Muscle State Estimation Device
The muscle condition estimating device and method accurately assess muscle condition by integrating circumference and myoelectric potential measurements with mathematical optimization and machine learning, addressing the limitations of existing technologies.
Patent Information
- Application Number
- JP2023543869
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2021-08-25
- Filing Date
- 2022-08-19
- Publication Date
- 2024-11-13
- Estimated Expiration
- 2042-08-19
AI Technical Summary
Existing technologies are unable to accurately estimate the muscle condition of a subject.
A muscle condition estimating device and method that includes circumference measurement, myoelectric potential measurement, and mathematical optimization to determine muscle tension, co-contraction, and weakness, utilizing models such as muscle deformation and tension models, and machine learning to correlate muscle activity with circumference data.
Enables accurate estimation of muscle condition, including active and passive deformations, and health/psychological states, providing quantitative assessment of muscle tension, co-contraction, and weakness.
Smart Images

Figure 0007582715000003 
Figure 0007582715000004 
Figure 0007582715000005
Abstract
Description
[Technical field]
[0001] The present invention relates to a technique for estimating the state of muscles (hereinafter simply referred to as "muscles") that constitute the body. [Background technology]
[0002] Various measuring devices for estimating physical conditions have been proposed so far. For example, the following Patent Document 1 discloses a physical information generating device 1 that detects changes in the circumference of the biceps of a user performing a golf swing, thereby evaluating how the user uses the muscles.
[0003] Furthermore, the following Patent Document 2 discloses an apparatus for measuring the grip strength of a subject in accordance with changes in the circumference of a specific part of the subject's arm.
[0004] Furthermore, the following Patent Document 3 discloses an apparatus for acquiring a subject's muscle strength from movement data acquired using a motion capture system based on a musculoskeletal model, and from data on myoelectric potential and floor reaction force. [Prior art documents] [Patent documents]
[0005] [Patent Document 1] JP 2017-108871 A [Patent Document 2] JP 2018-110658 A [Patent Document 3] WO2005 / 122900A1 publication Summary of the Invention [Problem to be solved by the invention]
[0006] As described above, the physical information generating device 1 described in Patent Document 1 is capable of evaluating how a user uses his or her muscles, and the technology described in Patent Document 2 is capable of measuring the grip strength of a subject.
[0007] Moreover, the technology described in Patent Document 3 is said to be capable of obtaining muscle strength of a subject.
[0008] However, none of these techniques has the problem of being able to accurately estimate the state of the muscles of the user or subject.
[0009] The present invention has been made to solve such problems, and has an object to provide an apparatus and method capable of accurately estimating the state of a subject's muscles. [Means for solving the problem]
[0010] In order to solve the above problems, the present invention provides a muscle state estimation device for estimating the state of a subject's muscles, comprising: circumference measurement means for measuring the circumference of the muscle; electromyography measurement means for measuring an electric potential generated by activity of the muscle; and muscle state estimation means for calculating the state of the muscle such that an evaluation function whose variables are the deviation of the circumference measured by the circumference measurement means from a circumference model that defines the relationship between the muscle state and the circumference, and the deviation of the electric potential measured by the electromyography measurement means from a muscle tension model that defines the relationship between the muscle state and the electric potential, is minimized.
[0011] Furthermore, in order to solve the above-mentioned problems, the present invention provides a muscle state estimation device for estimating a state of a subject's muscles, comprising: circumference measurement means for measuring the circumference of the muscle; motion state measurement means for measuring a motion state of the entire body of the subject; and muscle state estimation means for calculating the state of the muscle such that an evaluation function is minimized, the deviation of the circumference measured by the circumference measurement means from a circumference model which defines the relationship between the state of the muscle and the circumference, and the deviation of the motion state measured by the motion state measurement means from a joint torque model which defines the relationship between the motion state and a joint torque generated in the subject, as variables.
[0012] In addition, in order to solve the above-mentioned problems, the present invention provides a muscle state estimation device for estimating a muscle state of a subject, comprising: a circumference measurement means for measuring the circumference of the muscle; and a muscle state estimation means for calculating the state of the muscle such that an evaluation function, the deviation of which from a learned model which has been machine-learned to determine the correlation between the circumference and the activity state of the muscle, is minimized.
