Dynamic posture stability evaluation system and dynamic posture stability evaluation method

Abstract

[Problem] To provide a dynamic posture stability evaluation system and a dynamic posture stability evaluation method for evaluating posture stability by repeated measurement by a subject. [Solution] The present invention is provided with: an acquisition means for acquiring repeated measurement data by a subject 3; a calculation means for configuring a multidimensional variable space for evaluating the posture stability of the subject 3 on the basis of the repeated measurement data, and calculating a relative feature amount on the basis of each combination of a plurality of element variables; a generation means for developing the feature amount in the multidimensional variable space, and generating an image for evaluation indicating a state of posture stability; and an evaluation means for evaluating the stability of a dynamic posture of the subject on the basis of the image for evaluation.

Description

Dynamic Posture Stability Evaluation System and Dynamic Posture Stability Evaluation Method

[0001] The present invention relates to a dynamic posture stability evaluation system and a dynamic posture stability evaluation method for evaluating posture stability through repeated measurements by a subject.

[0002] Conventionally, as a device for analyzing human behavior by analyzing measurement data acquired by sensors worn on the human body, for example, a behavior analysis device disclosed in Patent Document 1 has been proposed. Also, as a system for calculating a floor reaction force index at low computational cost by connecting footwear equipped with wearable sensors and an external device via a wireless line, for example, a floor reaction force index estimation system disclosed in Patent Document 2 has been proposed.

[0003] The behavior analysis device disclosed in Patent Document 1 includes an analysis means and a storage means for analyzing human behavior using acceleration data output from a three-axis acceleration sensor worn on the human body. This analysis means separates the feature amounts of the human body's posture and the feature amounts of the human body's movement included in the temporal change of the acceleration data accumulated as time-series data, and defines a model consisting of the feature amounts of the movement and posture of the basic behavior pattern, and uses the model of the basic behavior pattern to perform processing to separate newly accumulated acceleration data as time-series data, and clusters the processed acceleration data to recognize an extended behavior pattern.

[0004] Further, the floor reaction force index estimation system disclosed in Patent Document 2 includes footwear, a wearable sensor that measures three-axis acceleration and / or three-axis angular velocity, and a calculation unit that calculates a floor reaction force index by multiple regression analysis using values used as explanatory variables calculated from the measurement results. The external device includes a control unit that evaluates the walking and running actions of the user based on the measurement results by the wearable sensor received from the footwear, the calculated floor reaction force index, and the determination criterion information, and a presentation unit that presents the result of the evaluation of the running action. In multiple regression analysis, the floor reaction force, which is the target variable, is calculated using a multiple regression equation in which the partial regression coefficient when the result calculated from the measurement results is used as the explanatory variable has been calculated in advance.

[0005] JP-A-2010-207488 JP-A-2022-182594

[0006] In contrast, in behavioral analysis devices such as Patent Document 1, or ground reaction force index estimation systems such as Patent Document 2, the measurement data of a subject obtained from sensors does not take repeatability into consideration. Furthermore, the acquired measurement data is evaluated by comparing it with basic behavioral patterns or judgment criterion information. Therefore, it is difficult to identify the subject's individual characteristics and potential problems and improve the accuracy of dynamic postural stability evaluation.

[0007] Therefore, the present invention was devised in view of the above-mentioned problems, and its objective is to provide a dynamic posture stability evaluation system and a dynamic posture stability evaluation method that can identify the individual characteristics and potential problems of a subject from the results of repeated measurements and improve the accuracy of dynamic posture stability evaluation.

[0008] The dynamic posture stability evaluation system according to the first invention is a dynamic posture stability evaluation system that evaluates posture stability by repeated measurements by a subject, and is characterized by comprising: acquisition means for acquiring repeated measurement data by the subject; calculation means for constructing a multidimensional variable space for evaluating the posture stability of the subject based on the repeated measurement data and calculating relative feature quantities based on each combination of a plurality of element variables; generation means for expanding the feature quantities into the multidimensional variable space and generating an evaluation image showing the state of posture stability; and evaluation means for evaluating the posture stability of the subject based on the evaluation image.

[0009] The dynamic posture stability evaluation system according to the second invention is characterized in that, in the first invention, the repeated measurement data acquired by the acquisition means includes ground reaction force data.

[0010] The dynamic posture stability evaluation system according to the third invention is characterized in that, in the first invention, the evaluation means includes evaluation images of a plurality of subjects and evaluates posture stability based on information including at least one of the attributes, personal characteristics, or time-series changes of a specific subject.

[0011] The dynamic posture stability evaluation system according to the fourth invention is characterized in that, in the first or third invention, the evaluation means further includes a machine learning evaluation means that evaluates the posture stability of a specific subject using a machine learning model.

[0012] The dynamic posture stability evaluation system according to the fifth invention is characterized in that, in the first invention, it further comprises a feedback providing means that provides feedback information for posture improvement to the subject based on the evaluation results from the evaluation means.

[0013] The dynamic posture stability evaluation method according to the sixth invention is characterized by having each step performed by the dynamic posture stability evaluation system in any of the first to fifth inventions.

[0014] According to the first invention, the acquisition means acquires repeated measurement data from the subject. Therefore, the calculation means can construct a multidimensional variable space based on the repeated measurement data and calculate relative features based on each combination of multiple element variables. This makes it possible to identify the subject's individual characteristics and potential problems from the results of repeated measurements and improve the accuracy of dynamic posture stability evaluation.

[0015] Furthermore, according to the first invention, the generation means expands feature quantities into a multidimensional variable space and generates an evaluation image that shows the state of posture stability. Therefore, the evaluation means can evaluate the posture stability of a specific subject based on the evaluation image. This makes it possible to identify the subject's individual characteristics and potential problems, and to improve the accuracy of dynamic posture stability evaluation.

[0016] Furthermore, according to the first to fifth inventions, repeated measurement data from multiple subjects is acquired, a multidimensional variable space is constructed for evaluating the postural stability of one subject compared to other subjects, and relative features are calculated based on each combination of multiple elemental variables. Therefore, each feature can be expanded into the multidimensional variable space, and evaluation images can be generated to compare the postural stability of one subject with that of other subjects. This makes it possible to identify the individual characteristics and potential problems of the subjects and improve the accuracy of dynamic postural stability evaluation.

[0017] In particular, according to the second invention, the repeated measurement data acquired by the acquisition means includes ground reaction force data. Therefore, the calculation means can construct a multidimensional variable space based on the ground reaction force data and calculate relative feature quantities based on each combination of multiple element variables. This makes it possible to identify the individual characteristics and potential problems of the subject and improve the accuracy of dynamic postural stability evaluation.

[0018] In particular, according to the third invention, the evaluation images include evaluation images of multiple subjects. Therefore, postural stability can be evaluated based on information including at least one of the attributes, personal characteristics, or time-series changes of a specific subject. This makes it possible to identify the personal characteristics and potential problems of the subjects and improve the accuracy of dynamic postural stability evaluation.

[0019] In particular, according to the fourth invention, the evaluation means further includes a machine learning evaluation means that uses a machine learning model to evaluate the postural stability of a specific subject. Therefore, it is possible to generate evaluation results based on the results of past evaluations of the subject's postural stability. This makes it possible to identify the subject's individual characteristics and potential problems, and to improve the accuracy of dynamic postural stability evaluation.

