JP2023017652A - Walking ability determination device, walking ability determination method, walking ability determination program, model generation device, model generation method, and model generation program - Google Patents

Abstract

To appropriately evaluate walking ability of an object person by a mechanical method.SOLUTION: A walking ability determination device acquires object data generated by observing walking of an object person, and determines walking ability of the object person on the basis of the acquired object data using a trained determination model generated by machine learning. The object data is configured so as to include at least one of first information indicating an area on the coronal plane of a trajectory of a human body's center of gravity at the time of walking of the object person, second information indicating an area on the transverse plane of the trajectory, third information indicating a degree of difference in the time for support with each of right and left legs in walking, fourth information indicting a degree of difference in the distance of movement while support is made with each of right and left legs in walking, and fifth information indicating a degree of repeating a walking motion during a fixed time.SELECTED DRAWING: Figure 1

Description

The present invention relates to a walking ability determination device, a walking ability determination method, a walking ability determination program, a model generation device, a model generation method, and a model generation program.

Conventionally, various methods for evaluating walking ability have been developed. FAC (Functional Ambulation Categories) is known as one of the evaluation methods (Non-Patent Document 1). FAC is an evaluation method that uses a walkway or stairs of about 15 m and evaluates a subject's walking ability into 6 stages based on motion observation by medical staff.

"Functional Ambulation Categories (FAC)", [online], [searched on May 11, 2018], Internet <URL: http://gunma-pt.com/wp-content/uploads/2018/04/Functional -Ambulation-Categories-FAC.pdf＞

In the conventional evaluation method, a skilled medical worker carefully observes the subject's gait and evaluates the subject's walking ability based on the observation results. However, it takes time and other costs to train medical personnel who have acquired such abilities. In addition, since the evaluation is performed by visual observation, there is a problem that the evaluation results may vary depending on the knowledge, experience, etc. of the medical staff.

SUMMARY OF THE INVENTION In one aspect, the present invention has been made in view of such circumstances, and an object thereof is to provide a technique for appropriately evaluating a subject's walking ability by a mechanical method.

The present invention adopts the following configurations in order to solve the above-described problems.

That is, a walking ability determination device according to one aspect of the present invention is a data acquisition unit that acquires target data generated by observing walking of a subject, and the target data is obtained when the subject walks. 1st information indicating the coronal area of the trajectory of the center of gravity of the human body, 2nd information indicating the area of the cross section of the trajectory, and 3rd information indicating the degree of difference in the time supported by one of the left and right legs during walking , at least one of fourth information indicating the degree of difference in the distance moved while being supported by one of the left and right legs in walking, and fifth information indicating the degree of repetition of the walking motion for a predetermined time. A data acquisition unit, a determination unit that determines the walking ability of the subject based on the acquired target data using a trained determination model generated by machine learning, and the walking of the subject. an output unit that outputs a result of determining the ability.

According to experimental examples described later, the present inventors found that, as explanatory variables of a trained model generated by machine learning, the area of the trajectory of the center of gravity of the human body during walking on the coronal plane, the area of the trajectory on the cross section, the walking at least one of the following: the degree of difference in the length of time that the left and right legs are supported during walking, the degree of difference in the distance that the person moves while being supported by one of the left and right legs during walking, and the degree of repetition of the walking motion for a predetermined period of time. By using , it was found that the walking ability of the subject can be estimated with high accuracy. Therefore, according to the said structure, a subject's walking ability can be evaluated appropriately by a mechanical method.

In the walking ability determination device according to the above aspect, the target data may be configured to include all of the first information, the second information, the third information, the fourth information, and the fifth information. . According to this configuration, by using all five pieces of information as explanatory variables, the walking ability of the subject can be estimated with higher accuracy.

A model generation device according to one aspect of the present invention is a data acquisition unit that acquires a plurality of learning data sets respectively generated by observing gait of a plurality of subjects, wherein each learning data set is , a combination of training data and correct labels, wherein the training data of each learning data set includes first information indicating the area on the coronal plane of the trajectory of the center of gravity of the human body during walking of each subject, the trajectory 2nd information indicating the cross-sectional area of the , 3rd information indicating the degree of difference in the time of supporting with one of the left and right legs during walking, and the difference in the distance traveled while supporting with one of the left and right legs during walking and at least one of fifth information indicating the degree of repetition of the walking motion for a predetermined period of time. A learning data acquisition unit configured to indicate a true value of walking ability of a person; and a learning processing unit that performs machine learning of a judgment model using a plurality of acquired learning data sets, In the machine learning, for each of the learning data sets, the judgment model is used so that the result of judging the walking ability of each subject by the judgment model based on the training data conforms to the true value indicated by the correct label. and a learning processing unit configured by training. According to this configuration, it is possible to generate a trained determination model for appropriately evaluating the walking ability of the subject by a mechanical method.

In the model generation device according to the above aspect, the training data of each learning data set includes all of the first information, the second information, the third information, the fourth information, and the fifth information. may be configured to According to this configuration, by using all five pieces of information as explanatory variables, it is possible to generate a trained determination model capable of more accurately estimating the walking ability of the subject.

Further, as another aspect of each of the walking ability determination device and the model generation device according to each of the above embodiments, one aspect of the present invention is an information processing method that realizes all or part of each of the above components. Alternatively, it may be a program, a storage medium storing such a program and readable by a computer, other device, machine, or the like, or an information processing system. Here, a computer-readable storage medium is a medium that stores information such as a program by electrical, magnetic, optical, mechanical, or chemical action. A walking ability determination system (information processing system) may be configured by the walking ability determination device and the model generation device according to any one of the above embodiments.

For example, a walking ability determination method according to one aspect of the present invention is a step of acquiring target data generated by observing a subject's walking by a computer, wherein the target data is the walking ability of the subject. First information indicating the area on the coronal plane of the trajectory of the center of gravity of the human body at the time, Second information indicating the area on the transverse plane of the trajectory, and Second information indicating the degree of difference in the time supported by one of the left and right legs during walking. 3 information, at least one of fourth information indicating the degree of difference in the distance traveled while being supported by one of the left and right legs in walking, and fifth information indicating the degree of repetition of the walking motion for a predetermined time. determining the walking ability of the subject based on the obtained subject data using a trained determination model generated by machine learning; and the walking ability of the subject. and a step of outputting the result of determining the information processing method.