[0013] Furthermore, in order to solve the above-mentioned problems, the present invention provides a muscle state estimation device for estimating the state of the muscles of a subject, comprising: circumference measurement means for measuring the circumference of the muscle; and muscle state estimation means for calculating the muscle activity data corresponding to the input circumference by inputting the measured circumference into a probabilistic circumference model constituted by muscle activity data correlated with the circumference and the circumference measured by the circumference measurement means, the measured circumference being an observation sequence and the muscle activity data being a latent variable sequence, and performing maximum likelihood sequence estimation.
[0014] In addition, in order to solve the above-mentioned problems, the present invention provides a muscle state estimation method for estimating the state of a subject's muscles, comprising: a first step of measuring the circumference of the muscle and an electric potential generated by activity of the muscle; and a second step of calculating the state of the muscle such that an evaluation function having variables, as variables, the deviation of the circumference measured in the first step from a circumference model that defines the relationship between the state of the muscle and the circumference, and the deviation of the electric potential measured in the first step from a muscle tension model that defines the relationship between the state of the muscle and the electric potential, is minimized.
[0015] In addition, in order to solve the above-mentioned problems, the present invention provides a muscle state estimation method for estimating a state of a subject's muscles, comprising: a first step of measuring the circumference of the muscle and a motion state of the subject's entire body; and a second step of calculating the state of the muscle such that an evaluation function is minimized, the variables being the deviation of the circumference measured in the first step from a circumference model that defines the relationship between the state of the muscle and the circumference, and the deviation of the motion state measured in the first step from a joint torque model that defines the relationship between the motion state and joint torque generated in the subject.
[0016] In addition, in order to solve the above-mentioned problems, the present invention provides a muscle state estimation method for estimating a muscle state of a subject, comprising: a first step of measuring the circumference of the muscle; and a second step of calculating the state of the muscle such that an evaluation function is minimized, the evaluation function being determined by, as a variable, a deviation from a learned model that has been machine-learned to determine the correlation between the circumference and the activity state of the muscle, and the circumference measured in the first step.
[0017] Furthermore, in order to solve the above-mentioned problems, the present invention provides a muscle state estimation method for estimating the state of a subject's muscles, comprising: a first step of measuring the circumference of the muscle; and a second step of calculating the muscle activity data corresponding to the input circumference by performing maximum likelihood sequence estimation by inputting the measured circumference into a probabilistic circumference model constituted by muscle activity data correlated with the circumference and a probability distribution of the entire system when the measured circumference is an observation sequence and the muscle activity data is a latent variable sequence, for the circumference measured in the first step. Effect of the Invention
[0018] According to the present invention, it is possible to provide a muscle state estimation device and a muscle state estimation method that can accurately estimate the muscle state of a subject. [Brief description of the drawings]
[0019] [Figure 1] 1 is a block diagram showing a configuration of a muscle state estimating device 1 according to a first embodiment of the present invention. [Diagram 2] 3 is a flowchart showing a muscle state estimating method according to the first embodiment of the present invention. [Diagram 3] FIG. 11 is a block diagram showing a configuration of a muscle state estimating device 10 according to a second embodiment of the present invention. [Figure 4] 10 is a flowchart showing a muscle state estimating method according to a second embodiment of the present invention. [Diagram 5] FIG. 11 is a block diagram showing a configuration of a muscle state estimating device 20 according to a third embodiment of the present invention. [Figure 6] 13 is a flowchart showing a muscle state estimating method according to a third embodiment of the present invention. [Figure 7] FIG. 11 is a block diagram showing the configuration of a muscle state estimating device 30 according to a fourth embodiment of the present invention. [Figure 8] 10 is a flowchart showing a muscle state estimating method according to a fourth embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0020] [Embodiment 1] Hereinafter, a first embodiment of the present invention will be described in detail with reference to the drawings, in which the same reference numerals indicate the same or corresponding parts.
[0021] Fig. 1 is a block diagram showing the configuration of a muscle state estimation device 1 according to embodiment 1 of the present invention. As shown in Fig. 1, muscle state estimation device 1 includes an input / output node N, a bus B connected to the input / output node N, and a circumference sensor 2, a myoelectric potential sensor 3, a muscle state estimation unit 4, and a memory 5, each connected to bus B.
[0022] Here, the circumference sensor 2 can be, for example, a known wearable device that is worn in a band-like manner around a part of the body, such as the upper arm, and measures the circumference of the muscle by sensing changes in length in a material made of elastic fiber, rubber, etc.
