Walking posture analysis device, method and program

Abstract

To analyze a posture of a subject during walking easily and inexpensively.SOLUTION: A walking posture analysis device 100 includes: an acceleration data acquisition unit 110; a walking cycle determination unit 130 which divides the acquired acceleration data into a first walking cycle associated with one foot of the subject and a second walking cycle associated with the other foot, and determines whether the first walking cycle and the second walking cycle are the walking cycle associated with the right foot; and a walking posture analysis unit 140 which uses a previously calculated multiple regression model to calculate right-side joint angle data being the angle data of the joint of the right half of the body in the walking cycle from the acceleration data in the walking cycle associated with the right foot, and calculate left-side joint angle data being the angle data of the joint of the left half of the body in the walking cycle from the acceleration data in the walking cycle associated with the left foot.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a technique for analyzing the posture of a pedestrian while walking, and more specifically, to a technique for analyzing joint angles of a pedestrian while walking. [Background technology]

[0002] Conventionally, in order to visualize and analyze posture during walking, more specifically, joint angles during walking, walking posture has been recorded and analyzed using a human skeletal model by motion capture. In motion capture technology, markers are attached to the joints and other parts of the subject, these markers are photographed by multiple cameras, and the walking posture of the subject is detected from the movement trajectory of these markers. However, this motion capture technology has the problem that it is inconvenient because it requires attaching markers to the subject, which takes time and effort to measure.

[0003] To address this issue, markerless motion capture technology has emerged (see Patent Document 1). This motion capture technology uses a 3D camera capable of capturing images including depth information, and estimates skeletal information while walking based on the images captured by the 3D camera. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2022-61910 [Patent Document 2] Patent No. 6111837 Summary of the Invention [Problem to be solved by the invention]

[0005] However, the method described in Patent Document 1 has a problem in that it requires an expensive device called a 3D camera.In addition, the method described in Patent Document 1 also has a problem in that it requires a photographer in addition to the subject, or a tripod or the like when a photographer cannot be prepared, which requires a large space and effort for measurement.

[0006] On the other hand, Patent Document 2 describes a technology for detecting walking posture based on detection data from an acceleration sensor. Acceleration sensors are inexpensive and are often built into mobile devices such as smartphones, which have become widespread in recent years. In addition, since the acceleration sensor is attached to the subject and measurements are taken, no space is required for photography. This solves the problems in Patent Document 1 described above. However, the technology described in Patent Document 2 calculates the "deviation of the center of gravity position" as the walking posture to be detected, and cannot detect the joint angles during walking.

[0007] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide an apparatus, method, and program capable of analyzing a subject's posture while walking simply and inexpensively. [Means for solving the problem]

[0008] In order to achieve the above-mentioned object, the walking posture analysis device of the present invention is characterized by comprising an acceleration data acquisition unit that acquires acceleration data of a subject walking measured by an acceleration sensor; a walking cycle determination unit that divides the acquired acceleration data into a first walking cycle, which is a walking cycle from when one foot of the subject touches the ground to when that foot touches the ground next, and a walking cycle determination unit that is a walking cycle from when the other foot of the subject touches the ground to when that other foot touches the ground next, and determines whether the first walking cycle and the second walking cycle are walking cycles related to the right foot or the left foot, respectively; and a walking posture analysis unit that calculates right side joint angle data, which is angle data of the joints of the right half of the body during a walking cycle, from the acceleration data during the walking cycle related to the right foot, and calculates left side joint angle data, which is angle data of the joints of the left half of the body during that walking cycle, from the acceleration data during the walking cycle related to the left foot, using a multiple regression model calculated in advance. Effect of the Invention

[0009] According to the present invention, the posture of the subject while walking is analyzed based on acceleration data while the subject is walking. This acceleration data can be easily and inexpensively obtained by an acceleration sensor. Therefore, the posture of the subject while walking can be analyzed simply and inexpensively. [Brief description of the drawings]

[0010] [Figure 1] Functional block diagram of walking posture analysis device [Diagram 2] Flowchart explaining the operation of the walking posture analysis device [Diagram 3] An example of measured acceleration data [Figure 4] An example of total acceleration data [Diagram 5] Diagram explaining heel contact timing [Figure 6] A diagram explaining the division of the first walking cycle into odd-numbered phases and the second walking cycle into even-numbered phases. [Figure 7] FIG. 1 is a diagram for explaining standardization and averaging of acceleration data; [Figure 8] A diagram explaining left / right determination during walking cycle [Figure 9] An example of averaged acceleration data [Figure 10] An example of averaged joint angle data [Figure 11] A diagram explaining the processing of the walking posture analysis unit [Figure 12] An example of a visualized image of joint angle data DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] A walking posture analysis device according to an embodiment of the present invention will be described with reference to the drawings. Fig. 1 is a functional block diagram of the walking posture analysis device, and Fig. 2 is a flow chart for explaining the operation of the walking posture analysis device.