For example, a program for determining walking ability according to one aspect of the present invention is a step of obtaining, in a computer, target data generated by observing the walking of a subject, wherein the target data is the walking of the subject. First information indicating the area on the coronal plane of the trajectory of the center of gravity of the human body at the time, Second information indicating the area on the transverse plane of the trajectory, and Second information indicating the degree of difference in the time supported by one of the left and right legs during walking. 3 information, at least one of fourth information indicating the degree of difference in the distance traveled while being supported by one of the left and right legs in walking, and fifth information indicating the degree of repetition of the walking motion for a predetermined time. determining the walking ability of the subject based on the obtained subject data using a trained determination model generated by machine learning; and the walking ability of the subject. and a step of outputting the result of determining the above.

For example, a model generation method according to one aspect of the present invention is a step in which a computer acquires a plurality of learning data sets respectively generated by observing the gait of each of a plurality of subjects, The data set is composed of a combination of training data and correct labels, and the training data of each learning data set is first information indicating the area on the coronal plane of the trajectory of the center of gravity of the human body during walking of each subject. , the second information indicating the area of the cross section of the trajectory, the third information indicating the degree of difference in the time of supporting by one of the left and right legs during walking, and the movement while supporting by one of the left and right legs during walking. It is configured to include at least one of fourth information indicating the degree of difference in distance and fifth information indicating the degree of repetition of the walking motion for a predetermined time, and the correct label of each learning data set includes the performing machine learning of a decision model using a plurality of acquired learning data sets, configured to indicate the true value of walking ability of each subject, wherein trains the judgment model for each learning data set so that the result of judging the walking ability of each subject by the judgment model based on the training data conforms to the true value indicated by the correct label. An information processing method comprising:

For example, the model generation program according to one aspect of the present invention is a step of acquiring a plurality of learning data sets generated by observing the gait of each of a plurality of subjects in a computer, The data set is composed of a combination of training data and correct labels, and the training data of each learning data set is first information indicating the coronal area of the trajectory of the center of gravity of the human body during walking of each subject; Second information indicating the area of the cross section of the trajectory, Third information indicating the degree of difference in the time of support by one of the left and right legs during walking, Distance moved while supporting by one of the left and right legs during walking and at least one of fifth information indicating the degree of repetition of the walking motion for a predetermined period of time. and performing machine learning of a decision model using the plurality of training data sets obtained, wherein the machine learning comprises: and training the judgment model for each learning data set so that the result of judging the walking ability of each subject by the judgment model based on the training data conforms to the true value indicated by the correct label. A program for executing a step and a.

ADVANTAGE OF THE INVENTION According to this invention, the technique which evaluates a subject's walking ability appropriately by a mechanical method can be provided.

FIG. 1 schematically shows an example of a scene to which the present invention is applied. FIG. 2 schematically shows an example of the hardware configuration of the model generation device according to the embodiment. FIG. 3 schematically shows an example of the hardware configuration of the walking ability determination device according to the embodiment. FIG. 4 schematically shows an example of the software configuration of the model generation device according to the embodiment. FIG. 5 schematically shows an example of the software configuration of the walking ability determination device according to the embodiment. 6 is a flowchart illustrating an example of a processing procedure of the model generation device according to the embodiment; FIG. FIG. 7 shows the measurement result of the subject's walking by the acceleration sensor. FIG. 8A shows the trajectory (coronal plane) of the center of gravity of the human body during walking obtained by analyzing the measurement results of FIG. FIG. 8B shows the trajectory (coronal plane) of the center of gravity of the human body during walking obtained by applying the filtering process to the calculation result of FIG. 8A. FIG. 9A shows the trajectory (cross section) of the human center of gravity during walking obtained by analyzing the measurement results of FIG. FIG. 9B shows the trajectory (cross section) of the human center of gravity during walking obtained by applying the filtering process to the calculation result of FIG. 9A. 10 is a flowchart illustrating an example of a processing procedure of the walking ability determination device according to the embodiment; FIG. FIG. 11 shows the result of calculating the contribution rate of each information.

Hereinafter, an embodiment (hereinafter also referred to as "this embodiment") according to one aspect of the present invention will be described based on the drawings. However, this embodiment described below is merely an example of the present invention in every respect. It goes without saying that various modifications or variations can be made without departing from the scope of the invention. In carrying out the present invention, a specific configuration according to the embodiment may be appropriately employed. Although the data appearing in this embodiment are explained in terms of natural language, more specifically, they are specified in computer-recognizable pseudo-language, commands, parameters, machine language, and the like.

§1 Application Example FIG. 1 schematically illustrates an example of a scene to which the present invention is applied. As shown in FIG. 1 , a determination system 100 according to this embodiment includes a model generation device 1 and a walking ability determination device 2 .

The model generation device 1 according to this embodiment is a computer configured to generate a trained determination model 6 that can be used for evaluating the walking ability of a subject. The model generation device 1 acquires a plurality of learning data sets 3 generated by observing the walking of each of a plurality of subjects.

Each learning data set 3 is composed of a combination of training data 31 and correct labels 32 . The training data 31 of each learning data set 3 includes first information indicating the area on the coronal plane of the trajectory of the center of gravity of the human body during walking of each subject, second information indicating the area of the trajectory on the cross section, walking 3rd information indicating the degree of difference in the time to support with one of the left and right legs in walking, fourth information indicating the degree of difference in the distance moved while supporting with one of the left and right legs in walking, and for a predetermined time at least one of the fifth information indicating the degree to which the walking motion is repeated. The correct label 32 of each training data set 3 is configured to indicate the true value of the walking ability of each subject.

The model generation device 1 performs machine learning of the judgment model 6 using the plurality of acquired learning data sets 3 . In machine learning, for each learning data set 3, the judgment model 6 is trained so that the result of judging the walking ability of each subject by the judgment model 6 based on the training data 31 conforms to the true value indicated by the correct label 32. It is constructed by As a result, a trained determination model 6 that has acquired the ability to estimate the walking ability of the subject can be generated based on at least one of the first information to the fifth information.

On the other hand, the walking ability determination device 2 is a computer configured to use the trained determination model 6 to determine the walking ability of the subject. The walking ability determination device 2 acquires target data 221 generated by observing the walking of the target person. A subject is a person whose walking ability is to be determined. The subject may be any one of the plurality of subjects at the time of the machine learning, or may be a person other than the plurality of subjects.

