feature obtained in this way. However, because the phase information of each frequency band is lost, the range of physiological state estimation is thought to be limited.

[0050] In the above-described embodiment, the physiological state is represented using 16-dimensional feature quantities, namely, the mean and standard deviation of instantaneous amplitude and the mean and standard deviation of instantaneous frequency for the four frequency bands LF, HF, VLF1, and VLF2. However, in other embodiments, the physiological state may be represented using feature quantities with more than 16 dimensions, using a frequency band lower than VLF2 or a frequency band higher than HF related to brain activity.

[0051] FIG. 5 is a flowchart showing an example of processing that the physiological state estimating device 1 executes based on the database generating program 100.

[0052] In step S10, multiple sets of physiological state measurement data before and after a specific environmental stimulus are input to the physiological state estimation device 1 to build a database. In step S11, the physiological state measurement data is band-pass filtered in a desired frequency band, then Hilbert transformed and analyzed to calculate instantaneous amplitude and instantaneous frequency. Each distribution is approximated with a Gaussian function to calculate the mean value and standard deviation, thereby calculating N (N=16 in the case of heartbeat interval data) physiological state feature quantities. In step S12, it is determined whether the changes in the physiological state feature quantities calculated in step S11 before and after the environmental stimulus are statistically significant. Specifically, it is determined whether the p-value of a t-test or the like for the change in the feature quantity before and after the environmental stimulus is equal to or less than a predetermined threshold (e.g., 0.05). If it is determined that the change before and after the environmental stimulus is not significant (NO), the process returns to step S10, and additional biological information is input. If it is determined that the change before and after the environmental stimulus is significant (YES), the process proceeds to step S13.

[0053] In step S13, the data before and after the environmental stimulus are standardized using the mean value and standard deviation calculated from the physiological state features before the environmental stimulus that are already stored in the displacement vector database, with the mean value set to 0 and the standard deviation set to 1.

[0054] In step S14, the variance of each dimension of the standardized N-dimensional feature before and after the environmental stimulus is calculated, and the process branches depending on the result. If it is determined that the variance of each dimension is small (YES), in other words, if it is determined that the variance of each dimension is equal to or smaller than a predetermined threshold, for example, if (maximum variance - minimum variance) / maximum variance≦0.5, the process proceeds to step S15. On the other hand, if it is determined that the variance of each dimension is large (NO), in other words, if it is determined that the variance of each dimension is larger than a predetermined threshold, for example, if (maximum variance - minimum variance) / maximum variance>0.5, the process proceeds to step S17.

[0055] In step S15, in the N-dimensional spatial coordinates consisting of the standardized feature amounts before and after the environmental stimulus calculated in step S13, a displacement vector from before the environmental stimulus to after the environmental stimulus is calculated for each data, and an average displacement vector for each environmental stimulus is calculated. In step S16, the average displacement vector for each environmental stimulus is registered in database 30, and the processing of FIG. 5 ends.

[0056] In step S17, the standardized feature quantities before and after the environmental stimulus are converted into a Cartesian coordinate system. In step S18, a displacement vector from before the environmental stimulus to after the environmental stimulus is calculated for each data, and an average displacement vector for each environmental stimulus is calculated. In step S19, the average displacement vector for each environmental stimulus is registered in database 30, and the process of FIG. 5 ends.

[0057] When new environmental stimulus data is added to the database, or when data is added to existing environmental stimuli, the processes from steps S13 to S16 or S13 to S19 must be executed for all data in the database. To do this, the feature data before standardization and the feature data after standardization must be registered in the displacement vector database 30 for each environmental stimulus, along with the displacement vector.

[0058] Here, since the displacement vectors registered in step S16 and the displacement vectors registered in step S19 are in completely different data spaces, when estimating a physiological state, it is necessary to distinguish which physiological state space to use for estimation. If the environmental stimuli used to estimate the physiological state do not include the environmental stimuli registered in step S19, estimation can be performed in the physiological state space registered in step S16, but if the environmental stimuli registered in step S19 are included, estimation must be performed in the physiological state space registered in step S19.

[0059] The correspondence between Fig. 2 and Fig. 5 will be described. The feature amount calculation unit 101 in Fig. 2 executes the processes of steps S10 to S13 in Fig. 5. The displacement vector calculation unit executes the processes of steps S14 to S19.