[0012] A walking posture analysis device according to this embodiment (hereinafter referred to as the "walking posture analysis device") is carried by the subject and acquires acceleration data of the subject while walking. The walking posture analysis device then analyzes the walking posture of the subject based on the acquired acceleration data. The walking posture analysis device also displays the analyzed walking posture as a video. Details of the walking posture analysis device are described below.

[0013] 1, the walking posture analysis device 100 according to this embodiment includes an acceleration data acquisition unit 110, a data range setting unit 120, a walking cycle determination unit 130, a walking posture analysis unit 140, a display control unit 150, and a storage unit 160. In addition, an acceleration sensor 200 and a display device 300 are connected to the walking posture analysis device 100.

[0014] Each unit of the walking posture analysis device 100 can be configured by a conventionally known computer equipped with a main processing unit, a main memory unit, an auxiliary memory unit, an input device, etc. The walking posture analysis device 100 can be implemented by installing a program that causes each unit described above to function in a computer. The walking posture analysis device 100 can be implemented as dedicated hardware. The walking posture analysis device 100 can be distributed and implemented in multiple devices. In this embodiment, the walking posture analysis device 100 is implemented by installing a program in a terminal that is a high-function mobile communication terminal called a "smartphone" and has an acceleration sensor 200 and a display device 300 built in.

[0015] The acceleration sensor 200 is made up of a well-known three-axis inertial sensor, and detects the acceleration of a pedestrian while walking. The accelerations detected here are the acceleration in the vertical direction (Z-axis direction), the acceleration in the walking direction (Y-axis direction), and the acceleration in the left and right horizontal directions (X-axis direction) relative to the walking direction. In this embodiment, the right direction of the X-axis is positive and the left direction is negative relative to the walking direction.

[0016] The acceleration data acquisition unit 110 acquires acceleration data of the subject walking measured by the acceleration sensor 200 at a predetermined measurement rate (for example, 100 Hz) (step S1 in FIG. 2). Here, during acceleration measurement, the relative position of the acceleration sensor 200 with respect to the subject, i.e., the relative position of the walking posture analysis device 100, is fixed so that each axial direction of the acceleration sensor 200 is constant. The fixed position of the walking posture analysis device 100 is preferably closer to the subject's median line in the left-right direction. Moreover, the fixed position of the walking posture analysis device 100 is preferably higher than above the crotch in the vertical direction, and is preferably lower than the neck. Therefore, during acceleration measurement, it is preferable to fix the walking posture analysis device 100 around the navel in the center of the abdomen of the subject or around the center of the waist on the back side in the left-right direction. The acceleration data acquisition unit 110 stores the acquired acceleration data 161 in the storage unit 160. An example of the acquired acceleration data 161 is shown in FIG. 3.

[0017] The data range setting unit 120 sets a time range to be analyzed from the acceleration data 161. This is because the acceleration data acquisition unit 110 sets the time range so that only the acceleration data related to stable walking is analyzed, since the period from the start of measurement to the end of measurement includes a period during which the user is not walking, and walking is not stable immediately after starting walking or immediately before stopping walking.

[0018] The data range setting unit 120 adds the accelerations of the X-axis, Y-axis, and Z-axis of the acceleration data 161, and applies a low-pass filter to the added acceleration data to remove noise (sine wave processing) to generate combined acceleration data (step S2 in FIG. 2). The data range setting unit 120 sets the time range to be analyzed based on this combined acceleration data (step S3 in FIG. 2). FIG. 4 shows an example of combined acceleration data. In FIG. 4, the time when the combined acceleration reaches a peak value (maximum value) equal to or greater than a predetermined value is the time of heel contact, and the period from when the first peak value is detected to when the last peak value is detected is considered to be the walking time range.