The target data 221 includes first information indicating the area of the trajectory of the center of gravity of the human body on the coronal plane when the subject walks, second information indicating the area of the trajectory on the cross section, and support by one of the left and right legs during walking. third information indicating the degree of difference in the time taken to walk, fourth information indicating the degree of difference in the distance traveled while being supported by one of the left and right legs during walking, and the degree of repetition of the walking motion within a predetermined time. It is configured to include at least one of the fifth information. The information contained in the target data 221 is configured to correspond to the information contained in the training data 31 described above.

The walking ability determination device 2 determines the walking ability of the subject based on the acquired target data 221 using the trained determination model 6 generated by the machine learning. Then, the walking ability determination device 2 outputs the result of determining the walking ability of the subject.

As described above, in this embodiment, at least one of the first information to the fifth information is adopted as an explanatory variable for estimating walking ability. Thereby, a subject's walking ability can be evaluated appropriately by a mechanical method.

In addition, in the example of FIG. 1, the model generation device 1 and the walking ability determination device 2 are connected to each other via a network. The type of network may be appropriately selected from, for example, the Internet, wireless communication network, mobile communication network, telephone network, dedicated network, and the like. However, the method of exchanging data between the model generation device 1 and the walking ability determination device 2 need not be limited to such an example, and may be appropriately selected according to the embodiment. As another example, data may be exchanged between the model generation device 1 and the walking ability determination device 2 using a storage medium.

Also, in the example of FIG. 1, the model generation device 1 and the walking ability determination device 2 are configured by separate computers. However, the configuration of the determination system 100 according to this embodiment does not have to be limited to such an example, and may be appropriately determined according to the embodiment. As another example, the model generation device 1 and the walking ability determination device 2 may be an integrated computer. As still another example, at least one of the model generation device 1 and the walking ability determination device 2 may be configured by a plurality of computers.

Any sensor may be used to observe the walking of the subject and the subject. The type of the sensor is not particularly limited as long as it can analyze the information used as the explanatory variable among the first information to the fifth information, and may be appropriately selected according to the embodiment. The sensors may be, for example, accelerometers, motion capture, cameras (eg, multiple cameras, depth cameras, stereo cameras, etc.), and the like. Also, the sensors may be appropriately installed so as to be able to analyze the information used for the explanatory variables. As an example, when an acceleration sensor is used as the sensor, the acceleration sensor may be worn on the waist of the subject/subject.

§2 Configuration example [Hardware configuration]
<Model generator>
FIG. 2 schematically illustrates an example of the hardware configuration of the model generating device 1 according to this embodiment. As shown in FIG. 2, the model generation device 1 according to the present embodiment includes a control unit 11, a storage unit 12, a communication interface 13, an external interface 14, an input device 15, an output device 16, and a drive 17 which are electrically connected. It is a computer that has been In addition, in FIG. 2, the communication interface and the external interface are described as "communication I/F" and "external I/F." The same notation is also used in FIG. 3 to be described later.

The control unit 11 includes a CPU (Central Processing Unit), which is a hardware processor, RAM (Random Access Memory), ROM (Read Only Memory), etc., and is configured to execute information processing based on programs and various data. be. The storage unit 12 is an example of memory, and is configured by, for example, a hard disk drive, a solid state drive, or the like. In this embodiment, the storage unit 12 stores various information such as the model generation program 81, the plurality of learning data sets 3, the learning result data 125, and the like.

The model generation program 81 is a program for causing the model generation device 1 to execute information processing (FIG. 6) regarding generation of the trained judgment model 6 (machine learning of the judgment model 6), which will be described later. The model generation program 81 includes a series of instructions for the information processing. Multiple learning data sets 3 are used to generate a trained decision model 6 . The learning result data 125 indicates information regarding the trained judgment model 6 . In this embodiment, the learning result data 125 is generated as a result of executing the model generation program 81. FIG. Details will be described later.

The communication interface 13 is, for example, a wired LAN (Local Area Network) module, a wireless LAN module, or the like, and is an interface for performing wired or wireless communication via a network. The model generation device 1 can use the communication interface 13 to perform data communication with other information processing devices via a network.

The external interface 14 is, for example, a USB (Universal Serial Bus) port, a dedicated port, or the like, and is an interface for connecting with an external device. The type and number of external interfaces 14 may be arbitrarily selected. When using sensors to generate each training data set 3 in the model generation device 1 , the sensors may be connected via the communication interface 13 or the external interface 14 .

The input device 15 is, for example, a device for performing input such as a mouse and a keyboard. Also, the output device 16 is, for example, a device for outputting such as a display and a speaker. An operator such as a user can operate the model generation device 1 by using the input device 15 and the output device 16 .

The drive 17 is, for example, a CD drive, a DVD drive, or the like, and is a drive device for reading various information such as programs stored in the storage medium 91 . The storage medium 91 stores information such as programs by electrical, magnetic, optical, mechanical or chemical action so that computers, other devices, machines, etc. can read various information such as programs. It is a medium that accumulates by At least one of the model generation program 81 and the learning data set 3 may be stored in the storage medium 91 . The model generation device 1 may acquire at least one of the model generation program 81 and the learning data set 3 from the storage medium 91 . In FIG. 2, as an example of the storage medium 91, a disk-type storage medium such as a CD or DVD is illustrated. However, the type of storage medium 91 is not limited to the disk type, and may be other than the disk type. As a storage medium other than the disk type, for example, a semiconductor memory such as a flash memory can be cited. The type of drive 17 may be arbitrarily selected according to the type of storage medium 91 .

Regarding the specific hardware configuration of the model generation device 1, it is possible to omit, replace, and add components as appropriate according to the embodiment. For example, control unit 11 may include multiple hardware processors. The hardware processor may consist of a microprocessor, a field-programmable gate array (FPGA), or the like. The storage unit 12 may be configured by RAM and ROM included in the control unit 11 . At least one of the communication interface 13, the external interface 14, the input device 15, the output device 16 and the drive 17 may be omitted. The model generation device 1 may be composed of a plurality of computers. In this case, the hardware configuration of each computer may or may not match. The model generation device 1 may be an information processing device designed exclusively for the service provided, a general-purpose server device, a general-purpose PC (Personal Computer), or the like.