[0023] Furthermore, the myoelectric potential sensor 3 may be, for example, a commercially available wearable device that includes two electrodes arranged along a muscle path in a body part such as the upper arm, and measures the action potential that appears during muscle activity as a potential difference (myoelectric potential) between the electrodes.
[0024] The muscle state estimation unit 4 is configured with a central processing unit (CPU) that operates according to a predetermined program, the operation of which will be described later. The program is stored in advance in the memory 5 via the input / output node N and the bus B.
[0025] Furthermore, the memory 5 also stores data measured by the circumference sensor 2 and the myoelectric potential sensor 3, and results obtained by the muscle state estimation unit 4.
[0026] The above results can be output to an external monitor or the like via the bus B and the input / output node N for display or the like.
[0027] Fig. 2 is a flowchart showing a muscle state estimation method according to embodiment 1 of the present invention. In the following, the muscle state estimation method according to embodiment 1 of the present invention will be described taking as an example a case where the method is performed by muscle state estimation device 1 shown in Fig. 1, however, it goes without saying that the method is not limited to the case where it uses muscle state estimation device 1 and can also be performed by other means.
[0028] In step S1a, the circumference sensor 2 measures the circumference of a muscle such as the upper arm that constitutes a body part, and in step S1b, the myoelectric potential sensor 3 measures the myoelectric potential of the muscle.
[0029] Next, in step S2a, the muscle state estimation unit 4 compares the circumference measured in step S1a with a circumference model prepared in advance. Note that this circumference model is stored in the memory 5 or an external device (not shown) such as the cloud.
[0030] Here, the circumference model is expressed as a function f with the circumference being the muscle length and the degree of muscle activity and relaxation as variables. This model is based on the fact that muscle volume is constant, that muscle activity causes the cross-sectional area of the muscle belly to increase, that muscle elasticity decreases as the muscle relaxes, and that the muscle passively deforms in response to external forces such as elasticity measured by a circumference sensor.
[0031] Specifically, as the first example of a circumference model, Murai, A., Hong, QY, Yamane, K., & Hodgins, JK, 2017, Dynamic skin deformation simulation using musculoskeletal It is possible to use the known model described in "Muscle Deformation Model and Soft Tissue Dynamics," Computational Visual Media, 3(1), pp. 49-60 (hereinafter referred to as the "muscle deformation model").
[0032] As a second example, a known model including mechanical simulations such as the finite element method, described in YAMAMURA, N., Alves, JL, ODA, T., KINUGASA, R., & TAKAGI, S., 2014, Effect of tendon stiffness on the generated force at the Achilles tendon-3D finite element simulation of a human triceps surae muscle during isometric contraction, Journal of Biomechanical Science and Engineering, 9(3), pp.13-00294 (hereinafter referred to as the "FEM model"), can also be used.
[0033] While the above first muscle deformation model and second FEM model merely represent the relationship between positive muscle activity and muscle deformation, the circumference model according to the embodiment of the present invention further specifies the relationship between muscle circumference and the state in which the muscle is relaxed. In other words, this circumference model further adds the circumference to the above function f as a correction term representing negative muscle activity, which indicates physical muscle softness caused by relaxation due to the intervention of external force in neurally controlled relaxation or stretching, and by adding the degree of relaxation, inertia, and external force functions to the above function f, it is possible to determine muscle softening and the associated passive movement. Deformations may also be taken into consideration.
[0034] Then, the muscle state estimation unit 4 calculates the circumference from previously obtained measurements and estimated values for the muscle length, which indicates the muscle shape, and the degree of muscle activity and relaxation, which indicate the muscle shape characteristics and stiffness, and compares it with the circumference actually measured in step S1a as described above.
[0035] On the other hand, in step S2b, the muscle state estimation unit 4 compares the myoelectric potential measured in step S1b with a muscle tension model prepared in advance. This muscle tension model is also stored in the memory 5 or an external device (not shown) such as the cloud.
[0036] Here, the muscle tension model defines muscle tension as a function of muscle activity evaluated by myoelectric potential and muscle length and muscle strength F l (l) Function of muscle length change rate and muscle force F v (v) and maximum muscle tension F max Expressed as a product of This is a model that is modeled to have muscle activity and muscle length as variables.
[0037] Specifically, as a model of muscle tension, for example, a known hill-type muscle model is used. model) can be used.