[0060] Examples of displacement vectors calculated for each environmental stimulus are shown in Figures 7 and 8. Figure 7 shows displacement vectors based on 16-dimensional features calculated using heart rate interval data obtained by performing 10-minute electrocardiogram measurements on 37 subjects before and after a fatigue task (20 minutes of 3-Back task). This is tentatively defined as mental fatigue. Figure 8 shows displacement vectors obtained by performing 10-minute electrocardiogram measurements on 46 subjects before and after a fatigue task (climbing up and down five flights of stairs twice). This is tentatively defined as physical fatigue. The horizontal axes mh, sh, ml, sl, mv, sv, mw, sw, mhf, shf, mlf, slf, mvf, svf, mwf, and swf in Figures 7 and 8 indicate the names of the displacement vector components of the standardized features. The letters m represent the mean, s represent the standard deviation, h represent the HF band, l represent the LF band, v represent the VLF1 band, and w represent the VLF2 band. The two letters represent features related to amplitude. The three letters with "f" represent features related to frequency. The vertical axis shows the deviation of the features normalized with the standard deviation set to 1.

[0061] FIG. 6 is a flowchart showing an example of processing executed by the physiological state estimation device 1 based on the physiological state estimation program 200.

[0062] In step S30, physiological state measurement data indicating a first physiological state and a second physiological state are input to the physiological state estimation device 1. In step S31, the above-mentioned feature quantities are calculated from the first and second physiological state measurement data. In step S32, the feature quantities of the first physiological state and the second physiological state calculated in step S31 are standardized using the feature quantities of the physiological states before the environmental stimuli stored in the displacement vector database 30. In step S33, the first feature quantity and the second feature quantity are mapped to an N-dimensional space, and a displacement vector indicating a spatial coordinate displacement from the first spatial coordinate to the second spatial coordinate is calculated.

[0063] In step S34, the displacement vector calculated in step S33 is expressed as a linear combination of multiple known displacement vectors of environmental stimuli stored in the displacement vector database 30, and the combination coefficient is solved as a mathematical optimization problem. In step S35, a linear combination ratio based on the calculated combination coefficient is output as a physiological state estimate, and the processing of FIG. 6 ends. By adopting the linear combination ratio calculated based on such highly accurate displacement vectors as a physiological state estimate, the reliability of the physiological state estimate can be improved. Corresponding to FIG. 3, steps S31 and S32 are executed by the feature calculation unit 201, step S33 is executed by the displacement vector calculation unit 202, and steps S34 and S35 are executed by the physiological state estimation unit 203.

[0064] Specifically, suppose that heartbeat interval measurement data of a person is measured in the morning and evening, and the displacement vector of the physiological state from morning to evening is ab. Then, the displacement vector of mental fatigue stored in the displacement vector database 30 is set to S, and the displacement vector of physical fatigue is set to R, and mathematical optimization is performed using Equation 1, Equation 2, and Equation 3 to find the coefficient K of the linear combination. j Ask for.

number

[0065] In Equation 11, a person's fatigue level from morning to evening is expressed by mental fatigue K1 and physical fatigue K2. If the residual ε can be expressed as a displacement vector of other environmental stimuli, mathematical optimization can be performed by adding the displacement vector T of the physiological state corresponding to the environmental stimuli and coefficient K3, and the physiological state can be estimated based on Equation 12 by adding a physiological state vector T other than mental fatigue and physical fatigue to the mental fatigue and physical fatigue.

number

[0066] The above calculation is explained using the vector diagram in Figure 12. In the above standardized 16-dimensional Cartesian coordinate system, when the physiological state A before environmental stimulation changes to physiological state B after an arbitrary environmental stimulation, the position coordinates of these states are a and b, respectively. When the mental fatigue vector p1q1 and the physical fatigue vector p2q2 are used to estimate the physiological state change vector ab, the ab' vector obtained by projecting the ab vector onto a plane containing the position coordinates p1, q1, p2, and q2 is synthesized using the p1q1 vector and the p2q2 vector. The synthesis ratio is the estimated value of the physiological state. In this case, an error occurs between b and b', and this error corresponds to the residual ε.