<Walking ability determination device>
FIG. 3 schematically illustrates an example of the hardware configuration of the walking ability determination device 2 according to this embodiment. As shown in FIG. 3, in the walking ability determination device 2 according to the present embodiment, the control unit 21, the storage unit 22, the communication interface 23, the external interface 24, the input device 25, the output device 26, and the drive 27 are electrically A connected computer.

The controllers 21 to 27 and the storage medium 92 of the walking ability determination device 2 may be configured similarly to the controllers 11 to 17 and the storage medium 91 of the model generation device 1, respectively. The control unit 21 includes a hardware processor such as a CPU, a RAM, and a ROM, and is configured to execute various types of information processing based on programs and data. The storage unit 22 is composed of, for example, a hard disk drive, a solid state drive, or the like. In this embodiment, the storage unit 22 stores various information such as the walking ability determination program 82 and the learning result data 125 .

The walking ability determination program 82 is a program for causing the walking ability determination device 2 to execute information processing (FIG. 10), which will be described later, regarding determination of the walking ability of the subject. The walking ability determination program 82 includes a series of instructions for the information processing. Details will be described later. At least one of the walking ability determination program 82 and the learning result data 125 may be stored in the storage medium 92 . Also, the walking ability determination device 2 may acquire at least one of the walking ability determination program 82 and the learning result data 125 from the storage medium 92 .

Regarding the specific hardware configuration of the walking ability determination device 2, it is possible to omit, replace, and add components as appropriate according to the embodiment. For example, the controller 21 may include multiple hardware processors. A hardware processor may comprise a microprocessor, FPGA, or the like. The storage unit 22 may be configured by RAM and ROM included in the control unit 21 . At least one of the communication interface 23, the external interface 24, the input device 25, the output device 26, and the drive 27 may be omitted. The walking ability determination device 2 may be composed of a plurality of computers. In this case, the hardware configuration of each computer may or may not match. In addition, the walking ability determination device 2 is an information processing device designed exclusively for the service provided, a general-purpose server device, a general-purpose PC, a mobile terminal (for example, a smartphone), a tablet PC, a microcomputer, etc. good too.

[Software configuration]
<Model generator>
FIG. 4 schematically illustrates an example of the software configuration of the model generating device 1 according to this embodiment. The control unit 11 of the model generation device 1 develops the model generation program 81 stored in the storage unit 12 in RAM. Then, the control unit 11 causes the CPU to execute instructions included in the model generation program 81 developed in the RAM, and controls each component. As a result, as shown in FIG. 4, the model generation device 1 according to this embodiment operates as a computer including a learning data acquisition unit 111, a learning processing unit 112, and a storage processing unit 113 as software modules.

The learning data acquisition unit 111 is configured to acquire a plurality of learning data sets 3 respectively generated by observing walking of a plurality of subjects. Each learning data set 3 is composed of a combination of training data 31 and correct labels 32 . The training data 31 is configured to include at least one of the first to fifth information of the corresponding subject. Correct label 32 is configured to indicate the true value of the corresponding subject's walking ability.

The learning processing unit 112 is configured to perform machine learning of the judgment model 6 using the acquired plurality of learning data sets 3 . In machine learning, for each learning data set 3, the judgment model 6 is trained so that the result of judging the walking ability of each subject by the judgment model 6 based on the training data 31 conforms to the true value indicated by the correct label 32. It is constructed by

The storage processing unit 113 is configured to generate information related to the trained judgment model 6 generated by machine learning as the learning result data 125 and save the generated learning result data 125 in an arbitrary storage area. Learning result data 125 may be configured as appropriate to include information for reproducing trained decision model 6 .

(Example of judgment model)
The judgment model 6 is composed of a machine learning model having one or more calculation parameters used in calculations for deriving inference results. The type of the machine learning model is not particularly limited as long as the process of inferring the walking ability can be executed, and may be appropriately selected according to the embodiment. As an example, the judgment model 6 may be configured by a neural network, a support vector machine, a regression model, a decision tree model, or the like.

Training the decision model 6 consists of using each learning data set 3 and adjusting (optimizing) the values of the calculation parameters so as to derive the true value of the corresponding correct label 32 from the training data 31. be. This machine learning method may be appropriately selected according to the type of machine learning model employed. By way of example, machine learning methods may employ methods such as backpropagation, solving optimization problems, performing regression analysis, random forests, and the like.

In one example of FIG. 4, the judgment model 6 is configured by a neural network. When employing a neural network to construct the decision model 6, the decision model 6 is typically configured to comprise an input layer, one or more hidden layers (hidden layers), and an output layer. Each layer may employ any type of layer, such as, for example, a fully bonded layer. The number of layers included in the judgment model 6, the number of nodes (neurons) in each layer, and the connection relationship of the nodes may be determined as appropriate according to the embodiment. The weight of the connection between each node, the threshold value of each node, and the like are examples of the calculation parameters.

As an example of training processing when a neural network is employed, the learning processing unit 112 inputs the training data 31 of each learning data set 3 to the judgment model 6 and executes computation processing of forward propagation of the judgment model 6 . The learning processing unit 112 acquires an output value corresponding to the result of determining the walking ability of each subject based on the training data 31 of each learning data set 3 as the calculation result of this forward propagation. The learning processing unit 112 calculates the error between the obtained output value and the true value indicated by the corresponding correct label 32, and further calculates the gradient of the calculated error. Subsequently, the learning processing unit 112 calculates the error of the value of the calculation parameter of the judgment model 6 by backpropagating the gradient of the calculated error using the error backpropagation method. Then, the learning processing unit 112 updates the value of the calculation parameter based on the calculated error.

Through this series of updating processes, the learning processing unit 112 adjusts the values of the parameters of the determination model 6 so that the sum of the errors between the determination result (output value) and the true value of each learning data set 3 becomes small. adjust. This parameter value adjustment may be repeated until a predetermined condition is satisfied, such as, for example, performing a specified number of times or the sum of calculated errors being equal to or less than a threshold. Also, for example, machine learning conditions such as a loss function and a learning rate may be appropriately set according to the embodiment. Through this machine learning process, it is possible to generate a trained determination model 6 that has acquired the ability to determine the walking ability of the subject based on at least one of the first to fifth information.