[0038] Then, muscle state estimation unit 4 obtains a myoelectric potential from a previously obtained measured value or an estimated value for the muscle length, which indicates the shape of the muscle, and compares it with the myoelectric potential actually measured in step S1b as described above.
[0039] Next, in step S2c, the muscle state estimation unit 4 calculates and estimates the subject's muscle state, such as muscle tension, muscle co-contraction, and degree of muscle relaxation, so as to minimize an evaluation function whose variables are the deviation of the circumference measured by the circumference sensor 2 from a circumference model that defines the relationship between the muscle state and circumference, and the deviation of the potential measured by the EMG sensor 3 from a muscle tension model that defines the relationship between the muscle state and the above potential.
[0040] Specifically, muscle state estimation unit 4 determines the degree of muscle tension, muscle co-contraction, and muscle relaxation that minimizes evaluation function Z1 expressed by, for example, the following equation (1).
number
[0041] In addition, C1, C2, and C3 in the formula (1) represent different constants.
[0042] In addition, the "function of muscle activity and movement based on exercise and physiological knowledge" in the third term on the right-hand side of equation (1) refers to a function of muscle activity and movement that is based on general exercise and physiological knowledge and expresses, for example, "small co-contraction during quick movements," "large co-contraction during precise manipulation," and "reducing compressive forces in joints, such as intervertebral disc pressure," and an example of such a function is the function disclosed in JP 2019-5344 A.
[0043] In the above, for example, in the case of the upper arm, a model consisting of multiple muscles such as the biceps and triceps can be considered, and the muscle state of each muscle such as muscle tension, muscle co-contraction, and the degree of muscle relaxation can be estimated by performing the optimization calculation as described above from the circumference and muscle activity obtained from the circumference sensor 2 and multiple myoelectric potential sensors 3 corresponding to the number of muscles included in the model. In particular, even when the muscle activity is 0, optimization is possible taking into account the negative inequality constraint condition for the muscle circumference and muscle activity, so the degree of muscle relaxation can be estimated.
[0044] According to the muscle state estimation device 1 of embodiment 1 of the present invention and the muscle state estimation method shown in FIG. 2 as described above, by measuring co-contraction including both antagonistic muscles using the circumference sensor 2, it is possible to avoid the difficulty of measuring muscle activity on only one side and quantitatively estimate not only the active deformation of muscles but also the passive deformation of muscles, i.e., the degree of relaxation, which cannot be obtained by conventional kinematic analysis.
[0045] Furthermore, since the measurement target can be limited to a part of the subject's body as described above, the muscle state estimating device 1 can also be configured as a wearable device.
[0046] [Embodiment 2] A second embodiment of the present invention will be described below, but a description of the parts common to the first embodiment will be omitted and differences will be mainly described.
[0047] Fig. 3 is a block diagram showing the configuration of a muscle state estimation device 10 according to embodiment 2 of the present invention. As shown in Fig. 3, muscle state estimation device 10 according to embodiment 2 includes a motion capture device 6 instead of the myoelectric potential sensor 3 in muscle state estimation device 1 according to embodiment 1.
[0048] Here, the motion capture 6 can use a known technique for measuring the whole body movement of the subject optically or by video.
[0049] Fig. 4 is a flowchart showing a muscle state estimation method according to embodiment 2 of the present invention. In the following, the muscle state estimation method according to embodiment 2 of the present invention will be described taking as an example a case where it is performed by muscle state estimation device 10 shown in Fig. 3, however, it goes without saying that this method is not limited to the case where it uses muscle state estimation device 10 and can also be performed by other means.
[0050] In step S1d, the motion state of the subject's entire body, that is, for example, muscle length and joint torque, is measured by the motion capture 6. Here, the muscle length and joint torque are calculated by the muscle state estimation unit 7 through inverse kinematic calculation using a musculoskeletal model.
[0051] As the musculoskeletal model, for example, a known model described in the above-mentioned Patent Document 3 can be used, and this model is stored in the memory 5 or an external device (not shown) such as a cloud.
[0052] Next, in step S2a, the muscle state estimating unit 7 compares the circumference measured in step S1a and the muscle length obtained in step S1d with a circumference model prepared in advance, as described above.
[0053] On the other hand, in step S2d, the muscle state estimating unit 7 compares the joint torque obtained in step S1b with a joint torque model prepared in advance.