[0020] In particular, according to the fifth invention, the feedback providing means provides feedback information for posture improvement to a specific subject based on the evaluation results. As a result, the subject can grasp information appropriate to their individual characteristics and potential problems and make appropriate improvements. This makes it possible to identify the subject's individual characteristics and potential problems and improve the accuracy of dynamic posture stability evaluation.

[0021] Furthermore, according to the sixth invention, the dynamic posture stability evaluation method comprises each step performed by the dynamic posture stability evaluation system. Therefore, each feature quantity can be expanded into a multidimensional variable space, and an evaluation image can be generated that compares the posture stability state of the subject with that of other subjects. This makes it possible to identify the subject's individual characteristics and potential problems, thereby improving the accuracy of the dynamic posture stability evaluation.

[0022] Figure 1 is a schematic diagram showing an example of the dynamic posture stability evaluation system in this embodiment. Figure 2 is a schematic diagram showing another example of the dynamic posture stability evaluation system in this embodiment. Figure 3(a) is a schematic diagram showing an example of the configuration of the dynamic posture stability evaluation device in this embodiment, and Figure 3(b) is a schematic diagram showing an example of the functions of the dynamic posture stability evaluation device in this embodiment. Figure 4 is a schematic diagram showing an example of the data in the measurement information table of the dynamic posture stability evaluation system in this embodiment. Figure 5 is a schematic diagram showing an example of the data in the subject information table of the dynamic posture stability evaluation system in this embodiment. Figure 6 is a schematic diagram showing an example of the evaluation image of the dynamic posture stability evaluation system in this embodiment. Figure 7 is a schematic diagram showing an example of the reference database in this embodiment. Figure 8 is a flowchart showing an example of the operation of the dynamic posture stability evaluation system in this embodiment.

[0023] Hereinafter, an example of a dynamic posture stability evaluation system in which the present invention is applied will be described with reference to the drawings.

[0024] First, an example of the dynamic posture stability evaluation system 100 in this embodiment will be described with reference to Figures 1 and 2.

[0025] (Dynamic Postural Stability Evaluation System 100) As shown in Figures 1 and 2, the dynamic posture stability evaluation system 100 in this embodiment evaluates the dynamic posture stability of subject 3 based on repeated measurement data obtained by repeatedly performing an exercise task (task, repeated measurement) on subject 3. The dynamic posture stability evaluation system 100 includes, for example, a dynamic posture stability evaluation device 1.

[0026] The dynamic posture stability evaluation system 100 includes a dynamic posture stability evaluation device 1, as shown in Figure 1, for example. The dynamic posture stability evaluation device 1 connects various sensors 2 (e.g., ground reaction force sensors, etc.) and acquires repeated measurement data (e.g., ground reaction force data) measured by the repeated execution of exercise tasks (e.g., single-leg drop landing test, etc.) given to multiple subjects 3 or a specific subject 3. The dynamic posture stability evaluation device 1 may also be connected to other terminals 5, servers 6, etc. In addition to inputting various information related to the subject 3 and setting exercise tasks, the dynamic posture stability evaluation device 1 may also evaluate the stability of the subject 3's dynamic posture based on repeated measurement data acquired via sensors 2, for example.

[0027] The dynamic posture stability evaluation device 1 acquires repeated measurement data measured by subject 3 through repeated execution of a set exercise task. Based on the acquired repeated measurement data, the dynamic posture stability evaluation device 1 calculates the structure of a multidimensional variable space for evaluating the stability of subject 3's dynamic posture and relative features based on each combination of multiple element variables. The dynamic posture stability evaluation device 1 expands the calculated features into the multidimensional variable space, generates an evaluation image showing the state of posture stability, and evaluates the stability of a specific subject 3's dynamic posture based, for example, on a comparison of the evaluation images of a specific subject 3 with those of other subjects 3.

[0028] As shown in Figure 2, the sensors 2 may be, for example, multiple sensors 2 (2a to 2e) installed in a space at another location A, or multiple sensors 2 (2f, 2g) installed in a space at location B, and are connected to the dynamic posture stability evaluation device 1 to measure repetitive measurement data for the exercise task of the subject 3 at each location. For example, ground reaction force sensors or other known sensors may be used as sensors 2 (2a to 2g). The multiple sensors 2 may be connected individually or all together to the dynamic posture stability evaluation device 1 and configured to work together as a function of the dynamic posture stability evaluation device 1.

[0029] Sensors 2 (2a to 2g) may be, for example, motion sensors that measure the movement of the entire body or specific body parts of the subject 3, or near-infrared or non-contact vital sensors that measure the body temperature and vital signs of the subject 3. Similarly, motion sensors or vital sensors may acquire repetitive measurement data by repeatedly performing an exercise task set for the subject 3. Motion sensors or vital sensors may, for example, measure different measurement data from other sensors 2 (ground reaction force sensors) in conjunction with repetitive movements due to the exercise task. This makes it possible for the dynamic posture stability evaluation device 1 to acquire, for example, measurement data specific to each sensor measured by multiple sensors 2, identify the individual characteristics and potential problems of the subject 3, and improve the accuracy of the evaluation of dynamic posture stability.

[0030] The dynamic posture stability evaluation system 100 also includes an interface for sending and receiving various information and measurement data. The communication network 4 may be, for example, the regular internet, an external cloud, a network connected to a server, or a network connected only within a closed environment.

[0031] Furthermore, the dynamic posture stability evaluation system 100 and dynamic posture stability evaluation device 1 may be electronic devices such as a cloud server or a personal computer (PC), or they may be electronic devices such as a smartphone, tablet, wearable device, IoT (Internet of Things) device, or a single-board computer such as Raspberry Pi®, and may have sensors 2 (2a to 2g) built in.

[0032] (Dynamic Posture Stability Evaluation Device 1) Next, an example of the dynamic posture stability evaluation device 1 in this embodiment will be described with reference to Figure 3. Figure 3(a) is a schematic diagram showing an example of the configuration of the dynamic posture stability evaluation device 1 in this embodiment, and Figure 3(b) is a schematic diagram showing an example of the function of the dynamic posture stability evaluation device 1 in this embodiment.

[0033] The dynamic attitude stability evaluation device 1, as shown in Figure 3(a) for example, comprises a housing 10, a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, a storage unit 104, and I / Fs 105 to 107. Each of the components 101 to 107 is connected by an internal bus 110.

[0034] The CPU 101 controls the entire dynamic attitude stability evaluation device 1. The ROM 102 stores the operation code of the CPU 101. The RAM 103 is a work area used when the CPU 101 is operating. The storage unit 104 stores various information such as notification-related information and reference databases. As the storage unit 104, a data storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive) can be used. For example, the dynamic attitude stability evaluation device 1 may also have a GPU (Graphics Processing Unit) which is not shown. Having a GPU enables faster computation processing than usual.

[0035] I / F105 is an interface for sending and receiving various types of information with sensors 2 (2a to 2g). I / F105 may also be an interface for sending and receiving various types of information with various sensors 2 (2a to 2g), other terminals 5 or servers 6, or remote controls that control multiple sensors 2 (2a to 2g) via a communication network 4 such as the regular internet or a closed local network (Local Area Network).