The storage processing unit 113 stores the trained determination model 6 generated by the machine learning as learning result data 125 . The configuration of the learning result data 125 is not particularly limited as long as it can hold information for executing calculations of the trained judgment model 6, and may be determined as appropriate according to the embodiment. As an example, the learning result data 125 may be configured to include information indicating the configuration of the judgment model 6 (for example, the structure of the neural network, etc.) and the values of the calculation parameters adjusted by machine learning. The learning result data 125 may be saved in any storage area. The learning result data 125 may be referred to as needed to set the trained judgment model 6 to a usable state on the computer.

<Walking ability determination device>
FIG. 5 schematically illustrates an example of the software configuration of the walking ability determination device 2 according to this embodiment. The control unit 21 of the walking ability determination device 2 expands the walking ability determination program 82 stored in the storage unit 22 to RAM. Then, the control unit 21 causes the CPU to execute instructions included in the walking ability determination program 82 developed in the RAM, and controls each component. Thereby, as shown in FIG. 5, the walking ability determination device 2 according to the present embodiment operates as a computer having a data acquisition unit 211, a determination unit 212, and an output unit 213 as software modules.

The data acquisition unit 211 is configured to acquire target data 221 generated by observing the walking of the target person. The target data 221 is configured to include at least one of the first to fifth information of the target person. The determination unit 212 has the trained determination model 6 generated by machine learning by holding the learning result data 125 . The determination unit 212 is configured to determine the walking ability of the subject based on the acquired subject data 221 using the trained determination model 6 . The output unit 213 is configured to output the result of determining the walking ability of the subject.

<Others>
Each software module of the model generation device 1 and the walking ability determination device 2 will be described in detail in operation examples described later. In this embodiment, an example in which each software module of the model generation device 1 and the walking ability determination device 2 is realized by a general-purpose CPU is described. However, some or all of the software modules may be implemented by one or more dedicated processors. That is, each module described above may be implemented as a hardware module. Further, regarding the software configurations of the model generation device 1 and the walking ability determination device 2, omission, replacement, and addition of software modules may be performed as appropriate according to the embodiment.

§3 Operation example [Model generation device]
FIG. 6 is a flow chart showing an example of a processing procedure relating to machine learning of the judgment model 6 by the model generation device 1 according to this embodiment. The following processing procedure of the model generation device 1 is an example of the model generation method. However, the following processing procedure of the model generation device 1 is only an example, and each step may be changed as much as possible. Also, in the following processing procedure of the model generation device 1, steps can be omitted, replaced, or added as appropriate according to the embodiment.

(Step S101)
In step S<b>101 , the control unit 11 operates as the learning data acquisition unit 111 and acquires multiple learning data sets 3 .

Each learning data set 3 may be appropriately generated by observing the subject's walking. Any sensor may be used to observe the subject's gait. The sensor may be, for example, an acceleration sensor, motion capture, camera, or the like. By appropriately analyzing the data obtained by the sensor, at least one of the first information to the fifth information can be obtained, thereby generating the training data 31 .

The first information and the second information can be obtained by three-dimensionally analyzing the movement of the center of gravity of a person from data obtained by sensors. In addition, the starting point of the walking motion can be set at an arbitrary timing, and one walking motion can be captured from the start of walking to the timing of returning to the starting point. As an example, assuming that the timing at which one leg touches the ground is the starting point of the walking motion, when one leg touches the ground, the other leg leaves the floor while the body is supported by both legs, and the body is lifted by only one leg. The series of movements from the period of support, the period when the other leg touches the ground, the period when both legs support the body, the period when one leg leaves the floor, the period when the body is supported only by the other leg, and the one leg touches the ground again, It corresponds to one walking motion. The third to fifth information can be obtained by analyzing such walking motion from data obtained by sensors.

FIG. 7 shows the results (measurement data) of the walking of a subject measured by an acceleration sensor in an experimental example to be described later. FIG. 8A shows the trajectory (coronal plane) of the center of gravity of the human body during walking obtained by analyzing the measurement results of FIG. FIG. 8B shows the trajectory (coronal plane) of the center of gravity of the human body during walking obtained by applying the filtering process to the calculation result of FIG. 8A. FIG. 9A shows the trajectory (cross section) of the human center of gravity during walking obtained by analyzing the measurement results of FIG. FIG. 9B shows the trajectory (cross section) of the human center of gravity during walking obtained by applying the filtering process to the calculation result of FIG. 9A.

In the measurement of the walking, the y-axis of the three-dimensional space corresponds to the front-back direction of the subject (the front direction is the walking direction), the x-axis corresponds to the left-right direction, and the z-axis corresponds to the up-down direction. are doing. However, the correspondence relationship in each direction need not be limited to such an example. Correspondence in each direction may be determined as appropriate according to the embodiment.

As an example, by integrating twice the measurement data of the acceleration sensor shown in FIG. 7 and plotting the left and right positions and the upper and lower positions at each time as a two-dimensional vector, the trajectory of the coronal plane shown in FIG. can be obtained. By applying a filtering process (Fourier transform, bandpass filtering for each peak, and then inverse Fourier transform) to this trajectory, the trajectory shown in FIG. 8B is obtained. can be done. The first information (D_xz) can be obtained by calculating the area of this trajectory by integration. On the other hand, by integrating the measurement data shown in FIG. 7 twice and plotting the left and right positions and the front and rear positions at each time as a two-dimensional vector, the trajectory of the cross section shown in FIG. 9A can be obtained. . Similar to the cross-sectional trajectory, filtering can be applied to this trajectory to obtain the trajectory shown in FIG. 9B. The second information (D_xy) can be obtained by calculating the area of this trajectory by integration.

Also, the measurement data of the acceleration sensor is converted as shown in FIG. By measuring the time (t2), the time when each of the left and right feet changes between the free leg and the stance leg can be known, and the time during which each leg is supported during walking can be calculated. Third information (γ_t) can be obtained from the calculated support time of each leg. Also, if the measurement data of the acceleration sensor is converted as shown in FIGS. The trajectory can be reconstructed three-dimensionally. As described above, the boundaries between the free and stance legs of the left and right feet of the trajectory are indicated by the minimum points in the vertical direction on the left and right, and the distance along the three-dimensional trajectory between them is measured by the time of the left leg and the time of the right leg. can calculate the distance traveled while each leg is supported by one leg during walking. Strictly speaking, there is a time for both legs to stand, but this time is assumed to be infinitely short, and the left and right stance and swing legs are exchanged at the minimum point. The fourth information (γ_d) can be obtained from the calculated moving distance for each leg support. It should be noted that any method of expressing the "degree of difference" in the third information and the fourth information may be used as long as the degree of left-right bias can be expressed. As an example, the “degree of difference” between the third information and the fourth information may be represented by a left-right ratio or a left-right difference.