[0019] The data range setting unit 120 may set the time range automatically using a predetermined algorithm, or may set the time range based on an input instruction from the user. In the former case, for example, as shown in Fig. 4, the data range setting unit 120 detects the time when the resultant acceleration reaches a peak value (maximum value) equal to or greater than a predetermined value as the time of heel strike, and sets the time range to be analyzed from the second heel strike to the penultimate heel strike. In the latter case, the data range setting unit 120 displays the resultant acceleration data on the display device 300, accepts input of the time range to be analyzed from the user, and sets this as the time range to be analyzed.

[0020] The walking cycle determination unit 130 divides the acceleration data 161 into a first walking cycle, which is a walking cycle from when one foot of the subject touches the ground to when the other foot touches the ground next, and a second walking cycle, which is a walking cycle from when the other foot of the subject touches the ground to when the other foot touches the ground next. Furthermore, the walking cycle determination unit 130 determines whether the first walking cycle and the second walking cycle are walking cycles related to the right foot or the left foot, respectively. The walking cycle determination unit 130 will be described in detail below.

[0021] The walking cycle determination unit 130 refers to the above-mentioned total acceleration data within the time range set by the data range setting unit 120, detects and counts the time when the resultant acceleration reaches a peak value (maximum value) as the time of heel strike, and stores it as heel strike timing information. Figure 5 shows the relationship between the resultant acceleration and the heel strike timing.

[0022] Next, the walking cycle determination unit 130 refers to the heel strike timing information and divides the acceleration data 161 into a plurality of sections between adjacent odd count values ​​(called "odd phases") and a plurality of sections between adjacent even count values ​​(called "even phases") (step S4 in FIG. 2). Each odd phase is a first walking cycle, which is a walking cycle from when one heel of the subject touches the ground to when the other heel touches the ground next. Specifically, the odd phases are from the first heel strike to the third heel strike, from the third heel strike to the fifth heel strike, ... from the 2n+1 heel strike to the 2n+3 heel strike. Similarly, each even phase is a second walking cycle, which is a walking cycle from when the other heel of the subject touches the ground to when the other heel touches the ground next. Specifically, the even phases are from the second heel strike to the fourth heel strike, from the fourth heel strike to the sixth heel strike, ... from the 2n heel strike to the 2n+2 heel strike. Figure 6 shows the first gait cycle with the divided odd phases and the second gait cycle with the divided even phases.

[0023] Next, the walking cycle determination unit 130 normalizes the unit of time axis direction for each walking cycle of the divided odd phase and even phase from real time to 0 to 1 (0 to 100%) (step S5 in FIG. 2). In the present invention, the walking posture is analyzed based on the waveform characteristics of the acceleration in each walking cycle, but since the walking of the subject is not stable, the periodic time of each walking cycle varies. For this reason, in the present invention, the normalization process in the time axis direction is performed as described above.

[0024] Next, the walking cycle determination unit 130 refers to the acceleration data 161 as shown in Fig. 7, calculates the average of the accelerations of the X-axis, Y-axis, and Z-axis in each first walking cycle for the odd-numbered phase, calculates the average acceleration data 162 of each of the X-axis, Y-axis, and Z-axis in one walking cycle, and stores this in the storage unit 160 (step S6 in Fig. 2). The walking cycle determination unit 130 similarly calculates the average acceleration data 162 in the even-numbered phase, and stores this in the storage unit 160. As a result of the left / right determination of each phase by the process described later, the two average acceleration data 162 stored in the storage unit 160 are treated as the average acceleration data 162a relating to the right foot and the average acceleration data 162b relating to the left foot, as shown in Fig. 1.

[0025] Next, the walking cycle determination unit 130 determines whether the first walking cycle in the odd phase is a walking cycle related to the right leg or the left leg based on the waveform characteristics of the averaged acceleration on the X-axis of either the odd phase or the even phase (here, the odd phase) (step S7 in FIG. 2).

[0026] Specifically, as shown in FIG. 8, the walking cycle determination unit 130 performs left / right determination based on the value of the average acceleration of the X-axis in the first half of the time axis direction of the walking cycle, preferably in a time range of 20 to 50%. More specifically, in the time range, the number of positive average accelerations of the X-axis and the number of negative average accelerations of the X-axis are counted and compared. Then, when the number of positive average accelerations of the X-axis is greater, it is determined that the first walking cycle of the odd phase is a walking cycle related to the left foot. On the other hand, when the number of negative average accelerations of the X-axis is greater, it is determined that the first walking cycle of the odd phase is a walking cycle related to the right foot. Hereinafter, a walking cycle determined by the walking cycle determination unit 130 to be a walking cycle related to the right foot is referred to as a right walking cycle. Similarly, a walking cycle determined by the walking cycle determination unit 130 to be a walking cycle related to the left foot is referred to as a left walking cycle.