[0036] I / F 106 is an interface for sending and receiving various types of information with the input unit 108. The input unit 108 can be, for example, a keyboard, sensors 2 (2a to 2g), or a remote control. Administrators and others using the dynamic posture stability evaluation device 1 input various types of information, control commands for the dynamic posture stability evaluation device 1, sensors 2 (2a to 2g), or the remote control via the input unit 108.

[0037] I / F 107 is an interface for sending and receiving various information with the display unit 109. The display unit 109 outputs various information such as the selection of the subject 3 stored in the storage unit 104, the setting of exercise tasks, the execution of exercise tasks, and the acquisition of execution result data (repeated measurement data), or the processing status of sensors 2 (2a to 2g), the dynamic posture stability evaluation device 1, and instructions to other various devices. A display is used as the display unit 109, and may be a touch panel type, for example.

[0038] Here, with reference to Figures 4 and 5, an example of the information table for the dynamic posture stability evaluation system 100 (dynamic posture stability evaluation device 1) in this embodiment will be described. Figure 4 is a schematic diagram showing an example of the data in the measurement information table of the dynamic posture stability evaluation system 100 in this embodiment.

[0039] The "measurement information table" shown in Figure 4 stores information such as various exercise tasks to be repeatedly performed by subject 3, and the results of repeated execution of the exercise tasks (repeated measurement data). The exercise tasks may be set by, for example, an administrator operating the dynamic posture stability evaluation device 1, which displays various menus for setting exercise tasks, and the administrator can select an appropriate task from various display screens (not shown).

[0040] The "Measurement Information Table" stores items related to the exercise task for subject 3, such as "A. Exercise Task ○○○ ... Perform ○○ in the ○○ posture," various items related to the specific measurement of the exercise task, and measured repeated measurement data such as "Measurement Target," "Number of Repetitions," and "Measurement Time." The "Measurement Information Table" may also store (pre-set) various elements (parameters, indicators, etc.) for evaluating the dynamic posture of subject 3, such as "Anterior-posterior ground reaction force (Fy)" and "Lateral ground reaction force (Fx)," along with the repeated measurement data.

[0041] The repeated measurement data stored in the "measurement information table" is referenced as appropriate for processing such as constructing a multidimensional variable space for evaluating the postural stability of subject 3 by the dynamic postural stability evaluation device 1, and calculating relative feature quantities based on various combinations of multiple element variables. The "measurement information table" may also be stored as a "calculation table," for example, various calculation formulas or groups of indicators (not shown) for the dynamic postural stability evaluation device 1 to calculate feature quantities. The "measurement information table" is referenced as appropriate by the dynamic postural stability evaluation device 1, other terminals 5, and servers 6 when evaluating the dynamic postural stability of a specific subject 3 and other subjects 3. The "measurement target," "number of repetitions," and "measurement time" set in the "measurement information table" may be set by an administrator via the dynamic postural stability evaluation device 1 (measurement target, specific number of repetitions, specific timing, etc.), and various repeated measurement data measured by various sensors 2, related specific data, etc., may be recorded based on the various setting information.

[0042] Next, Figure 5 is a schematic diagram showing an example of data in the subject information table of the dynamic posture stability evaluation system 100 in this embodiment. The "subject information table" shown in Figure 5 stores various information that identifies subject 3, for example. The "subject information table" stores, for example, a "subject ID" that identifies subject 3, and "subject information" as various information about subject 3 linked to the "subject ID". "Subject information" stores information such as basic information about subject 3, including "age," "gender," "height," and "weight," as well as information such as whether subject 3 regularly participates in sports or exercise, the content and frequency of such activities, "exercise presence / activity information," "injury presence / treatment information," various information about subject 3's past and present injuries and treatments, "personal characteristics / risk information," various information about subject 3's lifestyle and physical characteristics and potential risks, "measurement result ID information," which indicates various information regarding the evaluation of subject 3's dynamic postural stability, and "evaluation results / improvement information," which is feedback information such as various evaluation results, recommendations, and advice for subject 3. "Measurement result ID information" stores information such as information linked to the ID information indicating the measurement result, or detailed measurement result information.

[0043] The "measurement information table" and the "subject information table" may be referenced as appropriate by, for example, the dynamic posture stability evaluation device 1, other terminals 5 and servers 6 when setting exercise tasks for a specific subject 3 and other subjects 3, when acquiring repeated measurement data during repeated execution, and when evaluating the dynamic posture stability of subject 3. The "measurement information table" and the "subject information table" may be set by, for example, the dynamic posture stability evaluation system 100 (dynamic posture stability evaluation device 1) and stored in the server 6 of the dynamic posture stability evaluation system 100, the storage unit 104 of the dynamic posture stability evaluation device 1, etc.

[0044] Here, referring to FIG. 6, an example of an evaluation image of the dynamic postural stability evaluation system 100 in the present embodiment will be described. The evaluation image shown in FIG. 6 shows, for example, an image of the evaluation results of the dynamic postural stability of a specific subject 3 (for example, six subjects). The evaluation image constitutes a multi-dimensional variable space based on, for example, repeated measurement data repeated by a plurality of subjects 3, calculates relative feature amounts based on each combination of a plurality of element variables, and the feature amounts are developed in the multi-dimensional variable space and a discrimination line is given to create the evaluation image.

[0045] The evaluation image shown in FIG. 6 is composed of, for example, a two-dimensional space. On the X-axis, as evaluation dimension 1, for example, the numerical value of reaction force data is set, and on the Y-axis, as evaluation dimension 2, for example, the shock buffering coefficient obtained by a predetermined calculation is set. It is developed (plotted) based on the repeated measurement data of a plurality of subjects 3, optimized so that the dispersion ratio between a specific data group and the target data group is maximized, and shows the result of giving a discrimination line. The discrimination line will be given a plurality of patterns of discrimination lines in order to show the specificity among a plurality of subjects 3 and groups according to the degree of dispersion of the data group and the target group.

[0046] The discrimination line is appropriately grouped (for example, evaluation A, evaluation B) according to, for example, the number of plotted feature amounts and the statistical degree (range) of dispersion, and a discrimination line is drawn according to the specificity of each grouping, and is distinguished as a plurality of evaluation groups. The distinction of each evaluation group is referred to, for example, by the dynamic postural stability evaluation device 1 such as a "measurement information table", a "subject information table", etc., and according to various attributes of the subject 3, the type of movement task, the numerical value and number of repeated measurement data measured by the repeated actions of the subject 3 for the set task, etc., a discrimination line showing the specificity and features of each group is obtained based on the number and dispersion in the spatial coordinates, and for example, keywords related to each group are labeled.

[0047] The evaluation image shown in FIG. 6 uses a two-dimensional configuration for the multi-dimensional variable space. However, for example, by using more repeated measurement items, indicators, reference values, the attributes of subject 3, etc., a multi-dimensional variable space of two or more dimensions may be configured. Depending on the configuration of the multi-dimensional variable space, an evaluation image showing the posture stability state of a plurality of subjects 3 will be generated.