Furthermore, when the measurement data of the acceleration sensor is Fourier transformed, several peaks appear. The lowest frequency peak indicates the base frequency of the person's walking, which is the stride frequency unit/Hz), and its reciprocal is the time for two steps (unit/second). Also, the time for two steps corresponds to the time to go around from a starting point and return to the starting point in FIG. 8B or FIG. 9B. If you halve it, it will be the average time for one step. Peaks other than the base frequency are its harmonics, and the repetition of walking motion can be captured as a whole. By counting the number of times the walking motion is repeated for a predetermined period of time, the fifth information can be obtained with this base frequency. The predetermined time may be determined as appropriate. It should be noted that any expression method may be used for the "degree of repetition of walking motion" in the fifth information as long as it is possible to express whether the number of times of walking is large or small. As an example, the “degree of repetition of walking motion” may be expressed by frequency or period.

The training data 31 is configured to include at least one of the first to fifth information. The training data 31 of each learning data set 3 may be configured to include all of the first information, second information, third information, fourth information and fifth information for each subject. Also, information other than these may be further adopted as explanatory variables for inference. If other information is further adopted as explanatory variables, the training data 31 may be configured to further include the other information.

On the other hand, the correct label 32 can be generated by appropriately evaluating the walking ability of each subject. The evaluation of walking ability of each subject may be performed manually (physician). Alternatively, assessment of the walking ability of at least a portion of the subject may be performed by any mechanical technique. The correct label 32 is configured to indicate the true value of the subject's walking ability. The walking ability evaluation index may be arbitrary. Walking ability may be represented by continuous or discrete values (eg, categories, classes, etc.). As an example, FAC may be adopted as an evaluation index of walking ability. By associating the generated correct label 32 with the corresponding training data 31, each learning data set 3 can be generated. The number of learning data sets 3 generated from one subject may be arbitrary.

Each training data set 3 may be generated automatically by computer action, or may be generated manually, at least partially including operator action. Moreover, the generation of each learning data set 3 may be performed by the model generation device 1 or may be performed by a computer other than the model generation device 1 . When the model generation device 1 generates each learning data set 3, the control unit 11 acquires each learning data set 3 by executing the above generation processing automatically or manually by an operator's operation. On the other hand, when each learning data set 3 is generated by one or more other computers, the control unit 11 acquires each learning data set 3 via, for example, a network, a storage medium 91, an external storage device, or the like. A part of the learning data sets 3 may be generated by the model generation device 1 and the other learning data sets 3 may be generated by one or more other computers.

The number of learning data sets 3 to be acquired is not particularly limited, and may be determined as appropriate according to the embodiment. After acquiring the plurality of learning data sets 3, the control unit 11 proceeds to the next step S102.

(Step S102)
In step S<b>102 , the control unit 11 operates as the learning processing unit 112 and performs machine learning of the judgment model 6 using the acquired plurality of learning data sets 3 . As described above, the control unit 11 uses machine learning to reduce the error between the inference result (the result of determining the walking ability) for the training data 31 of each learning data set 3 and the true value indicated by the correct label 32. The values of the calculation parameters of the judgment model 6 are optimized as follows. As a result of this machine learning, it is possible to generate a trained determination model 6 that has acquired the ability to determine the walking ability of the subject based on at least one of the first to fifth information. When the training data 31 includes all of the first information to the fifth information, generating a trained judgment model 6 that has acquired the ability to judge the walking ability of the subject based on all of the first information to the fifth information. can be done. When the machine learning process is completed, the control unit 11 advances the process to the next step S103.

(Step S103)
In step S<b>103 , the control unit 11 operates as the storage processing unit 113 and generates information on the trained judgment model 6 generated by machine learning as the learning result data 125 . Then, the control unit 11 saves the generated learning result data 125 in an arbitrary storage area.

The storage destination of the learning result data 125 may be, for example, the RAM in the control unit 11, the storage unit 12, an external storage device, a storage medium, or a combination thereof. The storage medium may be, for example, a CD, DVD, or the like, and the control section 11 may store the learning result data 125 in the storage medium via the drive 17 . The external storage device may be, for example, a data server such as NAS (Network Attached Storage). In this case, the control unit 11 may use the communication interface 13 to store the learning result data 125 in the data server via the network. Also, the external storage device may be, for example, an external storage device connected to the model generation device 1 via the external interface 14 .

When the storage of the learning result data 125 is completed, the control unit 11 terminates the processing procedure of the model generation device 1 according to this operation example.

Note that the generated learning result data 125 may be provided to the walking ability determination device 2 at any timing. As an example, the control unit 11 may transfer the learning result data 125 to the walking ability determination device 2 as the process of step S103 or separately from the process of step S103. The walking ability determination device 2 may acquire the learning result data 125 by receiving this transfer. As another example, the walking ability determination device 2 may acquire the learning result data 125 by accessing the model generation device 1 or the data server via the network using the communication interface 23 . As another example, the walking ability determination device 2 may acquire the learning result data 125 via the storage medium 92 . As another example, the learning result data 125 may be incorporated in the walking ability determination device 2 in advance.

Further, the control unit 11 may update or newly generate the trained judgment model 6 (learning result data 125) by periodically or irregularly repeating the processing from step S101 to step S103. During this repetition, at least part of the learning data set 3 used for machine learning may be changed, corrected, added, deleted, etc., as appropriate. Then, the control unit 11 provides the walking ability determination device 2 with the updated or newly generated learning result data 125 in an arbitrary method and timing, thereby updating the learning result data 125 held by the walking ability determination device 2. You may

[Walking ability determination device]
FIG. 10 is a flowchart showing an example of a processing procedure for judging the walking ability of a subject by the walking ability judging device 2 according to this embodiment. The processing procedure of the walking ability determination device 2 described below is an example of the walking ability determination method. However, the processing procedure of the walking ability determination device 2 described below is merely an example, and each step may be changed as much as possible. Further, in the following processing procedures, steps may be omitted, replaced, or added as appropriate according to the embodiment.

