Waist swinging estimation device, estimation system, waist swinging estimation method, and recording medium

1 . A waist swinging estimation device comprising:

at least one memory storing instructions; and

at least one processor connected to the at least one memory and configured to execute the instructions to:

acquire, via a communication interface, feature-amount data generated by a measurement device mounted on footwear of the subject in a position corresponding to an arch of the subject, the feature-amount data being produced from time-series sensor data of spatial accelerations and spatial angular velocities measured by a sensor installed on footwear of left and right feet by (i) detecting heel-contact and toe-off gait events, (ii) first normalizing one gait cycle to 0-100%, (iii) second normalizing a stance/swing ratio to 60/40, and (iv) extracting first feature amounts from a gait-phase cluster selected by correlation analysis;

calculate second feature amounts as bilateral statistics including an average value and an absolute difference of at least the first feature amounts and gait parameters for both feet, and input the second feature amounts together with one or more subject attributes to an estimation model;

estimate the waist swinging of the subject according to an output of the estimation model; and

output information according to waist swinging of the subject including the fluctuation width and feedback information for gait correction;

wherein the estimation model is a trained machine-learning model learned using, for a plurality of subjects, an attribute of the subject, a cluster feature amount generated according to a gait of the subject, and a gait parameter, is trained to correlate foot-movement patterns with waist-movement characteristics, and is configured to output, for one gait cycle, a fluctuation width of waist swinging in at least one of a traveling direction, a left-right direction, and a vertical direction, the fluctuation width being defined as a difference between a maximum and a minimum of perpendicular distances from a regression line of a waist-position time series over one gait cycle detected by heel-contact and toe-off events.

2 . The waist swinging estimation device according to claim 1 , wherein:

the estimation model is configured to output an estimated value regarding the waist swinging according to an input of a gait parameter included in the feature amount data in addition to the second feature amounts, and

the at least one processor is configured to execute the instructions to:

acquire the feature amount data including a gait parameter extracted from a gait waveform of the spatial acceleration and the spatial angular velocity included in the sensor data;

input the gait parameter included in the acquired feature amount data to the estimation model in addition to the second feature amounts; and

estimate waist swinging of the subject according to the estimated value regarding the waist swinging output from the estimation model.

3 . The waist swinging estimation device according to claim 2 , wherein:

the estimation model is configured to output an estimated value regarding the waist swinging according to an input of the first feature amount included in the feature amount data in addition to the second feature amounts, and

the at least one processor is configured to execute the instructions to:

acquire the feature amount data including a first feature amount for each gait phase cluster extracted from a gait waveform of a spatial acceleration and a spatial angular velocity included in the sensor data;

input the first feature amount included in the acquired feature amount data to the estimation model in addition to the second feature amounts; and

estimate waist swinging of the subject according to the estimated value regarding the waist swinging output from the estimation model.

4 . The waist swinging estimation device according to claim 1 , wherein:

the estimation model is configured to output an estimated value regarding the waist swinging according to an input of the second feature amount corresponding to an average value or a difference regarding the first feature amount and the gait parameter for both feet,

the at least one processor is configured to execute the instructions to:

calculate, as the second feature amount, the average value and the difference regarding the first feature amount and the gait parameter to be used for estimation of the waist swinging among the first feature amount and the gait parameter for both feet of the subject;

input the calculated second feature amount to the estimation model; and

estimate waist swinging of the subject according to an estimated value regarding the waist swinging output from the estimation model, and

wherein the average value is (L+R)/2 and the difference is an absolute difference |L−R|, each computed for corresponding gait phases of the left and right feet.

5 . The waist swinging estimation device according to claim 4 , wherein wherein:

the estimation model is configured to output an estimated value regarding the waist swinging according to an input of an attribute of the subject and the second feature amount,

the at least one processor is configured to execute the instructions to:

input an attribute of the subject and the second feature amount to the estimation model; and

estimate waist swinging of the subject according to an estimated value regarding the waist swinging output from the estimation model, and

wherein the attribute of the subject comprises at least one of body weight, height, age, and gender, and is stored in the storage unit.

6 . The waist swinging estimation device according to claim 1 , wherein wherein:

the estimation model is configured to output a fluctuation width of the waist swinging regarding at least one of three directions of a traveling direction, a left-right direction, and a vertical direction in one gait cycle as an estimated value regarding the waist swinging according to an input of the feature amount data,

the at least one processor is configured to execute the instructions to:

input the feature amount included in the acquired feature amount data to the estimation model; and

estimate waist swinging of the subject according to the fluctuation width of the waist swinging regarding at least one of three directions of the traveling direction, the left-right direction, and the vertical direction output from the estimation model, and

wherein the fluctuation width is defined as a difference between a maximum and a minimum of perpendicular distances from a linear-regression line of a waist-position time series over the one gait cycle detected by heel-contact and toe-off events.