[0048] The evaluation of the posture stability of a specific subject 3 by the evaluation image, for example, when the evaluation image is generated including the repeated measurement data of a plurality of subjects 3, will be distinguished and displayed by specifying the "subject ID" of the specific subject 3 after generation. Also, when the evaluation image is generated using only the repeated measurement data of a specific subject 3 previously, by processing using the repeated measurement data of the specific subject 3 later, for example, the evaluation of the specific subject 3 may be superimposed and displayed on the previously generated evaluation image.

[0049] Next, referring to FIG. 7, an example of the reference database in the present embodiment will be described.

[0050] <Reference Database> The reference database is stored, for example, in the storage unit 104. The reference database stores the association between the previously acquired past evaluation images and the reference information associated with the past evaluation images. For example, a machine learning model having an association is stored. The reference database may store, for example, past evaluation images and reference information. The association is constructed, for example, by machine learning using a plurality of sets of learning data with past evaluation images and reference information as a set of learning data. As a learning method, deep learning such as a convolutional neural network is used, for example.

[0051] In this case, for example, correlation indicates the degree of connection between many-to-many pieces of information (multiple data points included in past evaluation images, multiple data points included in reference information). Correlation is updated as needed during the machine learning process. That is, correlation represents a function optimized based on past evaluation images (unfolded into a multidimensional variable space) and reference information. Therefore, evaluation results for evaluation images are generated using correlation constructed based on all the results of past evaluations of subject 3's dynamic posture. This makes it possible to generate optimal evaluation results even when subject 3's repeated execution results and subject 3's physical condition and behavior are in various states.

[0052] Furthermore, the system can quantitatively generate optimal evaluation results even when the evaluation image is identical or similar to past evaluation images, or even when it is dissimilar. Additionally, by improving the generalization ability during machine learning, the evaluation accuracy for unknown evaluation images can be improved.

[0053] Furthermore, the correlation may have multiple correlation degrees, which indicate the degree of connection between multiple data points contained in past evaluation images and multiple data points contained in reference information. The correlation degrees can be associated with weight variables, for example, when a machine learning model is constructed using a neural network.

[0054] Past evaluation images contain the same type of information as the evaluation images mentioned above. Past evaluation images include, for example, multiple evaluation images taken when subject 3 was evaluated in the past.

[0055] The reference information is linked to past evaluation images and shows information about the dynamic postural state of subject 3 based on repeated performance. The reference information shows an evaluation based on the dynamic postural state of subject 3 (for example, "normal," "prediction of sports injury," "no problem with returning to sports," "tendency towards high / low risk of XX," "60% chance of developing XX symptoms," etc.), and may also include physical information, response information, preparation information, prediction information, etc., related to the factors of subject 3's dynamic postural state.

[0056] Reference information may include, for example, predictions of dynamic postures that may occur in subject 3, trends in necessary training, and probabilities indicating potential future injuries or symptoms and conditions, as well as necessary training. The specific content included in the reference information can be arbitrarily set.

[0057] The correlation may indicate the degree of connection between past evaluation images and reference information, as shown in Figure 7, for example. In this case, by using correlation, the degree of relationship between each of the multiple data points contained in the past evaluation images ("Image Data A" to "Image Data C" in Figure 7) and the multiple data points contained in the reference information ("Reference A" to "Reference C" in Figure 7) can be linked and stored. Therefore, for example, by linking correlation, it is possible to link multiple data points contained in the reference information to a single data point contained in a past evaluation image, thereby enabling the generation of multifaceted evaluation results.

[0058] Furthermore, the correlation has multiple degrees of correlation, which link multiple data points contained in past evaluation images to multiple data points contained in reference information. The degree of correlation is expressed in three or more stages, such as a percentage, a 10-point scale, or a 5-point scale, and is indicated by line features (e.g., thickness). For example, "image data A" contained in past evaluation images shows a correlation degree AA "75%" with "reference A" contained in reference information, and a correlation degree AB "50%" with "reference B" contained in reference information. In other words, the "degree of correlation" indicates the degree of connection between each image data; for example, a higher degree of correlation indicates a stronger connection between the data. Note that when constructing correlation using the machine learning described above, the correlation may be set to have three or more degree of correlation.

[0059] Past evaluation images may be stored in a reference database by separating them into, for example, past image data A to C and past exercise task information or repetitive movement information. In this case, the degree of correlation is calculated based on the relationship between the combination of past image data and past exercise task information or repetitive movement information and the reference information. In addition to the above, past evaluation images may also be stored in the reference database by separating them into, for example, past measurement information (various sensor data, etc.).

[0060] Furthermore, past evaluation images may include, for example, composite data and similarity scores. The composite data is indicated by a similarity score of three or more levels between past image data, past exercise task information, or repetitive movement information. The composite data may be stored in a reference database in the form of numbers, matrices, or histograms, or it may be stored in the form of images or strings, for example.

[0061] Figure 3(b) is a schematic diagram showing an example of the functions of the dynamic posture stability evaluation device 1. The dynamic posture stability evaluation device 1 comprises an acquisition unit 11, a calculation unit 12, a generation unit 13, an evaluation unit 14, and a feedback provision unit 15, and may also include, for example, an update unit 16. Note that each function shown in Figure 3(b) is realized by the CPU 101 executing a program stored in the storage unit 104, etc., using the RAM 103 as a work area, and may be controlled by artificial intelligence using, for example, a machine learning algorithm.

[0062] <Acquisition Unit 11> The acquisition unit 11 acquires repeated measurement data from, for example, a specific subject 3 and other subjects 3. The acquisition unit 11 acquires the results (repeated measurement data) of repeated execution of an exercise task set for the subject 3, for example, via a plurality of sensors 2 (2a to 2g) connected to the dynamic posture stability evaluation device 1.

[0063] The acquisition unit 11 refers to, for example, the exercise task menu (not shown) for each subject 3 pre-set by the administrator, and enables the acquisition of repetitive measurement data by the sensors 2 (2a to 2g) set in the exercise task menu. The acquisition unit 11 then senses the state of repetitive measurement for the exercise task of subject 3 via the sensors 2 (2a to 2g) and acquires the repetitive measurement data. For example, if the exercise task set for subject 3 is a "single-leg drop landing test" and the connected sensor 2 is a "ground reaction force sensor", the acquisition unit 11 acquires a total of six types of measurement data, for example, numerical values ​​showing the three components of translational ground reaction force: "1. Ground reaction force in the anterior-posterior direction", "2. Ground reaction force in the lateral direction", and "3. Ground reaction force in the vertical direction", and numerical values ​​showing the three components of ground reaction moment around an axis: "1. Ground reaction moment around the anterior-posterior axis", "2. Ground reaction moment around the lateral axis", and "3. Ground reaction moment around the vertical axis".

[0064] The acquisition unit 11, for example, repeatedly measures 6 to 10 times on one leg of the subject 3, alternating between the left and right legs. The acquisition unit 11 acquires repeated measurement data of the six types of ground reaction force data mentioned above, measured over a predetermined period of time (for example, 5 seconds). The repeated measurement data acquired by the acquisition unit 11 may be acquired together with information indicating the acquisition conditions or environment during repeated execution. The repeated measurement data acquired by the acquisition unit 11 may be sequentially stored in the "measurement information table" and the "subject information table".