(Step S201)
In step S<b>201 , the control unit 21 operates as the data acquisition unit 211 and acquires the target data 221 .

The target data 221 is the same type of data as the training data 31 and may be generated by the same method as the training data 31 . The target data 221 is configured to include at least one of the first to fifth information of the target person. The target data 221 may be configured to include all of the first information, second information, third information, fourth information and fifth information of the target person.

In one example, the control unit 21 directly or indirectly acquires measurement data generated by observing the walking of the subject from a sensor, and analyzes the acquired measurement data to obtain the target data 221. may be obtained. In another example, target data 221 may be generated by another computer. The control unit 21 may acquire the target data 221 via, for example, a network, the storage medium 92, an external storage device, or the like. After acquiring the target data 221, the control unit 21 advances the process to the next step S202.

(Step S202)
In step S<b>202 , the control unit 21 operates as the determination unit 212 and refers to the learning result data 125 to set the trained determination model 6 . Then, the control unit 21 uses the trained determination model 6 to determine the walking ability of the subject based on the acquired target data 221 . That is, the control unit 21 inputs the acquired target data 221 to the trained judgment model 6 and executes the arithmetic processing of the trained judgment model 6 . As a result of executing this arithmetic processing, the control unit 21 acquires from the trained judgment model 6 an output value corresponding to the result of judging the walking ability of the subject. After acquiring the determination result, the control unit 21 advances the process to the next step S203.

(Step S203)
In step S203, the control unit 21 operates as the output unit 213 and outputs the result of determining the walking ability of the subject. The output destination of the determination result and the content of the information to be output may be appropriately determined according to the embodiment. For example, the control unit 21 may directly output the determination result obtained in step S202. Also, the control unit 21 may execute some information processing based on the obtained determination result. Then, the control unit 21 may output the result of executing the information processing as information on the determination result. As an example, the correspondence relationship between the determination result of walking ability and the information to be notified to the subject (for example, the content of rehabilitation for a subject with low walking ability, etc.) may be determined in advance. Based on this correspondence relationship, the control unit 21 may generate notification information for the subject from the determination result obtained in step S202, and output the generated notification information. The output destination may be, for example, the output device 26, an output device of another computer, or the like. The output format may be, for example, screen output, audio output, printing, or the like.

When the output of the information regarding the determination result is completed, the control unit 21 terminates the processing procedure of the walking ability determination device 2 according to this operation example. The control unit 21 may acquire the target data 221 of a plurality of subjects in step S201, and perform the processes of steps S202 and S203 on the target data 221 of each subject. Thereby, the walking ability determination device 2 may determine the walking ability of each of the plurality of subjects. Further, the control unit 21 may repeatedly execute a series of information processing from step S201 to step S203 for any target person. Thereby, the walking ability determination device 2 may continuously monitor the walking ability of the subject.

[feature]
As described above, in the present embodiment, it is possible to generate a trained determination model 6 that has acquired the ability to determine walking ability from at least one of the first information to the fifth information through the processing from step S101 to step S103. can. By using the generated trained determination model 6 in the process of step S202, the walking ability of the subject can be appropriately evaluated by a mechanical method. Further, by adopting all of the first information to the fifth information as explanatory variables of the determination model 6, the walking ability of the subject can be determined with higher accuracy.

§4 Experimental example <First experimental example>
An acceleration sensor (IMU-Z, manufactured by ZMP) was attached to the waist of each of the 27 subjects, and the walking of each subject was observed by the acceleration sensor to obtain measurement data (FIG. 7). Five pieces of information (walking data) from the first information to the fifth information were calculated from the obtained measurement data by the above method. A difference was used to express the third information and the fourth information. A frequency was used to express the fifth information (freq). In addition, 19 medical workers (doctors) evaluated the walking ability of each subject using the FAC index, and the average value of the evaluation results of each medical worker was obtained as the true value of the walking ability of each subject. bottom. Then, a data set was generated by associating the obtained true values with the walking data as correct labels.

Of the 29 generated datasets, 15 datasets were used for machine learning to generate a trained decision model. A layered neural network (number of layers: 3, number of nodes: 5, 3, 1) was adopted for the configuration of the judgment model, and the error backpropagation method was adopted for the machine learning method. The remaining 12 datasets were then used to evaluate the generated trained decision model. Table 1 below shows the evaluation results of the walking ability of each subject.

As shown in Table 1, the average value of the error between the judgment result of the generated trained judgment model and the average value (true value) of the walking ability evaluation result by the medical staff was 0.996. Considering that the number of data sets used for machine learning is 15, it was found that the walking ability of the subject can be determined with reasonable accuracy according to the above five pieces of information.

In addition, when referring to the judgment results of individual subjects, there was a slightly large discrepancy in the judgments of subjects around FACs 3 and 4. It is presumed that this is due to the fact that even experienced doctors are prone to make mistakes in evaluating walking ability per FAC3 and 4 (for example, FAC3 is mistaken for FAC4), and that the true value indicated by the correct label is blurred. rice field. Therefore, from these points, by increasing the number of data sets used for learning and increasing the number of doctors who evaluate the walking ability of each subject when obtaining the true value (correct label) of walking ability, It was speculated that a more accurate trained judgment model could be generated.

<Second Experimental Example>
Next, pairs of subjects were selected at random by random numbers, and one of the five pieces of information was exchanged (eg, the value of the fifth piece of information was exchanged between patient A and patient B). The difference in the output values of the trained judgment model generated in the first experimental example between before and after the exchange was calculated as the value of PI (permutation importance) of each piece of information. Then, the ratio of each PI value to the sum of the absolute values of the calculated PI values of the five information was calculated as the contribution ratio of each of the five information to the determination of the walking ability.

FIG. 11 shows the result of calculating the contribution rate of each information. As shown in FIG. 11, each of the five pieces of information had a certain degree of contribution. From this result, it was found that the walking ability of the subject can be determined without using all five pieces of information (that is, by at least one of the five pieces of information). Also, among the five pieces of information, the first information (the area of the trajectory of the center of gravity of the human body on the coronal plane) and the fourth information (difference in left-right movement distance when supporting one leg) had a relatively high contribution rate. From this result, it was found that the training data 31 and the target data 221 are preferably configured to include at least one of the first information and the fourth information.