7 . An estimation system comprising:

the waist swinging estimation device according to claim 1 ; and

the measurement device installed mounted on the footwear of the subject who is an estimation target of waist swinging that is an index regarding movement of the waist,

wherein the measurement device includes:

the sensor configured to:

measure a spatial acceleration and a spatial angular velocity;

generate the sensor data regarding a movement of a foot by using the measured spatial acceleration and spatial angular velocity;

and

output the generated sensor data;

the at least one memory storing the instructions; and

wherein the at least one processor connected to the at least one memory is configured to execute the instructions to:

acquire time-series data of the sensor data measured by the sensor;

extract gait waveform data for one gait cycle from the time-series data of the sensor data;

normalize the extracted gait waveform data;

extract a feature amount used for estimation of the waist swinging from the normalized gait waveform data from a gait phase cluster configured by at least one temporally continuous gait phase;

generate feature amount data including the extracted feature amounts; and

output the generated feature amount data to the waist swinging estimation device.

8 . The estimation system according to claim 7 , wherein:

the waist swinging estimation device is mounted on a terminal device having a screen visually recognizable by the subject, and

the at least one processor of the waist swinging estimation device is configured to execute the instructions to display information regarding the waist swinging estimated according to the movement of the foot of the subject on a screen of the terminal device.

9 . A waist swinging estimation method causing a computer to execute:

acquiring feature-amount data generated by a measurement device mounted on footwear of the subject in a position corresponding to an arch of the subject, the feature-amount data being produced from time-series sensor data of spatial accelerations and spatial angular velocities measured by a sensor installed on footwear of left and right feet by (i) detecting heel-contact and toe-off gait events, (ii) first normalizing one gait cycle to 0-100%, (iii) second normalizing a stance/swing ratio to 60/40, and (iv) extracting first feature amounts from a gait-phase cluster selected by correlation analysis;

calculating second feature amounts as bilateral statistics including an average value and an absolute difference of at least the first feature amounts and gait parameters for both feet, and input the second feature amounts together with one or more subject attributes to an estimation model;

estimating the waist swinging of the subject according to an output of the estimation model; and

outputting information according to waist swinging of the subject including the fluctuation width and feedback information for gait correction;

wherein the estimation model is a trained machine-learning model learned using, for a plurality of subjects, an attribute of the subject, a cluster feature amount generated according to a gait of the subject, and a gait parameter, is trained to correlate foot-movement patterns with waist-movement characteristics, and is configured to output, for one gait cycle, a fluctuation width of waist swinging in at least one of a traveling direction, a left-right direction, and a vertical direction, the fluctuation width being defined as a difference between a maximum and a minimum of perpendicular distances from a regression line of a waist-position time series over one gait cycle detected by heel-contact and toe-off events.

10 . The waist swinging estimation method according to claim 9 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of a gait parameter included in the feature amount data in addition to the second feature amounts, and

wherein the waist swinging estimation method causing the computer to execute:

acquiring the feature amount data including a gait parameter extracted from a gait waveform of the spatial acceleration and the spatial angular velocity included in the sensor data;

inputting the gait parameter included in the acquired feature amount data to the estimation model in addition to the second feature amounts; and

estimating waist swinging of the subject according to the estimated value regarding the waist swinging output from the estimation model.

11 . The waist swinging estimation method according to claim 10 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of the first feature amount included in the feature amount data in addition to the second feature amounts, and

wherein the waist swinging estimation method causing the computer to execute:

acquiring the feature amount data including a first feature amount for each gait phase cluster extracted from a gait waveform of a spatial acceleration and a spatial angular velocity included in the sensor data;

input the first feature amount included in the acquired feature amount data to the estimation model in addition to the second feature amounts; and

estimate waist swinging of the subject according to the estimated value regarding the waist swinging output from the estimation model.

12 . The waist swinging estimation method according to claim 9 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of the second feature amount corresponding to an average value or a difference regarding the first feature amount and the gait parameter for both feet,

wherein the waist swinging estimation method causing the computer to execute:

calculating, as the second feature amount, the average value and the difference regarding the first feature amount and the gait parameter to be used for estimation of the waist swinging among the first feature amount and the gait parameter for both feet of the subject;

inputting the calculated second feature amount to the estimation model; and

estimating waist swinging of the subject according to an estimated value regarding the waist swinging output from the estimation model, and

wherein the average value is (L+R)/2 and the difference is an absolute difference |L−R|, each computed for corresponding gait phases of the left and right feet.

13 . The waist swinging estimation method according to claim 12 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of an attribute of the subject and the second feature amount,

wherein the waist swinging estimation method causing the computer to execute:

inputting an attribute of the subject and the second feature amount to the estimation model; and

estimating waist swinging of the subject according to an estimated value regarding the waist swinging output from the estimation model, and

wherein the attribute of the subject comprises at least one of body weight, height, age, and gender, and is stored in the storage unit.