[0065] The acquisition unit 11 may also include ground reaction force data in the repeated measurement data acquired by, for example, sensors 2 (2a to 2g). The acquisition unit 11 may also acquire other repeated measurement data, such as motion sensor data acquired by a motion sensor, and vital sensor data acquired by a vital sensor.

[0066] <Calculation Unit 12> Based on the repeated measurement data of the subject 3 acquired by the acquisition unit 11, the calculation unit 12 constructs a multidimensional variable space for evaluating the postural stability of the subject 3 and calculates relative feature quantities based on each combination of multiple element variables.

[0067] The calculation unit 12 extracts data portions that are useful for calculating features from acquired repeated measurement data (for example, raw data measured by a ground reaction force sensor). If subject 3 is performing a "single-leg drop landing test," the calculation unit 12 extracts measurement data from subject 3's repeated measurement data for several seconds (for example, 5 seconds) after subject 3 lands on one leg. The calculation unit 12 may also apply a low-pass filter (for example, a cutoff frequency of 70 Hz) to the extracted measurement data for the extracted number of seconds to remove noise in the frequency band above 70 Hz from the extracted measurement data.

[0068] The calculation unit 12 calculates several specific features using, for example, the measurement data after noise reduction. The features calculated by the calculation unit 12 may include, for example, "ground reaction force data" and "center of pressure (COP) point" in the evaluation data used as features when a multidimensional space is constructed in a two-dimensional space. As "ground reaction force data," the calculation unit 12 calculates, for example, the impact cushioning coefficient (peak value of vertical component of ground reaction force / time of appearance, normalized by body weight [N / ms / kg]) which quantifies the cushioning characteristics at the moment of contact, as the first dimension (X axis) from a known group of indicators that show the stability of the dynamic posture of the subject 3. As "center of pressure point," the calculation unit 12 may calculate, for example, the COP trajectory length (for example, between 20ms and 5s after landing of the subject 3, normalized by foot length [%]) which quantifies the magnitude of postural sway of the subject 3, as the second dimension (Y axis) from a group of indicators.

[0069] The calculation unit 12 calculates the "center of foot pressure" as a feature quantity corresponding to the anterior-posterior and lateral directions of the body orientation of the subject 3 upon landing, for example, the X and Y axes of the coordinate system of the ground reaction force sensor. Alternatively, the anterior-posterior center of foot pressure (CoPx) of the subject 3 may be obtained by dividing the ground reaction moment (My) on the lateral axis by the vertical ground reaction force (Fz), and the lateral center of foot pressure (CoPy) may be obtained by dividing the ground reaction moment (Mx) on the anterior-posterior axis by the vertical ground reaction force (Fz).

[0070] The calculation unit 12 may, for example, repeatedly perform the above-described process to calculate the following as feature quantities: "anterior-posterior ground reaction force (Fx)", "lateral ground reaction force (Fy)", "vertical ground reaction force (Fz)", "ground reaction moment around the anterior-posterior axis (Mx)", "ground reaction moment around the lateral axis (My)", "ground reaction moment around the vertical axis (Mz)", "center of pressure in the anterior-posterior direction (CoPx)", or "center of pressure in the lateral direction (CoPy)".

[0071] The calculation unit 12 further calculates multiple feature quantities. The calculation unit 12 constructs a two-dimensional space by using two features (two indicators), for example, "ground reaction force data" and "center of foot pressure," as the feature quantities to be calculated. However, by selecting multiple other indicators (for example, various calculation items, etc.), it is possible to calculate feature quantities of a dimension corresponding to the number of selected indicators. For example, if there are two feature quantities (two indicators) to be calculated for subject 3, the calculation unit 12 calculates them from evaluation target data, which is repeated measurement data of repeated executions corresponding to these two feature quantities (for example, the number of measurements taken for the exercise task, number of repetitions, and repetition time, after noise reduction). This makes it possible to identify the individual characteristics and potential problems of subject 3 and improve the accuracy of the evaluation of dynamic postural stability.

[0072] The calculation unit 12 may calculate more effective relative features based on, for example, multiple element variables (3, 4 indicators, etc.). By calculating effective features, the calculation unit 12 can enable the generation unit 13, described later, to construct a higher-dimensional multi-dimensional variable space, such as 3D or 4D. The calculation unit 12 may also refer to various information, such as subject information of subject 3, to determine which indicators are effective for subject 3, and calculate multiple features and indicators that are effective for subject 3 based on the determination result.

[0073] <Generation Unit 13> The generation unit 13, for example, maps the feature quantities of subject 3 calculated by the calculation unit 12 into a multidimensional variable space and generates an evaluation image that shows the characteristics of subject 3. If the generation unit 13 expands the feature quantities of multiple subjects 3 into the same multidimensional variable space, it may expand them by superimposing them onto the multidimensional variable space with identification displays (e.g., various colorings, different icons, enlarged / smalled displays, etc.) so that each feature quantity of subject 3 can be distinguished. The identification displays by the generation unit 13 should be displayed according to, for example, the number of subjects 3, the feature quantities of subjects 3, indicators, or the trends, quantities, time series, etc. of the data to be evaluated, and how they are displayed is arbitrary.

[0074] If the calculation unit 12 calculates two types of features for the subject 3 (for example, two indicators, "ground reaction force data" and "center of pressure (COP) point"), the generation unit 13 constructs a two-dimensional variable space representing the two types of features, and generates an image in which multiple features of the subject 3 are unfolded as a planar evaluation image, as shown in Figure 6, which will be described later. If the calculation unit 12 calculates ten features, the generation unit 13 maps a feature vector (multidimensional feature) consisting of the ten calculated features to a point in a 10-dimensional space. The generation unit 13 repeatedly unfolds a part or all of the calculated features sequentially into the multidimensional variable space, generating multiple or specific evaluation images of the subject 3.

[0075] If there are 10 features, they will include various features calculated from evaluation target data such as force data, moment data, and center of pressure. The evaluation image generated by the generation unit 13 will be a 10-dimensional information-rich feature vector that includes various aspects of subject 3, such as motor ability, balance ability, fall risk, or injury risk. In other words, if, for example, M subjects 3 are each characterized by a feature vector (multidimensional feature vector with N elements) consisting of N features, an evaluation image will be generated that is equivalent to unfolding M individual characteristics in an N-dimensional variable space. This makes it possible to identify the individual characteristics and potential problems of subject 3 and improve the accuracy of the evaluation of dynamic posture stability.

[0076] <Evaluation Unit 14> The evaluation unit 14 evaluates the stability of the dynamic posture of the subject 3 based on the evaluation images generated by the generation unit 13, for example. The evaluation unit 14 may evaluate the stability of the dynamic posture of each subject 3 based on evaluation images of multiple subjects 3, or a specific subject 3. Based on the evaluation images generated by the generation unit 13, the evaluation unit 14 determines which plane (dimension) to divide the image by, for example, reducing the number of dimensions using principal component analysis and performing linear (or nonlinear) discriminant analysis, sets the main axis, and evaluates the evaluation image along the set axis.

