Estimation device, estimation system, estimation method, and recording medium

1 . An estimation device comprising:

a communication interface;

a memory storing instructions; and

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

receive, via a communication interface, time-series sensor data including three-axis acceleration and angular velocity from a data acquisition device installed in footwear at a back-of-arch position of a foot of a user;

convert the received sensor data from a local coordinate system of the data acquisition device to a world coordinate system;

detect gait events in the converted, received sensor data, the detected gait events including heel-strike events and toe-off events;

generate, for one gait cycle normalized between consecutive ones of the heel-strike events, a gait waveform of a pitch angle representing rotation about a Y-axis in a coronal plane;

extract, from the pitch-angle gait waveform, a feature amount according to an attribute of the user selected from gender and age in a section defined as a predetermined sub-range of the normalized gait cycle identified relative to the detected heel-strike and toe-off events, the feature amount comprising at least one of an arithmetic mean or an integral of the pitch angle;

input the feature amount to an attribute-specific inference model stored in the memory, the model being trained with a data set in which a training feature amount is an explanatory variable and a center of pressure excursion index (CPEI) obtained from a foot pressure distribution measured by a pressure sensor is an objective variable for a plurality of subjects;

execute the attribute-specific inference model to estimate the CPEI for the user based on the inputted feature amount;

classify a degree of pronation/supination of a foot of the user based on the estimated CPEI, the classification being determined using thresholds stored in the memory including a pronation threshold and a supination threshold; and

cause an output unit to present the estimated CPEI and the classification among pronation/normal/supination.

2 . The estimation device according to claim 1 , wherein the processor is configured to execute the instructions to:

input the feature amount extracted from the gait waveform of the user to an inference model that outputs an estimation result regarding a physical condition according to the attribute according to an input of the feature amount extracted according to the attribute; and

estimate the physical condition of the user based on the estimation result output from the inference model.

3 . The estimation device according to claim 1 , wherein the processor is configured to execute the instructions to:

estimate a physical condition of the user by using the inference model learned by machine learning; and

output information to assist the user's decision of contacting a medical institution.

4 . An estimation system comprising:

the estimation device according to claim 1 ; and

a data acquisition device that is configured to be installed at a foot portion of the user, is configured to measure the acceleration and the angular velocity, is configured to generate the time-series sensor data based on the measured acceleration and the measured angular velocity, and is configured to transmit the generated time-series sensor data to the estimation device.

5 . The estimation device according to claim 1 , wherein:

in a case where the attribute is the user being a woman, the predetermined sub-range of the normalized gait cycle is a female feature amount extraction period including a period immediately after heel-strike and a period of single-leg support from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the female feature amount extraction period, and the plurality of subjects is a plurality of female subjects, and

in a case where the attribute is the user being a male, the predetermined sub-range of the normalized gait cycle is a male feature amount extraction period including a period immediately before toe-off and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted from the male feature amount extraction period, and the plurality of subjects is a plurality of male subjects.

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

in a case where the attribute is the user being an elderly person, the predetermined sub-range of the normalized gait cycle is an elderly person feature amount extraction period including a period immediately after heel-strike and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the elderly person feature amount extraction period, and the plurality of subjects is a plurality of elderly subjects, and

in a case where the attribute is the user being a young person, the predetermined sub-range of the normalized gait cycle is a young person feature amount extraction period including a period immediately before toe-off and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the feature amount is extracted in the young person feature amount extraction period, and the plurality of subjects is a plurality of young subjects.

7 . An estimation method executed by a computer, the method comprising:

receiving, via a communication interface, time-series sensor data including three-axis acceleration and angular velocity from a data acquisition device installed in footwear at a back-of-arch position of a foot of a user;

converting the received sensor data from a local coordinate system of the data acquisition device to a world coordinate system;

detecting gait events in the converted, received sensor data, the detected gait events including heel-strike events and toe-off events;

generating, for one gait cycle normalized between consecutive ones of the heel-strike events, a gait waveform of a pitch angle representing rotation about a Y-axis in a coronal plane;

extracting, from the pitch-angle gait waveform, a feature amount according to an attribute of the user selected from gender and age in a section defined as a predetermined sub-range of the normalized gait cycle identified relative to the detected heel-strike and toe-off events, the feature amount comprising at least one of an arithmetic mean or an integral of the pitch angle;

inputting the feature amount to an attribute-specific inference model stored in a memory, the model being trained with a data set in which a training feature amount is an explanatory variable and a center of pressure excursion index (CPEI) obtained from a foot pressure distribution measured by a pressure sensor is an objective variable for a plurality of subjects;

executing the attribute-specific inference model to estimate the CPEI for the user based on the inputted feature amount;

classifying a degree of pronation/supination of a foot of the user based on the estimated CPEI, the classification being determined using thresholds stored in the memory including a pronation threshold and a supination threshold; and

causing an output unit to present the estimated CPEI and the classification among pronation/normal/supination.

