US-12626535-B2 - Posture estimating device, posture estimating system, and posture estimating method

US12626535B2US 12626535 B2US12626535 B2US 12626535B2US-12626535-B2

Abstract

A posture estimating device includes: at least one sensor attached to the person to be measured, the at least one sensor acquiring walking data based on an acceleration of three axes and an angular velocity of the three axes at at least one location of the person to be measured; a processing circuitry that estimates, using an estimation model trained by machine learning, the whole body posture of the person to be measured during the predetermined motion from the walking data acquired by the sensor; and an output device that outputs estimation data. The estimation model is generated in advance by training by machine learning based on the walking data acquired by the sensor during the predetermined motion and based on posture data obtained by measuring the whole body posture of the person to be measured at a timing when the walking data is acquired.

Inventors

Toshiaki Okamoto
Satoru Abe
Shunsuke Yamagata
Masanori Sakaguchi

Assignees

ASICS CORPORATION

Dates

Publication Date: 20260512
Application Date: 20240513
Priority Date: 20230516

Claims (9)

1 . A posture estimating device that estimates a whole body posture of a person to be measured during a predetermined motion, the posture estimating device comprising: a first sensor configured to be disposed in a shoe to acquire measurement data of an acceleration of three axes of a foot of the person to be measured and an angular velocity of three axes of the foot of the person to be measured; a processing circuitry configured to estimate, using an estimation model trained by machine learning on training data of representative acceleration of three axes of feet of a plurality of persons during the predetermined motion, representative angular velocity of three axes of the feet of the plurality of persons during the predetermined motion, and whole body postures of the plurality of persons during the predetermined motion, the whole body posture of the person to be measured during the predetermined motion from the measurement data acquired by a second sensor configured to measure the whole body posture of the person; and an output configured to output estimation data of the whole body posture of the person to be measured that has been estimated by the processing circuitry.
2 . The posture estimating device according to claim 1 , wherein the estimation model is retrained by machine learning for each of persons to be measured who uses the posture estimating device.
3 . The posture estimating device according to claim 1 , wherein the predetermined motion is a motion in which a whole body of the person to be measured moves in conjunction with a motion of the foot of the person to be measured to which the first sensor is attached.
4 . The posture estimating device according to claim 1 , wherein the first sensor is configured to acquire a plurality of parameters based on one of walking or running of the person to be measured based on the measurement data.
5 . The posture estimating device according to claim 4 , wherein the plurality of parameters includes at least one parameter among a stride, a pitch, a walking speed or a running speed, a time taken for one step, a stance period, a swing period, a toe angle at a time of grounding, a pronation, a maximum foot height, and a landing impact.
6 . The posture estimating device according to claim 1 , further comprising: the second sensor configured to measure posture data of the whole body posture of the person to be measured at a timing when the measurement data is acquired, wherein the posture data measured by the second sensor is used when the estimation model is trained by machine learning.
7 . The posture estimating device according to claim 6 , wherein the second sensor is a motion capture camera, and the posture data is skeletal coordinate data acquired using the motion capture camera.
8 . The posture estimating device according to claim 7 , wherein the estimation model is trained by machine learning using the skeletal coordinate data dimensionally compressed by main component analysis.
9 . A posture estimating method for estimating a whole body posture of a person to be measured during a predetermined motion, the posture estimating method comprising: acquiring, from a first sensor configured to be disposed in a shoe, measurement data of an acceleration of three axes of a foot of the person to be measured and an angular velocity of three axes of the foot of the person to be measured; estimating, using an estimation model trained by machine learning on training data of representative acceleration of three axes of feet of a plurality of persons during the predetermined motion, representative angular velocity of three axes of the feet of the plurality of persons during the predetermined motion, and whole body postures of the plurality of persons during the predetermined motion, the whole body posture of the person to be measured during the predetermined motion from the measurement data acquired a second sensor configured to measure the whole body posture of the person; and outputting estimation data of the whole body posture of the person to be measured that has been estimated.

Description

REFERENCE TO RELATED APPLICATIONS This nonprovisional application is based on Japanese Patent Application No. 2023-080778 filed on May 16, 2023 with the Japan Patent Office, the entire contents of which are hereby incorporated by reference. BACKGROUND Technical Field The present disclosure relates to a posture estimating device, a posture estimating system, and a posture estimating method. Background Information In order to improve the form of walking or running, it is necessary to continuously measure a whole body posture during walking or running. In addition, studies for detecting a mental disease from a posture during walking and estimating a risk of fall have also been conducted, and it is necessary to measure a posture during walking on a daily basis in these studies. Known examples of a method of measuring the whole body posture include a method using motion capture, a method using a moving image, and a method using a wearable device. For example, Japanese Patent Laying-Open No. 2023-16825 discloses a system for measuring a whole body posture using a wearable device. According to this system, six sensor devices are attached respectively to six sites of a body of a person to be measured, specifically, joint sites serving as a reference of the body (for example, the waist or the head), or the vicinities of ends of the body (wrist, ankle, head, etc.), and the whole body posture is measured on the basis of measurement values from the sensor devices. SUMMARY However, in the method using motion capture and the method using a moving image, it is necessary to continuously capture the whole body with the whole body always included in the imaging range. It is difficult to continuously capture the whole body while the whole body is always included in the imaging range in order to continuously grasp the whole body posture on a daily basis. On the other hand, in the method using a wearable device, since the posture can be measured by attaching the wearable device to a person to be measured, it is suitable to continuously grasp the posture on a daily basis. However, it is only possible to grasp a partial posture including a part to which the wearable device is attached. Therefore, in order to grasp the whole body posture using the wearable device, it is necessary to wear many sensor devices at joint sites and the vicinities of ends of the body as disclosed in Japanese Patent Laying-Open No. 2023-16825, and thus it is difficult to continuously wear many sensors to grasp the whole body posture on a daily basis. The present disclosure has been made to solve such a problem, and an object thereof is to provide a posture estimating device, a posture estimating system, and a posture estimating method capable of continuously grasping a whole body posture on a daily basis. A posture estimating device according to an aspect of the present disclosure is a posture estimating device that estimates a whole body posture of a person to be measured during a predetermined motion. The posture estimating device includes: at least one sensor attached to the person to be measured, the at least one sensor acquiring measurement data based on an acceleration of three axes and an angular velocity of the three axes at at least one location of the person to be measured; a processing circuitry that estimates, using an estimation model trained by machine learning, the whole body posture of the person to be measured during the predetermined motion from the measurement data acquired by the sensor; and an output device that outputs estimation data of the whole body posture of the person to be measured that has been estimated by the processing circuitry. The estimation model is generated in advance by training by machine learning based on the measurement data acquired by the sensor during the predetermined motion and based on posture data obtained by measuring the whole body posture of the person to be measured at a timing when the measurement data is acquired. A posture estimating system according to an aspect of the present disclosure includes: the measuring device that measures posture data; and the posture estimating device described above, in which the posture data measured by the measuring device is used when the estimation model is trained by machine learning. A posture estimating method according to an aspect of the present disclosure is a posture estimating method for estimating a whole body posture of a person to be measured during a predetermined motion. A posture estimating method includes: acquiring, from at least one sensor attached to the person to be measured, measurement data based on an acceleration of three axes and an angular velocity of the three axes at at least one location of the person to be measured; estimating, using an estimation model trained by machine learning, the whole body posture of the person to be measured during the predetermined motion from the measurement data acquired by the sensor; and outputting estimatio