[0077] The evaluation unit 14, for example, uses evaluation images from M subjects 3 that constitute an N-dimensional variable space, based on the evaluation images from M subjects 3 in the generation unit 13. The evaluation unit 14 then performs principal component analysis on the features of the M subjects, which have been expanded into an N-dimensional (N=11, 11 feature quantities) variable space, to obtain M (N-1, 10) new feature vectors from the first principal component to the Nth (11) principal component. The evaluation unit 14 then performs a coordinate transformation from the coordinate system of the N-dimensional (11-dimensional) variable space, which was originally represented by N (11) feature axis axes, to a coordinate system of the N-dimensional (11-dimensional) variable space that uses N (11) principal component axis axes.

[0078] The evaluation unit 14 may, for example, expand the first axis, which is the first principal component, so that M (N-1, 10) features are most varied, and transform the coordinates of the second principal component so that M (N-1, 10) features are most varied in an N-1 dimensional variable space (plane) orthogonal to the first principal component. The evaluation unit 14 may, for example, perform trials of multiple patterns to determine which axis or plane is most effective for setting the N dimensional variable space newly spanned from the first principal component axis to the Nth principal component axis using the coordinate-transformed principal component analysis, determine a priority based on the trial results, and make a judgment based on the determined priority.

[0079] The evaluation unit 14 may determine the target of dynamic postural stability for evaluating subject 3 based, for example, on the attributes of subject 3 acquired by the acquisition unit 11, or on the results of repeated measurement data. The evaluation unit 14 may also evaluate the evaluation image by finding the plane or axis that can most effectively distinguish between "a. points of people with knee ligament injuries" and "b. points of people without injuries" from M people of subject 3. The evaluation unit 14 may also repeatedly search for an N-1 dimensional variable plane (for example, 1 dimension, i.e., a line) that can most accurately separate the features of the two groups to be separated in an N-dimensional (2-dimensional) variable space (for example, N=2, 2-dimensional plane) using, for example, a known linear discriminant analysis, visualize the plane orthogonal to the N-1 dimensional variable plane, and evaluate the dynamic postural stability of subject 3 based on the visualized plane.

[0080] The evaluation unit 14 may divide multiple axes based on an evaluation image that shows a two-dimensional variable space (for example, "a. Points for people with knee ligament injuries" and "b. Points for people without injuries"), and if, for example, the feature quantities of a newly acquired specific subject 3 (the M+1th person) are mapped to the region on the "people with knee ligament injuries" side (for example, each region divided by multiple axes is pre-labeled to give meaning to each region), the evaluation unit 14 may determine the indicators and values ​​of the X and Y axes that constitute the two-dimensional variable space, and the position (value) of the specific subject 3 mapped to the specific region, and evaluate that "the specific subject 3 is at risk of knee injury."

[0081] Furthermore, the evaluation unit 14 may, for example, evaluate (group) multiple subjects 3 based on evaluation images generated by the generation unit 13, according to predetermined purposes and content. Alternatively, it may use pre-generated evaluation images based on the features of multiple subjects 3 to evaluate the stability of dynamic posture for newly acquired repeated measurement data of a specific subject 3. By evaluating the mapping of a specific subject 3 to regions in a two-dimensional space separated by axes such as "a. people with large postural sway" and "b. people with small postural sway," the evaluation unit 14 can present suggestions and recommendations for balance training, suggestions and recommendations for training facilities and trainers, etc.

[0082] Furthermore, the evaluation unit 14 can present "fall prevention training," "suggestions from medical institutions," etc., by evaluating the mapping of a specific subject 3 to regions in a three-dimensional space that are separated by axes such as "a. elderly people," "b. people with large postural sway," and "c. people with small postural sway."

[0083] Furthermore, the evaluation unit 14 may evaluate the postural stability of a subject 3 based on information including, for example, evaluation images of multiple subjects 3, and at least one of the attributes, personal characteristics, or time-series changes of a specific subject 3. Based on the evaluation images generated by the generation unit 13, the evaluation unit 14 may perform principal component analysis by referring to, for example, a "measurement information table" or a "subject information table" to identify multiple axes and each region separated by the axes. This makes it possible to set evaluation axes and regions based on other parameters and indicators in addition to the evaluation images generated by the generation unit 13, thereby identifying the personal characteristics and potential problems of the subject 3 in detail and improving the accuracy of dynamic postural stability evaluation.

[0084] Furthermore, the evaluation unit 14 includes a machine learning evaluation unit that, for example, uses a machine learning model to evaluate the postural stability of a specific subject 3. In the machine learning evaluation unit, the machine learning model is constructed using machine learning with multiple training data. As a training method, deep learning such as a convolutional neural network is used. The machine learning evaluation unit, for example, refers to a reference database and uses a machine learning model constructed based on the correlation between the evaluation image generated by the generation unit 13, previously acquired past evaluation images, and reference information associated with past evaluation images to evaluate whether the evaluation image is identical or similar to past evaluation images, and quantitatively generates the optimal evaluation result. This makes it possible to improve the evaluation accuracy for unknown evaluation images.

[0085] <Feedback Provision Unit 15> Furthermore, the evaluation unit 14 includes a feedback provision unit 15 that provides feedback information for posture improvement to a specific subject 3 based on the evaluation results. The feedback from the feedback provision unit 15 is provided based on the mapped coordinate information, numerical values, etc., when the evaluation of the specific subject 3 by the evaluation unit 14 is mapped to regions in a two-dimensional space that are separated by axes such as "a. People with large postural sway" and "b. People with small postural sway".

[0086] The various types of feedback information provided by the feedback provision unit 15 may be stored, for example, in the server 6 or storage unit 104. The feedback provision unit 15 refers to, for example, a "feedback information table (not shown)" and obtains feedback information related to the evaluated "measurement results / ID information". The feedback information may include, for example, "suggestions and recommendations for balance training", "suggestions and recommendations from training facilities and trainers", "fall prevention training", or "suggestions from medical institutions", and the feedback information may be obtained by referring to a third-party database linked to the dynamic posture stability evaluation system 100, for example.

[0087] <Update Unit 16> The update unit 16 reflects the relationship in the correlation when it acquires new relationships, such as repeated measurement data, past evaluation images, and reference information. The update unit 16 may also update the correlation stored in the reference database based on the evaluation results, for example, when the dynamic posture stability evaluation device 1 acquires the results of an evaluation of the accuracy of the evaluation results by an administrator or the like based on the evaluation results.

[0088] Each of the other dynamic posture stability evaluation devices 1 and sensors 2 (2a to 2g) is equipped with a memory unit. The memory unit retrieves various information, such as a reference database stored in the storage unit 104, as needed. The memory unit stores various information acquired or generated by each of the components 11, 13 to 17 in the storage unit 104.

[0089] <Display Unit 109> The display unit 109 displays the exercise task for the subject 3, repeated measurement data from the sensors 2 (2a to 2g), evaluation images, other related information, and various other information. The display unit 109 may also display various types of information depending on the data and content (e.g., text, messages, images, patterns, etc.) and form (normal display, flashing, etc.) measured by the sensors 2 (2a to 2g).

[0090] <Sensors 2, 2a-2g> For example, if the exercise task for subject 3 is a "single-leg drop landing test," it is preferable that sensors 2 (2a-2g) be known ground reaction force sensors that measure ground reaction force data. Alternatively, if measuring the movement of the subject 3's whole body or specific body parts, known motion sensors may be used, and if measuring the subject 3's body temperature or vital signs, known near-infrared or non-contact vital signs sensors may be used. Multiple sensors 2 (2a-2g) may be configured to acquire repeated measurement data of subject 3 individually or in conjunction with other sensors 2 (2a-2g).