[0054] Here, the joint torque model is, for example, T., & Nakamura, Y., 2014, September, Musculoskeletal modeling and physiological As described in 2014 IEEE International Workshop on Advanced Robotics and its Social Impacts, pp. 108-113, IEEE, the function g shown in Fig. 4 is used for validation. A known model for distributing joint torque to muscle tensions based on joint motion can be used.
[0055] Then, in step S2e, the muscle state estimation unit 7 calculates and estimates the muscle states of the subject, such as muscle tension, muscle co-contraction, and degree of muscle relaxation, so as to minimize mathematical optimization, i.e., an evaluation function whose variables are the deviation of the circumference measured by the circumference sensor 2 from a circumference model that defines the relationship between the muscle state and circumference, and the deviation of the movement state measured by the motion capture 6 from the joint torque model that defines the relationship between the movement state and the joint torque generated in the subject.
[0056] Specifically, muscle state estimation unit 7 calculates and estimates the degree of muscle tension, muscle co-contraction, and muscle relaxation of the subject so as to minimize evaluation function Z2 expressed by, for example, the following equation (2).
number
[0057] Here, C21, C22, and C23 in formula (2) represent different constants.
[0058] Furthermore, the muscle state estimation device 10 according to the second embodiment further includes the myoelectric potential sensor 3 shown in FIG. 1, and the muscle state estimation unit 7 adds the deviation of the potential measured by the myoelectric potential sensor 3 from the muscle tension model according to the first embodiment, which defines the relationship between the muscle state and the potential, to the variable of evaluation function Z2, and then calculates the muscle state that minimizes evaluation function Z2, thereby enabling more accurate estimation of the muscle state.
[0059] The muscle state estimation device 10 according to the second embodiment of the present invention and the muscle state estimation method shown in FIG. 4 as described above have the same effects as the muscle state estimation device 1 and muscle state estimation method according to the first embodiment. In addition, by measuring the physical quantity of the muscle circumference instead of the myoelectric potential, it is possible to avoid the effects of changes in impedance that occur between the electrodes and the skin, and as a result, it is possible to realize stable measurement over a long period of time.
[0060] [Embodiment 3] In the following, a description of the parts common to the first and second embodiments will be omitted, and differences will be mainly described.
[0061] Fig. 5 is a block diagram showing the configuration of a muscle state estimating device 20 according to embodiment 3 of the present invention. As shown in Fig. 5, muscle state estimating device 20 according to embodiment 3 has a configuration obtained by excluding myoelectric potential sensor 3 from muscle state estimating device 1 according to embodiment 1.
[0062] Fig. 6 is a flowchart showing a muscle state estimation method according to embodiment 3 of the present invention. In the following, the muscle state estimation method according to embodiment 3 of the present invention will be described taking as an example a case where it is performed by muscle state estimation device 20 shown in Fig. 5, however, it goes without saying that this method is not limited to the case where it uses muscle state estimation device 20 and can also be performed by other means.
[0063] When data on circumference, myoelectric potential, whole body movement, and muscle state (muscle activity and relaxation) obtained by implementing the above-mentioned first and second embodiments are accumulated, a certain circumference pattern corresponds to a muscle state (muscle activity and relaxation) pattern, and a correlation model can be constructed between the two. Therefore, by utilizing this correlation model, muscle state data can be estimated from circumference data.
[0064] That is, in step S2f, the muscle state estimation unit 8 compares (inputs) the circumference measured in step S1 with the learned circumference model or the probabilistic circumference model. Here, the learned circumference model is constructed by machine learning the correlation F between the change in muscle circumference (circumference) and the muscle activity state (muscle state) using the circumference and muscle activity data obtained separately as described above as teacher data. On the other hand, the probabilistic circumference model is constructed by obtaining a probability distribution of the entire system for the circumference and muscle activity data obtained separately as described above in a probabilistic transition model represented by a hidden Markov model (HMM), with the circumference as an observed sequence and the muscle activity data as a latent variable sequence. Note that the probabilistic transition model refers to a general model that expresses the relationship between multiple synchronous or asynchronous time series data and estimates or generates other time series data from some time series data, and includes a time series regression model and other statistical models that are not probabilistic models.
[0065] The learned circumference model or the probabilistic circumference model is stored in the memory 5 or an external device (not shown) such as a cloud.