1 ... model generation device,
11... control unit, 12... storage unit, 13... communication interface,
14 ... external interface,
15... input device, 16... output device, 17... drive,
81 ... model generation program, 91 ... storage medium,
111... learning data acquisition unit, 112... learning processing unit,
113... Storage processing unit, 125... Learning result data,
2 ... walking ability determination device,
21... control unit, 22... storage unit, 23... communication interface,
24 ... external interface,
25... input device, 26... output device, 27... drive,
82... walking ability determination program, 92... storage medium,
211 ... data acquisition unit, 212 ... determination unit, 213 ... output unit,
221 ... target data,
3 ... learning data set,
31... training data, 32... correct label,
6 ... Judgment model

Claims (8)

A data acquisition unit that acquires target data generated by observing walking of a subject, wherein the target data is a first data indicating an area on a coronal plane of a trajectory of the center of gravity of a person during walking of the subject. information, second information indicating the area of the cross section of the trajectory, third information indicating the degree of difference in the time of supporting with one of the left and right legs during walking, movement while supporting with one of the left and right legs during walking a data acquisition unit configured to include at least one of fourth information indicating the degree of difference in walking distance and fifth information indicating the degree of repetition of the walking motion for a predetermined time;
a determination unit that determines the walking ability of the subject based on the acquired target data using a trained determination model generated by machine learning;
an output unit that outputs a result of determining the walking ability of the subject;
comprising
Walking ability judgment device. the target data is configured to include all of the first information, the second information, the third information, the fourth information and the fifth information;
The walking ability determination device according to claim 1. the computer
a step of acquiring target data generated by observing the walking of a subject, wherein the target data is first information indicating an area on the coronal plane of the trajectory of the center of gravity of the person during walking of the subject; Second information indicating the area of the cross section of the trajectory, Third information indicating the degree of difference in the time of support by one of the left and right legs during walking, Distance moved while supporting by one of the left and right legs during walking a step configured to include at least one of fourth information indicating the degree of difference between and fifth information indicating the degree of repetition of the walking motion for a predetermined time;
determining the walking ability of the subject based on the obtained subject data using a trained determination model generated by machine learning;
a step of outputting a result of determining the walking ability of the subject;
run the
Walking ability judgment method. to the computer,
a step of acquiring target data generated by observing the walking of a subject, wherein the target data is first information indicating an area on the coronal plane of the trajectory of the center of gravity of the person during walking of the subject; Second information indicating the area of the cross section of the trajectory, Third information indicating the degree of difference in the time of support by one of the left and right legs during walking, Distance moved while supporting by one of the left and right legs during walking a step configured to include at least one of fourth information indicating the degree of difference between and fifth information indicating the degree of repetition of the walking motion for a predetermined time;
determining the walking ability of the subject based on the obtained subject data using a trained determination model generated by machine learning;
a step of outputting a result of determining the walking ability of the subject;
to run the
Walking ability judgment program. A data acquisition unit that acquires a plurality of learning data sets respectively generated by observing the gait of each of a plurality of subjects,
Each learning data set is composed of a combination of training data and correct labels,
The training data of each learning data set includes first information indicating an area on the coronal plane of the trajectory of the center of gravity of the human body during walking of each subject, second information indicating an area on the transverse plane of the trajectory, Third information indicating the degree of difference in the time during which the right and left leg is supported during walking, fourth information indicating the degree of difference in the distance traveled while the right and left leg is supported during walking, and a predetermined time period. configured to include at least one of the fifth information indicating the degree of repeating the walking motion between
wherein the correct label for each training data set is configured to indicate the true value of the walking ability of each subject;
a learning data acquisition unit;
A learning processing unit that performs machine learning of a judgment model using a plurality of acquired learning data sets, wherein the machine learning is performed on each of the learning data sets based on the training data for each subject a learning processing unit configured by training the judgment model so that the result of judging the walking ability of the judgment model by the judgment model matches the true value indicated by the correct answer label;
comprising
model generator. the training data of each training data set is configured to include all of the first information, the second information, the third information, the fourth information and the fifth information;
The model generation device according to claim 5. the computer
A step of acquiring a plurality of learning data sets respectively generated by observing the gait of each of a plurality of subjects,
Each learning data set is composed of a combination of training data and correct labels,
The training data of each learning data set includes first information indicating an area on the coronal plane of the trajectory of the center of gravity of the human body during walking of each subject, second information indicating an area on the transverse plane of the trajectory, Third information indicating the degree of difference in the time during which the right and left leg is supported during walking, fourth information indicating the degree of difference in the distance traveled while the right and left leg is supported during walking, and a predetermined time period. configured to include at least one of the fifth information indicating the degree of repeating the walking motion between
wherein the correct label for each training data set is configured to indicate the true value of the walking ability of each subject;
a step;
a step of performing machine learning of a judgment model using the acquired plurality of learning data sets, wherein the machine learning is performed on each of the learning data sets based on the training data to determine the gait of each subject; training the decision model so that a result of determining ability by the decision model matches a true value indicated by the correct answer label;
run the
Model generation method. to the computer,
A step of acquiring a plurality of learning data sets respectively generated by observing the gait of each of a plurality of subjects,
Each learning data set is composed of a combination of training data and correct labels,
The training data of each learning data set includes first information indicating an area on the coronal plane of the trajectory of the center of gravity of the human body during walking of each subject, second information indicating an area on the transverse plane of the trajectory, Third information indicating the degree of difference in the time during which the right and left leg is supported during walking, fourth information indicating the degree of difference in the distance traveled while the right and left leg is supported during walking, and a predetermined time period. configured to include at least one of the fifth information indicating the degree of repeating the walking motion between
wherein the correct label for each training data set is configured to indicate the true value of the walking ability of each subject;
a step;
a step of performing machine learning of a judgment model using the acquired plurality of learning data sets, wherein the machine learning is performed on each of the learning data sets based on the training data to determine the gait of each subject; training the decision model so that a result of determining ability by the decision model matches a true value indicated by the correct answer label;
to run the
Model generation program.

Images