[0091] <Communication Network 4> Communication network 4 is, for example, the Internet, a local network in a closed environment, etc., to which the dynamic attitude stability evaluation device 1, multiple sensors 2, etc., are connected via a communication circuit. Communication network 4 may be composed of a so-called optical fiber communication network. In addition, communication network 4 may be implemented using a known communication network such as a wired communication network or a wireless communication network.

[0092] <Other Terminal 5> Other terminal 5 is implemented as an electronic device, similar to the dynamic posture stability evaluation device 1. Other terminal 5 represents, for example, a central control unit that can communicate with multiple dynamic posture stability evaluation devices 1. Other terminal 5 can connect to multiple dynamic posture stability evaluation devices 1 and acquire evaluation results generated by each dynamic posture stability evaluation device 1. This makes it possible to output evaluation results based on repeated measurement data of the subject 3 measured at multiple locations, provide feedback to the subject 3, and improve the subject 3's dynamic posture.

[0093] <Server 6> Server 6 stores various types of information, for example. Server 6 stores various measurement information and subject information sent via the communication network 4, as well as various repetitive measurement data about the subject 3 measured by various sensors 2 (2a to 2g). Server 6 may store the same information as the storage unit 104, and may store the transmission and reception results of various information, data, and evaluation results performed by one or more dynamic posture stability evaluation devices 1 via the communication network 4, or the dynamic posture stability evaluation device 1 may use Server 6 instead of Storage Unit 104.

[0094] (An example of the operation of the dynamic posture stability evaluation system 100) Next, an example of the operation of the dynamic posture stability evaluation system 100 in this embodiment will be described. Figure 8 is a flowchart showing an example of the operation of the dynamic posture stability evaluation system 100 in this embodiment.

[0095] <Acquisition means S110> First, the acquisition unit 11 acquires repeated measurement data from the subject 3 (acquisition means S110). The acquisition unit 11 acquires repeated measurement data of an exercise task by, for example, a specific subject 3 or other subject 3. The repeated measurement data of subject 3 acquired by the acquisition unit 11 may be, for example, repeated measurement data of multiple subject 3. The acquisition unit 11 acquires the results (repeated measurement data) of repeated execution of an exercise task set for a specific or multiple subject 3 via, for example, a plurality of sensors 2 (2a to 2g) connected to the dynamic posture stability evaluation device 1.

[0096] The acquisition unit 11 stores the acquired repeated measurement data of the subject 3 in, for example, the "measurement information table" of the server 6. The acquisition unit 11 may also acquire, for example, various physical information (body temperature, heart rate, health checkup results, etc.) and environmental information (temperature, room temperature, humidity, etc.) related to the subject 3 performing the exercise task, along with the acquired repeated measurement data of the subject 3, as various information related to the exercise task menu.

[0097] For example, if the exercise task menu set for subject 3 is a "single-leg drop landing test," the acquisition unit 11 uses sensor 2 (ground reaction force sensor) to acquire repeated measurement data such as "ground reaction force in the anterior-posterior direction," "ground reaction force in the lateral direction," "ground reaction force in the vertical direction," "ground reaction moment around the anterior-posterior axis," "ground reaction moment around the lateral axis," and "ground reaction moment around the vertical axis."

[0098] The acquisition unit 11, for example, repeatedly measures 6 to 10 times on one leg of the subject 3, and acquires repeated measurement data of 6 types of ground reaction force data measured over a predetermined period of time (for example, 5 seconds). The repeated measurement data acquired by the acquisition unit 11 may be sequentially stored in the "measurement information table" and the "subject information table".

[0099] <Calculation means S120> Next, the calculation unit 12 constructs a multidimensional space for evaluating the postural stability of the subject 3 based on the repeated measurement data acquired by the acquisition unit 11, and calculates relative feature quantities based on each combination of multiple element variables (calculation means S120). The calculation unit 12 extracts, for example, the data portion that is effective for calculating feature quantities from the acquired repeated measurement data. In the case of the "single-leg drop landing test" by the subject 3, the calculation unit 12 extracts, for example, the measurement data for the few seconds (for example, 5 seconds) after the subject 3 lands on one leg from the repeated measurement data of the subject 3.

[0100] The calculation unit 12 may also perform noise reduction on the acquired repeated measurement data and then calculate feature quantities using the corrected measurement data. If the feature quantities to be calculated are two features (two indicators), namely "ground reaction force data" and "center of foot pressure," the calculation unit 12 constructs a two-dimensional space corresponding to the two feature quantities.

[0101] The calculation unit 12 may also calculate feature quantities with dimensions corresponding to the number of selected indicators if, for example, multiple other indicators (e.g., various calculation items, etc.) are selected by the administrator. The calculation unit 12 calculates more relative feature quantities based on multiple element variables, for example. The calculation unit 12 refers to various information, such as subject information of subject 3, for example. The calculation unit 12 may also determine which indicators are effective for subject 3 and calculate multiple feature quantities and multiple indicators that are effective for subject 3.

[0102] When the calculation unit 12 calculates each feature quantity in a two-dimensional variable space composed of, for example, "ground reaction force data" and "center of pressure," it calculates the impact cushioning coefficient (peak value of the vertical component of the ground reaction force / time of its appearance, normalized by body weight [N / ms / kg]) which quantifies the cushioning characteristics at the moment of contact, as the first dimension (X axis), and calculates the COP trajectory length (for example, between COP 20ms and 5s after landing of subject 3, normalized by foot length [%]) as the "center of pressure" which quantifies the magnitude of postural sway of subject 3, as the second dimension (Y axis).

[0103] The feature index calculated by the calculation unit 12 may be selected in advance by the administrator from a predetermined group of indexes, or, for example, the index may be selected based on the results of preprocessing of the repeated measurement data of the subject 3. The calculation unit 12 may configure a multidimensional variable space according to the number of selected indexes. If there are two indexes, it will configure a two-dimensional variable space, but if there are three or four indexes, it will configure a three-dimensional variable space and a four-dimensional variable space, respectively, and calculate the feature in each multidimensional variable space.

[0104] <Generation means S130> Next, the generation unit 13 expands the feature quantities into a multidimensional variable space and generates an evaluation image that shows the state of posture stability (Generation means S130). The generation unit 13 expands (maps) the feature quantities of the subject 3 calculated by the calculation unit 12 into a multidimensional variable space and generates an evaluation image that shows the characteristics of the subject 3. The generation unit 13 expands the feature quantities of the subject 3 calculated by the calculation unit 12 into a multidimensional variable space (for example, a two-dimensional variable space, etc.). If the feature quantities of the subject 3 calculated by the calculation unit 12 are of two types (for example, two indicators, "ground reaction force data" and "center of pressure (COP) point"), the generation unit 13 constructs a two-dimensional variable space that shows the two types of characteristics and generates an evaluation image on which the multiple feature quantities of the subject 3 are expanded as a planar evaluation image.