[0067] A specific example of the physiological state estimation method described above is described below. For a given subject, the state before environmental stimulation (r1) was designated as r1. Then, a 3-Back task was performed for 20 minutes. Then, the state was designated as r2. The state was then subjected to 10 minutes of ECG measurement while resting. Then, the subject climbed up and down five flights of stairs twice. Then, the state was designated as r3. The state was subjected to 10 minutes of ECG measurement while resting. The 16-dimensional feature values ​​from the data r1 through r3 were standardized using the 16-dimensional feature values ​​before environmental stimulation in the database as the population. Figure 9 shows the results of calculating the displacement vector r13 from r1 to r3. Here, the horizontal axis indicates the displacement vector components of the standardized feature values. The letters m, s, ml, sl, mv, sv, mw, sw, mhf, shf, mlf, slf, mvf, svf, mwf, and swf represent the mean, standard deviation, h, HF band, l, LF band, v, VLF1 band, and w, respectively. The two letters represent amplitude. The three letters followed by "f" represent frequency. The vertical axis shows the amount of change in the feature, standardized with the standard deviation set to 1.

[0068] Using the method described above, when the displacement vector in Figure 9 is combined with the displacement vectors of mental fatigue in Figure 7 and physical fatigue in Figure 8, K1 = 2.4 (mental fatigue) and K2 = 1.1 (physical fatigue), making it possible to quantitatively estimate the ratio of mental fatigue to physical fatigue for a given subject. If the displacement vectors of environmental stimuli other than the displacement vectors of physical fatigue and mental fatigue are known, the ratios of other physiological states can also be estimated by performing similar mathematical optimization.

[0069] The lengths of the vectors in 16-dimensional space and the angles between the two vectors can be grasped, and their relationships are shown in FIGS. 10 and 13. Specifically, FIG. 10 shows the relationship between the position vectors r1 to r3, the mental fatigue vector h1, and the physical fatigue vector h2. FIG. 10 also shows the angles formed by each vector. The angle between h1 and h2 is 112.5°. The angle between h1 and vector r12 is 73.9°. The angle between h2 and vector r23 is 73.0°. The angle between h1 and vector r13 is 70.9°. The angle between vectors h2 and r13 is 81.2°. FIG. 13 shows the relationship between the displacement vectors shown in FIG. 10, but with the same vector arrangement as FIG. 12.

[0070] Looking at these vector diagrams, we can see that there is a large error from the plane subtended by the r13 vector and the vectors of mental and physical fatigue, suggesting that physiological state displacements are due to causes other than those associated with mental and physical fatigue. By examining the displacement vectors of other environmental stimuli, such as rest, sleep, massage, heat, cold, vibration, and noise, it is hoped that it will be possible to synthesize components that do not contribute to vector synthesis.

[0071] In the above embodiment, physiological state estimation was performed based on mental fatigue tasks and physical fatigue tasks. For the physical fatigue task, a resting measurement was performed followed by exercise and then another resting measurement. Because heart rate increases immediately after exercise, features were calculated using data from 10 minutes after the heart rate returned to normal after 5 minutes of rest. Therefore, the physical fatigue displacement vector is a vector that combines the effects of physical fatigue and fatigue recovery (15 minutes of rest). The displacement vector component of the fatigue recovery effect introduces errors into the estimated value. In Figure 10, the angle between the mental fatigue and physical fatigue displacement vectors is 112.5°. However, if the displacement vector component corresponding to the fatigue recovery effect is removed, the angle becomes orthogonal (90°), potentially making the two independent. Below, we explain how to calculate the physical fatigue displacement vector after removing the fatigue recovery effect.

[0072] Because the mental fatigue displacement vector's features are calculated using resting measurement data immediately after the task, it is believed that factors other than mental fatigue are hardly mixed in. Possible candidates for such environmental stimuli are hot and cold stimuli. Although such experiments have not yet been conducted, let's assume that the thermal stimulus displacement vector can be calculated. Here, let's assume that the mental fatigue vector and the thermal stimulus vector are orthogonal. If a 16-dimensional Cartesian coordinate system can be calculated with the mental fatigue displacement vector and the thermal stimulus displacement vector as physiological basis vectors, then the basis vector that is closest to the physical fatigue vector, i.e., the vector with the smallest angle between the displacement vectors, can be used as the physical fatigue vector, which is thought to be able to eliminate the displacement vector component of the fatigue recovery effect.