14 . The waist swinging estimation method according to claim 9 , wherein the estimation model is configured to output a fluctuation width of the waist swinging regarding at least one of three directions of a traveling direction, a left-right direction, and a vertical direction in one gait cycle as an estimated value regarding the waist swinging according to an input of the feature amount data,

wherein the waist swinging estimation method causing the computer to execute:

inputting the feature amount included in the acquired feature amount data to the estimation model; and

estimating waist swinging of the subject according to the fluctuation width of the waist swinging regarding at least one of three directions of the traveling direction, the left-right direction, and the vertical direction output from the estimation model, and

wherein the fluctuation width is defined as a difference between a maximum and a minimum of perpendicular distances from a linear-regression line of a waist-position time series over the one gait cycle detected by heel-contact and toe-off events.

15 . A non-transitory recording medium with a program recorded therein executed by a computer to execute:

acquiring feature-amount data generated by a measurement device mounted on footwear of the subject in a position corresponding to an arch of the subject, the feature-amount data being produced from time-series sensor data of spatial accelerations and spatial angular velocities measured by a sensor installed on footwear of left and right feet by (i) detecting heel-contact and toe-off gait events, (ii) first normalizing one gait cycle to 0-100%, (iii) second normalizing a stance/swing ratio to 60/40, and (iv) extracting first feature amounts from a gait-phase cluster selected by correlation analysis;

calculating second feature amounts as bilateral statistics including an average value and an absolute difference of at least the first feature amounts and gait parameters for both feet, and input the second feature amounts together with one or more subject attributes to an estimation model;

estimating the waist swinging of the subject according to an output of the estimation model; and

outputting information according to waist swinging of the subject including the fluctuation width and feedback information for gait correction;

wherein the estimation model is a trained machine-learning model learned using, for a plurality of subjects, an attribute of the subject, a cluster feature amount generated according to a gait of the subject, and a gait parameter, is trained to correlate foot-movement patterns with waist-movement characteristics, and is configured to output, for one gait cycle, a fluctuation width of waist swinging in at least one of a traveling direction, a left-right direction, and a vertical direction, the fluctuation width being defined as a difference between a maximum and a minimum of perpendicular distances from a regression line of a waist-position time series over one gait cycle detected by heel-contact and toe-off events.

16 . The non-transitory recording medium according to claim 15 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of a gait parameter included in the feature amount data in addition to the second feature amounts, and

wherein the non-transitory recording medium with the program recorded therein executed by the computer to execute:

acquiring the feature amount data including a gait parameter extracted from a gait waveform of the spatial acceleration and the spatial angular velocity included in the sensor data;

inputting the gait parameter included in the acquired feature amount data to the estimation model in addition to the second feature amounts; and

estimating waist swinging of the subject according to the estimated value regarding the waist swinging output from the estimation model.

17 . The non-transitory recording medium according to claim 16 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of the first feature amount included in the feature amount data in addition to the second feature amounts, and

wherein the non-transitory recording medium with the program recorded therein executed by the computer to execute:

acquiring the feature amount data including a first feature amount for each gait phase cluster extracted from a gait waveform of a spatial acceleration and a spatial angular velocity included in the sensor data;

inputting the first feature amount included in the acquired feature amount data to the estimation model in addition to the second feature amounts; and

estimating waist swinging of the subject according to the estimated value regarding the waist swinging output from the estimation model.

18 . The non-transitory recording medium according to claim 15 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of the second feature amount corresponding to an average value or a difference regarding the first feature amount and the gait parameter for both feet,

wherein the non-transitory recording medium with the program recorded therein executed by the computer to execute:

calculating, as the second feature amount, the average value and the difference regarding the first feature amount and the gait parameter to be used for estimation of the waist swinging among the first feature amount and the gait parameter for both feet of the subject;

inputting the calculated second feature amount to the estimation model; and

estimating waist swinging of the subject according to an estimated value regarding the waist swinging output from the estimation model, and

wherein the average value is (L+R)/2 and the difference is an absolute difference |L−R|, each computed for corresponding gait phases of the left and right feet.

19 . The non-transitory recording medium according to claim 18 , wherein the estimation model is configured to output an estimated value regarding the waist swinging according to an input of an attribute of the subject and the second feature amount,

wherein the non-transitory recording medium with the program recorded therein executed by the computer to execute:

inputting an attribute of the subject and the second feature amount to the estimation model; and

estimating waist swinging of the subject according to an estimated value regarding the waist swinging output from the estimation model, and

wherein the attribute of the subject comprises at least one of body weight, height, age, and gender, and is stored in the storage unit.

20 . The non-transitory recording medium according to claim 15 , wherein the estimation model is configured to output a fluctuation width of the waist swinging regarding at least one of three directions of a traveling direction, a left-right direction, and a vertical direction in one gait cycle as an estimated value regarding the waist swinging according to an input of the feature amount data,

wherein the non-transitory recording medium with the program recorded therein executed by the computer to execute:

inputting the feature amount included in the acquired feature amount data to the estimation model; and

estimating waist swinging of the subject according to the fluctuation width of the waist swinging regarding at least one of three directions of the traveling direction, the left-right direction, and the vertical direction output from the estimation model, and

wherein the fluctuation width is defined as a difference between a maximum and a minimum of perpendicular distances from a linear-regression line of a waist-position time series over the one gait cycle detected by heel-contact and toe-off events.