Analysis device, analysis method, analysis program, and generation device

1 . An analysis device that analyzes a motion of a target with constitution bodies connected to each other, the analysis device comprising:

angular momentum information acquisition processor circuitry configured to acquire angular momentum information representing a time series of angular momentum for each of the constitution bodies; and

characteristic information acquisition processor circuitry configured to acquire characteristic information, the characteristic information being based on a first singular vector and representing a characteristic of the motion, the first singular vector corresponding to a first singular value and having elements respectively corresponding to the constitution bodies, the first singular value being the largest among singular values of a first matrix whose elements are the angular momentum information for the constitution bodies, respectively,

output processor circuitry configured to output information representing a degree of normality of the motion based upon the characteristic information, wherein

the characteristic information is based on an element of a second singular vector, the element corresponding to a target of interest among targets, the second singular vector corresponding to a second singular value and having elements respectively corresponding to the targets, the second singular value being the second largest among singular values of a second matrix whose elements are the first singular vectors for the targets, respectively.

2 . The analysis device according to claim 1 , wherein

the characteristic information acquisition processor circuitry configured to acquire the characteristic information based on a trained model and the angular momentum information for the target of interest, and

the trained model is generated by:

acquiring a second singular vector, the second singular vector corresponding to a second singular value and having elements respectively corresponding to learning targets, the second singular value being the second largest among singular values of a second matrix whose elements are the first singular vectors for the respective-learning targets, respectively; and

learning teaching data for each of the learning targets, the teaching data including the angular momentum information of the learning target and information based on an element of the second singular vector, the element corresponding to the learning target.

3 . The analysis device according to claim 1 , wherein

the target is a human with paralysis in a half of the body on left side or right side, and

the constitution bodies include:

a pelvic region;

a forearm, an upper arm, a thigh, a lower leg, and a foot, each included in the half of the body with paralysis; and

a forearm, an upper arm, a thigh, a lower leg, and a foot, each included in another half of the body without paralysis.

4 . The analysis device according to claim 1 , wherein

the motion is a gait, and

the angular momentum information corresponds to a period with two steps composed of one step on each side.

5 . The analysis device according to claim 1 , wherein

the characteristic information is acquired to reflect a characteristic which is unique to the motion of the target of interest in the targets.

6 . An analysis method that analyzes a motion of a target with constitution bodies connected to each other, the analysis method including:

acquiring angular momentum information representing a time series of angular momentum for each of the constitution bodies; and

acquiring characteristic information, the characteristic information being based on a first singular vector and representing a characteristic of the motion, the first singular vector corresponding to a first singular value and having elements respectively corresponding to the constitution bodies, the first singular value being the largest among singular values of a first matrix whose elements are the angular momentum information for the constitution bodies, respectively,

outputting information representing a degree of normality of the motion based upon the characteristic information, wherein

the characteristic information is based on an element of a second singular vector, the element corresponding to a target of interest among targets, the second singular vector corresponding to a second singular value and having elements respectively corresponding to the targets, the second singular value being the second largest among singular values of a second matrix whose elements are the first singular vectors for the targets, respectively.

7 . The analysis method according to claim 6 , wherein

the characteristic information is acquired to reflect a characteristic which is unique to the motion of the target of interest in the targets.

8 . The analysis method according to claim 6 , wherein

the acquiring of the characteristic information is based on a trained model and the angular momentum information for the target of interest, and

the trained model is generated by:

acquiring a second singular vector, the second singular vector corresponding to a second singular value and having elements respectively corresponding to learning targets, the second singular value being the second largest among singular values of a second matrix whose elements are the first singular vectors for the learning targets, respectively; and

learning teaching data for each of the learning targets, the teaching data including the angular momentum information of the learning target and information based on an element of the second singular vector, the element corresponding to the learning target.

9 . The analysis method according to claim 6 , wherein

the target is a human with paralysis in a half of the body on left side or right side, and

the constitution bodies include:

a pelvic region;

a forearm, an upper arm, a thigh, a lower leg, and a foot, each included in the half of the body with paralysis; and

a forearm, an upper arm, a thigh, a lower leg, and a foot, each included in another half of the body without paralysis.

10 . The analysis method according to claim 6 , wherein

the motion is a gait, and

the angular momentum information corresponds to a period with two steps composed of one step on each side.

11 . A generation device that generates a trained model used to analyze a motion of a target with constitution bodies connected to each other, the generation device comprising:

angular momentum information acquisition processor circuitry configured to acquire angular momentum information for each of learning targets, the angular momentum information representing a time series of angular momentum for each of the constitution bodies;

first singular vector acquisition processor circuitry configured to acquire a first singular vector for each of the learning targets, the first singular vector corresponding to a first singular value and having elements respectively corresponding to the constitution bodies, the first singular value being the largest among singular values of a first matrix whose elements are the angular momentum information for the constitution bodies, respectively;

second singular vector acquisition processor circuitry configured to acquire a second singular vector, the second singular vector corresponding to a second singular value and having elements respectively corresponding to the learning targets, the second singular value being the second largest among singular values of a second matrix whose elements are the first singular vectors for the learning targets, respectively;

a model generator configured to generate the trained model by learning teaching data for each of the learning targets, the teaching data including the angular momentum information of the learning target and information based on an element of the second singular vector, the element corresponding to the learning target, wherein the trained model is used to acquire characteristic information representing a characteristic of a motion of a target of interest; and

output processor circuitry configured to output information representing a degree of normality of the motion based upon the characteristic information.

12 . The generation device according to claim 11 , wherein

the target is a human with paralysis in a half of the body on left side or right side, and

the constitution bodies include:

a pelvic region;

a forearm, an upper arm, a thigh, a lower leg, and a foot, each included in the half of the body with paralysis; and

a forearm, an upper arm, a thigh, a lower leg, and a foot, each included in another half of the body without paralysis.

13 . The generation device according to claim 11 , wherein

the motion is a gait, and

the angular momentum information corresponds to a period with two steps composed of one step on each side.

14 . The generation device according to claim 11 , wherein

the characteristic information is acquired to reflect a characteristic which is unique to the motion of the target of interest in the learning targets.