[0066] Next, in step S2g, the muscle state estimation unit 8 calculates and estimates the muscle state by mathematical optimization, i.e., by minimizing an evaluation function having a circumference model that specifies the relationship between the muscle state and the circumference measured by the circumference sensor 2 and a deviation from the learned model as a variable. On the other hand, when the probabilistic circumference model is used, the muscle state estimation unit 8 calculates muscle activity data, which is a latent variable sequence that generates a specific observational knowledge sequence by maximum likelihood sequence estimation when the circumference is observed.
[0067] As with muscle state estimation device 10 and muscle state estimation method according to embodiment 2, muscle state estimation device 20 according to embodiment 3 of the present invention and the muscle state estimation method shown in FIG. 6 do not measure myoelectric potentials, and therefore can avoid the effects of changes in impedance between the electrodes and the skin, making it possible to achieve stable measurement over a long period of time.
[0068] [Embodiment 4] Fig. 7 is a block diagram showing the configuration of a muscle state estimation device 30 according to embodiment 4 of the present invention. As shown in Fig. 7, muscle state estimation device 30 according to embodiment 4 has a similar configuration to muscle state estimation device 20 according to embodiment 3 shown in Fig. 5, but differs in that it further includes a health / mental state estimation unit 9 connected to bus B.
[0069] Fig. 8 is a flowchart showing a muscle state estimation method according to embodiment 4 of the present invention. In the following, the muscle state estimation method according to embodiment 4 will be described with reference to Fig. 8. However, a description of parts common to above embodiments 1 to 3 will be omitted, and mainly differences will be described.
[0070] Furthermore, the muscle state estimation method according to embodiment 4 can be implemented, for example, by muscle state estimation device 30 shown in FIG. 7. In the following, the muscle state estimation method will be described as being implemented by muscle state estimation device 30; however, it goes without saying that the method can also be implemented by other means.
[0071] When data on circumference, myoelectric potential, whole body movement and muscle state (muscle activity and relaxation) measured by the methods described in the first and second embodiments, and data on health conditions such as fatigue and illness and psychological conditions such as tension and anxiety measured at the same or different granularity simultaneously with all or some of these are accumulated, a correlation model can be constructed between the two because the patterns of the health and psychological conditions correspond to a certain circumference pattern or a long-term circumference pattern such as intraday or interday fluctuations. Therefore, by utilizing this correlation model, health and psychological data can be estimated from circumference data.
[0072] Specifically, in step S3a, the health and mental state estimation unit 9 generates a second calculation result expressing the health and mental states as a function i of the muscle state calculated by the muscle state estimation means 8. Contrast with the health and psychological state model.
[0073] In addition, in step S3b, the health / mental state estimation unit 9 compares the muscle state calculated by the muscle state estimation unit 8 with a learned second health / mental state model in which the correlation H between the muscle state calculated over time by the muscle state estimation unit 8 and the corresponding data on the health state and mental state described above are used as teacher data, and the health / mental state estimation unit 9 compares the muscle state calculated by the muscle state estimation unit 8 with a learned second health / mental state model in which the correlation H between the muscle state calculated over time by the muscle state estimation unit 8 and the corresponding data on the health state and mental state described above are machine-learned, or with a probabilistic health / mental state model based on a probabilistic transition model such as a hidden Markov model (HMM) as described above.
[0074] Similarly, in step S4a, the health and mental state estimation unit 9 compares the intra-day and inter-day variation data regarding the muscle state, muscle tension, muscle co-contraction, and degree of muscle relaxation calculated by the muscle state estimation unit 8 based on the circumference with a first health and mental state model that expresses the health state and mental state as a function h of the muscle state.
[0075] In addition, in step S4b, the health / mental state estimation unit 9 compares the muscle state calculated by the muscle state estimation unit 8 over time with the corresponding data on the health and mental states described above as teacher data, and compares the muscle state calculated by the muscle state estimation unit 8 with a learned first health / mental state model in which correlation G between them is machine-learned, or with a probabilistic health / mental state model based on a probabilistic transition model such as a hidden Markov model (HMM) as described above.
[0076] Then, in steps S3c and 4c, the health and mental state estimation unit 9 calculates health and mental state data that minimizes an evaluation function whose variables are the deviation from the second and first health and mental state models and the learned second and first health and mental state models, or performs maximum likelihood sequence estimation using the probabilistic health and mental model, thereby estimating the subject's health and mental state, such as fatigue, illness, tension and anxiety.