[0105] If, for example, the calculation unit 12 calculates 10 features, the generation unit 13 maps the feature vector (multidimensional feature) consisting of the 10 calculated features to a point in a 10-dimensional space. The generation unit 13 then characterizes, for example, M subjects 3 using a feature vector consisting of N features, constructs an N-dimensional variable space, and generates an evaluation image in which the M individual characteristics are displayed.

[0106] Furthermore, if, for example, the calculation unit 12 calculates multiple features, the generation unit 13 maps the feature vector (multidimensional feature) consisting of the calculated multiple features to a single point in the multidimensional variable space. The generation unit 13 sequentially expands a part or all of the calculated multiple features into the corresponding multidimensional variable space to generate evaluation images for multiple or specific subjects 3. The evaluation images may be in a two-dimensional variable space (plane), or for example, a three-dimensional variable space (solid), etc.

[0107] <Evaluation means S140> Next, the evaluation unit 14 evaluates the stability of the subject's dynamic posture based on the evaluation image (evaluation means S140). Based on the evaluation image of the subject 3 generated in the generation unit 13, the evaluation unit 14 sets the main evaluation axes by reducing the number of dimensions, for example by principal component analysis, and performing linear (or nonlinear) discriminant analysis, and evaluates the stability of the subject's dynamic posture.

[0108] For example, if the generation unit 13 generates evaluation images that constitute an N-dimensional variable space from M subjects 3, the evaluation unit 14 applies principal component analysis to the M subjects' features expanded into an N-dimensional (N=11, 11 features) variable space, and obtains M (N-1, 10) new feature vectors for the M subjects up to the Nth (11) principal component as the first principal component. The evaluation unit 14 then performs a coordinate transformation from the coordinate system of the N-dimensional (11-dimensional) variable space, which was represented by the axes of N (11) features, to a coordinate system of the N-dimensional (11-dimensional) variable space that uses N (11) principal component axes.

[0109] If the evaluation image generated by the generation unit 13 has a two-dimensional variable space structure (two dimensions) with two indicators, the evaluation unit 14 may, for example, expand and set the first axis, which is the first principal component, so that M (10) features are most varied, and then perform a coordinate transformation on the second axis, which is the second principal component, in an N-1 dimensional variable space (plane) orthogonal to the first principal component so that M (10) features are most varied.

[0110] Furthermore, the evaluation unit 14 may include, for example, evaluation images of multiple subjects 3 and evaluate the stability of the subject's dynamic posture based on information including at least one of the attributes, personal characteristics, or time-series changes of a specific subject 3. Based on the evaluation images generated by the generation unit 13, the evaluation unit 14 refers to, for example, a "measurement information table" or a "subject information table" and performs principal component analysis to identify multiple axes and each region separated by the axes. This makes it possible to set evaluation axes and regions based on other parameters and indicators, to identify the personal characteristics and potential problems of the subject 3 in detail, and to improve the accuracy of the evaluation of the stability of dynamic posture.

[0111] The evaluation unit 14 may further include a machine learning evaluation unit that uses a machine learning model to evaluate the stability of the dynamic posture of a specific subject 3. The machine learning evaluation unit constructs the machine learning model using machine learning with multiple training data. The machine learning evaluation unit, for example, refers to a reference database and uses a machine learning model constructed based on the relationship between the evaluation image generated by the generation unit 13, previously acquired past evaluation images, and reference information associated with past evaluation images to evaluate whether the evaluation image is identical or similar to past evaluation images. The machine learning evaluation unit can quantitatively generate optimal evaluation results and improve the evaluation accuracy for unknown evaluation images.

[0112] <Feedback Provisioning Means S150> Next, the feedback provisioning unit 15 provides feedback information for posture improvement to a specific subject 3 based on the evaluation results from the evaluation unit 14. When the evaluation of a specific subject 3 by the evaluation unit 14 is mapped to regions in a two-dimensional space separated by axes such as "a. People with large postural sway" and "b. People with small postural sway", the feedback provisioning unit 15 refers to, for example, a "feedback information table (not shown)" based on the mapped coordinate information, numerical values, etc., and obtains feedback information related to the evaluated "measurement result / ID information".

[0113] The information that the feedback unit 15 provides to the subject 3 may include, for example, "suggestions and recommendations for balance training," "suggestions and recommendations from training facilities and trainers," "fall prevention training," or "suggestions from medical institutions." The feedback unit 15 may also obtain feedback information by referring to a third-party database linked to, for example, the dynamic posture stability evaluation system 100.

[0114] Furthermore, according to this embodiment, a dynamic posture stability evaluation method for evaluating posture stability through repeated measurements by a subject 3 can be provided by: an acquisition step by an acquisition means S110 for acquiring repeated measurement data by the subject 3; a calculation step by a calculation means S120 for calculating relative feature quantities from a multidimensional variable space and element variables; a generation step by a generation means S130 for mapping feature quantities and generating an evaluation image; an evaluation step by an evaluation means S140 for evaluating the posture stability of a specific subject 3; and a feedback provision step by a feedback provision means S150 for providing feedback information for posture improvement to a specific subject 3.

[0115] This completes the operation of the dynamic posture stability evaluation system 100 in this embodiment. The timing of when the update unit 16 performs an update is arbitrary.

[0116] While embodiments of the present invention have been described, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents.

[0117] 1: Dynamic posture stability evaluation device 2: Sensor 2a-2g: Sensor 3: Subject 4: Communication network 5: Other terminals 6: Server 10: Housing 11: Acquisition unit 12: Calculation unit 13: Generation unit 14: Evaluation unit 15: Feedback provision unit 16: Update unit 100: Dynamic posture stability evaluation system 101: CPU 102: ROM 103: RAM 104: Storage unit 105: I / F 106: I / F 107: I / F 108: Input unit 109: Display unit 110: Internal bus A: Location B: Location S110: Acquisition means S120: Calculation means S130: Generation means S140: Evaluation means S150: Feedback provision means

Claims

1. A dynamic posture stability evaluation system for evaluating posture stability through repeated measurements by a subject, comprising: acquisition means for acquiring repeated measurement data by the subject; calculation means for constructing a multidimensional variable space for evaluating the subject's posture stability based on the repeated measurement data and calculating relative feature quantities based on each combination of multiple element variables; generation means for expanding the feature quantities into the multidimensional variable space and generating an evaluation image showing the state of posture stability; and evaluation means for evaluating the subject's posture stability based on the evaluation image.

2. The dynamic posture stability evaluation system according to claim 1, characterized in that the repeated measurement data acquired by the acquisition means includes ground reaction force data.

3. The dynamic posture stability evaluation system according to claim 1, characterized in that the evaluation means includes evaluation images of a plurality of subjects and evaluates posture stability based on information including at least one of the attributes, personal characteristics, or time-series changes of a specific subject.

4. The dynamic posture stability evaluation system according to claim 1 or 3, characterized in that the evaluation means further includes a machine learning evaluation means that evaluates the posture stability of a specific subject using a machine learning model.

5. The dynamic posture stability evaluation system according to claim 1, further comprising a feedback providing means for providing feedback information for posture improvement to the specific subject based on the evaluation results from the evaluation means.

6. A method for evaluating dynamic posture stability, characterized by comprising each step performed by the dynamic posture stability evaluation system described in any one of claims 1 to 5.

Images