[0073] Specifically, the 16-dimensional orthogonal coordinate system used in the calculation in step S15 of Fig. 5 is orthonormalized by setting the length to 1. Then, a rotational coordinate transformation is performed on the mutually orthogonal mental fatigue vector and thermal stimulus vector so that their directions coincide with two axes in the orthogonalized orthogonal coordinate system. This method can be achieved by using Equation 13, as described in Non-Patent Document 1.

number

[0074] For example, suppose the angle between mental fatigue basis vector b of length 1 and one of the basis vectors a of the 16-dimensional normal orthogonal coordinate system is δ. In Figure 16, if vector b is at an angle with vector a and δ, and vector c of length 1 that is perpendicular to a on the plane spanned by vectors a and b is found, then we obtain Equation 14.

number

[0075] Next, we consider using the mental fatigue vector as the axis to rotate the thermal stimulus vector, orthogonal to it, to align its direction with that of one vector in an orthonormal system. Specifically, we define the thermal stimulus vector orthogonal to the mental fatigue vector as b, one of the 15 orthonormal system vectors with a different direction from the mental fatigue vector as a, and the angle between it and the thermal stimulus vector as δ, and then perform the same process. Because vectors a and b are orthogonal to the mental fatigue vector, a rotational coordinate transformation is performed around the mental fatigue vector axis, resulting in a 16-dimensional basis vector space containing the basis vectors for the mental fatigue vector and the thermal stimulus vector. If we define the physical fatigue vector as the basis vector of the 14 basis vectors other than the mental fatigue vector and the thermal stimulus vector, which has the smallest angle with the physical fatigue vector, it may be possible to extract the pure physical fatigue vector component after a physical fatigue task. Future experiments with other environmental stimuli will allow us to define those with the smallest angle with the displacement vector as the basis vector for each environmental stimulus. We have confirmed that the orthogonality of the orthonormal system is maintained by the above rotational coordinate transformation.

[0076] The above explanation assumes that the mental fatigue vector and the thermal stimulus vector are physiological basis vectors, but they may not be physiological basis vectors. In the future, it may be possible to calculate displacement vectors of various environmental stimuli and define better physiological basis vectors.

[0077] Once the physiological basis vectors are obtained, it is possible to express a displacement vector due to any change in physiological state with better basis vectors, without the need to solve the simultaneous equations consisting of Equations 1 to 3. Therefore, the ratio of each physiological state can be estimated simply by calculating the dot product with the basis vectors.

[0078] In the above examples, the program includes instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more functions described in the embodiments. The program may be stored on a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, computer-readable medium or tangible storage medium includes random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technology, CD-ROM, digital versatile disk (DVD), Blu-ray® disk or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The program may also be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, transitory computer-readable medium or communication medium includes electrical, optical, acoustic, or other forms of propagated signals.

[0079] The present invention is not limited to the above-described embodiment, and can be modified as appropriate within the scope of the invention.