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

in a case where the attribute is the user being a woman, the predetermined sub-range of the normalized gait cycle is a female feature amount extraction period including a period immediately after heel-strike and a period of single-leg support from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the female feature amount extraction period, and the plurality of subjects is a plurality of female subjects, and

in a case where the attribute is the user being a male, the predetermined sub-range of the normalized gait cycle is a male feature amount extraction period including a period immediately before toe-off and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted from the male feature amount extraction period, and the plurality of subjects is a plurality of male subjects.

9 . The estimation method according to claim 7 , wherein:

in a case where the attribute is the user being an elderly person, the predetermined sub-range of the normalized gait cycle is an elderly person feature amount extraction period including a period immediately after heel-strike and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the elderly person feature amount extraction period, and the plurality of subjects is a plurality of elderly subjects, and

in a case where the attribute is the user being a young person, the predetermined sub-range of the normalized gait cycle is a young person feature amount extraction period including a period immediately before toe-off and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the young person feature amount extraction period, and the plurality of subjects is a plurality of young subjects.

10 . A non-transitory program recording medium storing a program for causing a computer to execute:

receiving, via a communication interface, time-series sensor data including three-axis acceleration and angular velocity from a data acquisition device installed in footwear at a back-of-arch position of a foot of a user;

converting the received sensor data from a local coordinate system of the data acquisition device to a world coordinate system;

detecting gait events in the converted, received sensor data, the detected gait events including heel-strike events and toe-off events;

generating, for one gait cycle normalized between consecutive ones of the heel-strike events, a gait waveform of a pitch angle representing rotation about a Y-axis in a coronal plane;

extracting, from the pitch-angle gait waveform, a feature amount according to an attribute of the user selected from gender and age in a section defined as a predetermined sub-range of the normalized gait cycle identified relative to the detected heel-strike and toe-off events, the feature amount comprising at least one of an arithmetic mean or an integral of the pitch angle;

inputting the feature amount to an attribute-specific inference model stored in the memory, the model being trained with a data set in which a training feature amount is an explanatory variable and a center of pressure excursion index (CPEI) obtained from a foot pressure distribution measured by a pressure sensor is an objective variable for a plurality of subjects;

executing the attribute-specific inference model to estimate the CPEI for the user based on the inputted feature amount;

classifying a degree of pronation/supination of a foot of the user based on the estimated CPEI, the classification being determined using thresholds stored in the memory including a pronation threshold and a supination threshold; and

causing an output unit to present the estimated CPEI and the classification among pronation/normal/supination.

11 . The non-transitory program recording medium according to claim 10 , wherein:

in a case where the attribute is the user being a woman, the predetermined sub-range of the normalized gait cycle is a female feature amount extraction period including a period immediately after heel-strike and a period of single-leg support from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the female feature amount extraction period, and the plurality of subjects is a plurality of female subjects, and

in a case where the attribute is the user being a male, the predetermined sub-range of the normalized gait cycle is a male feature amount extraction period including a period immediately before toe-off and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted from the male feature amount extraction period, and the plurality of subjects is a plurality of male subjects.

12 . The non-transitory program recording medium according to claim 10 , wherein the program is configured to cause the computer to execute:

in a case where the attribute is the user being an elderly person, the predetermined sub-range of the normalized gait cycle is an elderly person feature amount extraction period including a period immediately after heel-strike and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the elderly person feature amount extraction period, and the plurality of subjects is a plurality of elderly subjects; and

in a case where the attribute is the user being a young person, the predetermined sub-range of the normalized gait cycle is a young person feature amount extraction period including a period immediately before toe-off and a period immediately before heel-strike from the gait waveform regarding an angle in the coronal plane, the training feature amount is extracted in the young person feature amount extraction period, and the plurality of subjects is a plurality of young subjects.