[0077] As in the third embodiment, the probabilistic transition model in the above refers to a general model that expresses the relationship between multiple synchronous or asynchronous time series data and estimates or generates some time series data from other time series data, and includes time series regression models and other statistical models that are not probabilistic models.
[0078] In addition, in the above, the health / mental state estimation unit 9 may be integrated with the muscle state estimation unit 8 to estimate the health state and mental state of the subject by calculating data on the health state and mental state directly from the circumference without using the muscle state.
[0079] According to the muscle state estimation device 30 according to the fourth embodiment of the present invention and the muscle state estimation method shown in FIG. 8 as described above, the health state and psychological state of the subject can be quantitatively estimated by the simple method of measuring the muscle circumference using the circumference sensor 2. [Explanation of symbols]
[0080] 1, 10, 20, 30 muscle state estimation device, 2 circumference sensor, 3 myoelectric potential sensor, 4, 7, 8 muscle state estimation unit, 6 motion capture, 9 health and mental state estimation unit.
Claims
1. A muscle state estimation device that estimates a muscle state of a subject, comprising: A circumference measuring means for measuring the circumference of the muscle; a myoelectric potential measuring means for measuring an electric potential generated by the activity of the muscle; a muscle state estimation means for calculating a state of the muscle such that an evaluation function having variables, the deviation of the circumference measured by the circumference measuring means from a circumference model which defines the relationship between the muscle state and the circumference, and the deviation of the potential measured by the myoelectric potential measuring means from a muscle tension model which defines the relationship between the muscle state and the potential, is minimized.
2. A muscle state estimation device that estimates a muscle state of a subject, comprising: A circumference measuring means for measuring the circumference of the muscle; A motion state measuring means for measuring a motion state of the subject's whole body; a muscle state estimation means for calculating a state of the muscle such that an evaluation function having as variables a deviation of the circumference measured by the circumference measuring means from a circumference model which defines the relationship between the state of the muscle and the circumference, and a deviation of the motion state measured by the motion state measuring means from a joint torque model which defines the relationship between the motion state and a joint torque generated in the subject, is minimized.
3. The device further includes a myoelectric potential measuring means for measuring an electric potential generated by the activity of the muscle.
3. The muscle state estimation device according to claim 2, wherein the muscle state estimation means further calculates the muscle state such that the evaluation function is minimized after adding a deviation of the potential measured by the myoelectric potential measuring means from a muscle tension model that defines a relationship between the muscle state and the potential to a variable of the evaluation function.
4. A muscle state estimation device that estimates a muscle state of a subject, comprising: A circumference measuring means for measuring the circumference of the muscle; A muscle state estimation device comprising: a circumference model that defines the relationship between the muscle state and the circumference measured by the circumference measurement means; and a muscle state estimation means that calculates the muscle state such that an evaluation function, the deviation from a trained model that has been machine-learned to determine the correlation between the circumference and the activity state of the muscle, is minimized.
5. The muscle state estimation device according to claim 4, further comprising a health and psychological state estimation means for calculating the health and psychological state of the muscle calculated by the muscle state estimation means so as to minimize an evaluation function having as variables a deviation from a learned health and psychological state model which defines a relationship between the subject's health and psychological state and the muscle state, and which has been machine-learned to determine the correlation between the muscle state and the health and psychological state.
6. The muscle state estimation device according to claim 1 , wherein the circumference model defines a relationship between the circumference and a state in which the muscle is relaxed.
7. A muscle state estimation device that estimates a muscle state of a subject, comprising: A circumference measuring means for measuring the circumference of the muscle; and a muscle state estimation means for calculating the muscle activity data corresponding to the input circumference by inputting the measured circumference into a probabilistic circumference model constituted by a probability distribution of the entire system when the measured circumference is an observation sequence and the muscle activity data is a latent variable sequence, for muscle activity data correlated with the circumference and the circumference measured by the circumference measurement means, and performing maximum likelihood sequence estimation on the measured circumference.
Citation Information
Patent Citations
Motion measuring method, motion measuring device, and motion measuring program
JP2006014837A
Construction method of muscle tension database, muscle tension database and muscle tension calculation method and device using muscle tension database
JP2010029340A
Body information generation device
JP2017108871A
Grip strength measurement method, grip strength measurement device, method, and grip strength measurement system
JP2018110658A
Mascular strength acquiring method and device based on musculo skeletal model
WO2005122900A1