[0080] <Additional Notes> (Appendix 1) a feature calculation unit that divides a time-series signal representing a physiological state of a living organism into a plurality of frequency bands and calculates a feature from time-series fluctuations of amplitude and frequency in each frequency band of the time-series signal; a displacement vector calculation unit that represents the physiological state of the living body in a multidimensional vector space made up of the calculated feature amounts and calculates a displacement vector that indicates a change in the physiological state of the living body in the multidimensional vector space; a storage unit that stores the displacement vectors before and after a plurality of environmental stimuli; a physiological state estimation unit that uses displacement vectors before and after a desired environmental stimulus stored in the storage unit to synthesize the displacement vectors before and after an arbitrary physiological state change, and estimates the physiological state of the living body based on a synthesis ratio of the synthesized displacement vectors; A physiological state estimation system comprising: (Appendix 2) The physiological state estimation system described in Appendix 1, wherein the displacement vector stored in the memory unit is calculated by determining position coordinates in the multidimensional vector space of features based on time series signals representing physiological states before and after multiple types of environmental stimuli for multiple people or multiple times, and averaging, for each environmental stimulus, displacement vectors from before the environmental stimulus to after the environmental stimulus when changes in the physiological state of the living organism before and after the environmental stimulus become statistically significant. (Appendix 3) The physiological state estimation system described in Appendix 1, wherein the feature calculation unit calculates an average value and a standard deviation using multiple feature values ​​before the environmental stimulus stored in the memory unit, and standardizes the data before and after the environmental stimulus using the calculated average value and standard deviation, with the average value set to 0 and the standard deviation set to 1. (Appendix 4) the feature calculation unit calculates a first feature, which is a standardized feature before an environmental stimulus, using a time-series signal representing a first physiological state; the feature calculation unit calculates a second feature, which is a standardized feature after the environmental stimulus, using the time-series signal representing the second physiological state; the displacement vector calculation unit calculates a displacement vector indicating a change from the first physiological state to the second physiological state in the multidimensional vector space based on the first feature amount and the second feature amount; The physiological state estimation system described in Appendix 3, wherein the physiological state estimation unit represents a displacement vector indicating a change from the first physiological state to the second physiological state as a linear combination using the displacement vectors stored in the memory unit, and outputs a ratio of combination coefficients of the linear combination as an estimated value indicating a change in the physiological state of the living body. (Appendix 5) the feature calculation unit calculates an instantaneous amplitude and an instantaneous frequency of a signal waveform representing a physiological state in a certain frequency band, and calculates, as the feature, an average value and a standard deviation obtained by aggregating the calculated instantaneous amplitude and instantaneous frequency in a section regarded as a physiological steady state. (Appendix 6) the frequency band represents a physiological state of the living body, The physiological state estimation system described in Appendix 5, wherein the feature calculation unit calculates the feature for at least four frequency bands: HF (0.15 to 0.4 Hz) representing the respiratory cycle of the living body, LF (0.04 to 0.15 Hz) representing the blood pressure fluctuation cycle of the living body, and VLF1 (15 to 40 mHz) and VLF2 (4 to 15 mHz) representing fluctuations in the autonomic nervous system of the living body. (Appendix 7) The frequency band includes ULF1 (0.15 to 0.4 mHz) and ULF2 (0.04 to 0.15 mHz), 7. The physiological state estimation system according to claim 6, wherein the feature amount calculation unit calculates the feature amount for at least one of the ULF1 and the ULF2. (Appendix 8) The frequency bands include frequency bands δ (0.4 to 4 Hz), θ (4 to 8 Hz), α (8 to 13 Hz), β (13 to 30 Hz), and γ (30 Hz or higher); 7. The physiological state estimation system according to claim 6, wherein the feature amount calculation unit calculates the feature amount for at least one of the δ, the θ, the α, the β, and the γ. (Appendix 9) The physiological state estimation system described in Appendix 3, characterized in that when the variance variation of the standardized feature before and after the environmental stimulus stored in the memory unit is larger than a predetermined threshold, the displacement vector calculation unit converts the feature space into an orthogonal space and calculates the displacement vector. (Appendix 10) When there are two orthogonal displacement vectors due to environmental stimuli, The physiological state estimation system according to any one of appendices 1 to 3, wherein the physiological state estimation unit calculates basis vectors by rotationally transforming the feature space so that directions of any two vectors in a multidimensional eigenvalue vector matrix obtained by transforming a feature space before a plurality of environmental stimuli into an orthogonal space coincide with directions of the two orthogonal displacement vectors, and estimates the physiological state using the matrix of the calculated basis vectors and an inner product value of each basis vector with a displacement vector indicating a change from a first physiological state to a second physiological state. (Appendix 11) The feature amount calculation unit Dividing each of the time series signals of the pulse interval and blood flow volume acquired from the living body into a plurality of frequency bands; For each frequency band, a distribution of instantaneous amplitudes and instantaneous frequencies of the time series signals of the pulse intervals and the blood flow volume, and a distribution of instantaneous phase differences between the time series signals of the pulse intervals and the blood flow volume are calculated; 5. The physiological state estimation system according to claim 4, wherein the mean, variance, or concentration of each of the calculated distributions is calculated as a feature. (Appendix 12) A computer-implemented physiological state estimation method, comprising: Dividing a time-series signal representing a physiological state of a living organism into a plurality of frequency bands, and calculating a feature amount from time-series fluctuations of amplitude and frequency in each frequency band of the time-series signal; expressing the physiological state of the living body in a multidimensional vector space composed of the calculated feature quantities, and calculating a displacement vector indicating a change in the physiological state of the living body in the multidimensional vector space; synthesizing the displacement vectors before and after an arbitrary physiological state change using the displacement vectors before and after a desired environmental stimulus stored in a storage unit that stores the displacement vectors before and after a plurality of environmental stimuli, and estimating the physiological state of the living body based on a synthesis ratio of the synthesized displacement vectors. Physiological state estimation method. (Appendix 13) A physiological state estimation program executed by a computer, the program including: a step of dividing a time-series signal representing a physiological state of a living organism into a plurality of frequency bands, and calculating a feature amount from time-series fluctuations of amplitude and frequency in each frequency band of the time-series signal; expressing the physiological state of the living body in a multidimensional vector space consisting of the calculated feature amounts, and calculating a displacement vector indicating a change in the physiological state of the living body in the multidimensional vector space; a step of synthesizing the displacement vectors before and after an arbitrary physiological state change using the displacement vectors before and after a desired environmental stimulus stored in a storage unit that stores the displacement vectors before and after a plurality of environmental stimuli, and estimating the physiological state of the living body based on a synthesis ratio of the synthesized displacement vectors; A physiological state estimation program that executes the above. [Explanation of symbols]

[0081] 1. Physiological state estimation device, physiological state estimation device 10 Arithmetic unit 11 Communication Interface 12 Storage device 13 Display device 100 Database Generator 101 Feature calculation unit 102 Displacement vector calculation unit 200 Physiological state estimation program 201 Feature calculation unit 202 Displacement vector calculation unit 203 Physiological State Estimation Unit 30 Displacement vector database, memory section

Claims

1. a feature calculation unit that divides a time-series signal representing a physiological state of a living organism into a plurality of frequency bands and calculates a feature from time-series fluctuations of amplitude and frequency in each frequency band of the time-series signal; a displacement vector calculation unit that represents the physiological state of the living body in a multidimensional vector space made up of the calculated feature amounts and calculates a displacement vector that indicates a change in the physiological state of the living body in the multidimensional vector space; a storage unit that stores the displacement vectors before and after a plurality of environmental stimuli; a physiological state estimation unit that uses displacement vectors before and after a desired environmental stimulus stored in the storage unit to synthesize the displacement vectors before and after an arbitrary physiological state change, and estimates the physiological state of the living body based on a synthesis ratio of the synthesized displacement vectors; A physiological state estimation system comprising:

2. 2. The physiological state estimation system of claim 1, wherein the displacement vectors stored in the memory unit are calculated by determining position coordinates in the multidimensional vector space of features based on time series signals representing physiological states before and after multiple types of environmental stimuli for multiple people or multiple times, and averaging, for each environmental stimulus, displacement vectors from before the environmental stimulus to after the environmental stimulus when changes in the physiological state of the living organism before and after the environmental stimulus become statistically significant.

3. 2. The physiological state estimation system according to claim 1, wherein the feature calculation unit calculates an average value and a standard deviation using the plurality of feature values ​​before the environmental stimulus stored in the memory unit, and standardizes the data before and after the environmental stimulus by using the calculated average value and standard deviation, with the average value set to 0 and the standard deviation set to 1.

4. the feature calculation unit calculates a first feature, which is a standardized feature before an environmental stimulus, using a time-series signal representing a first physiological state; the feature calculation unit calculates a second feature, which is a standardized feature after the environmental stimulus, using the time-series signal representing the second physiological state; the displacement vector calculation unit calculates a displacement vector indicating a change from the first physiological state to the second physiological state in the multidimensional vector space based on the first feature amount and the second feature amount; 4. The physiological state estimation system according to claim 3, wherein the physiological state estimation unit represents a displacement vector indicating a change from the first physiological state to the second physiological state as a linear combination using the displacement vectors stored in the memory unit, and outputs a ratio of combination coefficients of the linear combination as an estimated value indicating a change in the physiological state of the living body.

5. 5. The physiological state estimation system according to claim 4, wherein the feature calculation unit calculates an instantaneous amplitude and an instantaneous frequency of a signal waveform representing a physiological state in a certain frequency band, and calculates, as the feature, an average value and a standard deviation obtained by aggregating the calculated instantaneous amplitude and the instantaneous frequency in a section regarded as a physiological steady state.

6. the frequency band represents a physiological state of the living body, 6. The physiological state estimation system according to claim 5, wherein the feature calculation unit calculates the feature for at least four frequency bands: HF (0.15 to 0.4 Hz) representing a respiratory cycle of the living body; LF (0.04 to 0.15 Hz) representing a blood pressure fluctuation cycle of the living body; and VLF1 (15 to 40 mHz) and VLF2 (4 to 15 mHz) representing fluctuations in the autonomic nervous system of the living body.

7. The frequency bands include ULF1 (0.15-0.4 mHz) and ULF2 (0.04-0.15 mHz), The physiological state estimation system according to claim 6 , wherein the feature amount calculation unit calculates the feature amount for at least one of the ULF1 and the ULF2.

8. The frequency bands include frequency bands δ (0.4-4 Hz), θ (4-8 Hz), α (8-13 Hz), β (13-30 Hz), and γ (30 Hz or higher); The physiological state estimation system according to claim 6 , wherein the feature amount calculation unit calculates the feature amount for at least one of the δ, the θ, the α, the β, and the γ.

9. 4. The physiological state estimation system according to claim 3, wherein when the variance variation of the standardized feature values ​​before and after the environmental stimulus stored in the memory unit is greater than a predetermined threshold, the displacement vector calculation unit converts the feature value space into an orthogonal space to calculate the displacement vector.

10. When there are two orthogonal displacement vectors due to environmental stimuli, 4. The physiological state estimation system according to claim 1, wherein the physiological state estimation unit calculates basis vectors by rotationally transforming the feature space so that directions of any two vectors in a multidimensional eigenvalue vector matrix obtained by transforming a feature space before a plurality of environmental stimuli into an orthogonal space coincide with directions of the two orthogonal displacement vectors, and estimates the physiological state using the matrix of the calculated basis vectors and an inner product value of a displacement vector indicating a change from a first physiological state to a second physiological state with each basis vector.

11. The feature amount calculation unit Dividing each of the time series signals of the pulse interval and blood flow volume acquired from the living body into a plurality of frequency bands; For each frequency band, a distribution of instantaneous amplitudes and instantaneous frequencies of the time series signals of the pulse intervals and the blood flow volume, and a distribution of instantaneous phase differences between the time series signals of the pulse intervals and the blood flow volume are calculated; The physiological state estimation system according to claim 4 , wherein the mean, the variance, or the concentration of each of the calculated distributions is calculated as a feature.

12. A computer-implemented physiological state estimation method, comprising: Dividing a time-series signal representing a physiological state of a living organism into a plurality of frequency bands, and calculating a feature amount from time-series fluctuations of amplitude and frequency in each frequency band of the time-series signal; expressing the physiological state of the living body in a multidimensional vector space composed of the calculated feature quantities, and calculating a displacement vector indicating a change in the physiological state of the living body in the multidimensional vector space; synthesizing the displacement vectors before and after an arbitrary physiological state change using the displacement vectors before and after a desired environmental stimulus stored in a storage unit that stores the displacement vectors before and after a plurality of environmental stimuli, and estimating the physiological state of the living body based on a synthesis ratio of the synthesized displacement vectors. Physiological state estimation method.

13. A physiological state estimation program executed by a computer, the program including: a step of dividing a time-series signal representing a physiological state of a living organism into a plurality of frequency bands, and calculating a feature amount from time-series fluctuations of amplitude and frequency in each frequency band of the time-series signal; expressing the physiological state of the living body in a multidimensional vector space consisting of the calculated feature amounts, and calculating a displacement vector indicating a change in the physiological state of the living body in the multidimensional vector space; a step of synthesizing the displacement vectors before and after an arbitrary physiological state change using the displacement vectors before and after a desired environmental stimulus stored in a storage unit that stores the displacement vectors before and after a plurality of environmental stimuli, and estimating the physiological state of the living body based on a synthesis ratio of the synthesized displacement vectors; A physiological state estimation program that executes the above.

Images