[0027] In another embodiment, instead of the positive and negative number of the averaged acceleration on the X-axis, other statistical values ​​such as the integral value of the averaged acceleration on the X-axis can be used as a criterion for left / right determination. In another embodiment, the range in the time axis direction to be determined can be the latter half of the range in the time axis direction instead of the first half of the range. In this case, it is determined that the positive and negative of the averaged acceleration on the X-axis is reversed at the time of determination. In another embodiment, it is possible to perform left / right determination based on a plurality of second walking cycles in the even phase. In this case, it is determined that the positive and negative of the averaged acceleration on the X-axis is reversed at the time of determination.

[0028] The walking posture analysis unit 140 calculates right side joint angle data 163a, which is angle data of the joints of the right half of the body in a right walking cycle, from right averaged acceleration data 162a in the right walking cycle, and calculates left side joint angle data 163b, which is angle data of the joints of the left half of the body in the left walking cycle, from left averaged acceleration data 162b in the left walking cycle, using a multiple regression model calculated in advance. First, the multiple regression model used in this calculation process will be described in detail.

[0029] This multiple regression model is calculated in advance using another computer prior to the analysis process by the walking posture analysis device 100. The parameters of the calculated multiple regression model are stored in advance in the storage unit 160 as analysis parameters 169.

[0030] This multiple regression model is calculated by performing principal component analysis and multiple regression analysis using samples of acceleration data (acceleration data for model generation) during walking of multiple subjects measured in advance and joint angle data obtained by imaging and image analysis at the time of measurement corresponding to the acceleration data for model generation. Here, the acceleration data for model generation and joint angle data used in the calculation of the multiple regression model include both data related to right and left walking cycles. In other words, the multiple regression model is not differentiated for right and left feet, but is a single model common to both right and left feet.

[0031] The joint angle data indicates the walking posture of the subject, and represents the angles between the bones that constitute the joints of the human body. The joint angle data includes joint angle data related to the right half of the body (right joint angle data) and joint angle data related to the left half of the body (left joint angle data). The right joint angle data corresponds to acceleration data in a right walking cycle. The left joint angle data corresponds to acceleration data in a left walking cycle. In this embodiment, the subject's lower body, i.e., joints below the waist, are the subject's target. Specifically, the left and right pelvis, hip joints, knee joints, and ankle joints are the subject's target. Each joint angle data includes an angle in the sagittal plane (angle seen from the side), an angle in the frontal plane (angle seen from the front), and an angle in the horizontal plane (angle seen from above). Therefore, the joint angle data includes 4 locations x 3 directions = 12 measurement values ​​per measurement timing.

[0032] The time range to be analyzed for the model generation acceleration data and the corresponding joint angle data used to create the multiple regression model is set in advance by the same process as that performed by the data range setting unit 120 described above.

[0033] Furthermore, by the same process as that performed by the walking cycle determination unit 130, the acceleration data for generating the model and the corresponding joint angle data are divided into a plurality of right walking cycles and left walking cycles, and normalized in the time axis direction.

[0034] Furthermore, the acceleration data for model generation and the corresponding joint angle data are averaged by the same process as that of the walking cycle determination unit 130. Specifically, the arithmetic average of the accelerations of the X-axis, Y-axis, and Z-axis in each of a plurality of right walking cycles is calculated to calculate the averaged acceleration data of each of the X-axis, Y-axis, and Z-axis in one right walking cycle. The same is true for the left walking cycle. The arithmetic average of the right joint angle data in each direction of the pelvis (sagittal plane, frontal plane, horizontal plane) in each of a plurality of right walking cycles is also calculated to calculate the averaged acceleration data of each direction in one right walking cycle. The same is true for the right joint angle data of the hip joint, knee joint, and ankle joint. Furthermore, the same process is also performed on the left joint angle data.

[0035] The explanatory variable (independent variable) A of the multiple regression model is the principal component [A1, A2, …, A N ] T Here, the number of samples of the observation variables subject to the principal component analysis is two per one walk (a series of movements from the start of walking to the end of walking), the average acceleration data pertaining to the right walking cycle and the average acceleration data pertaining to the left walking cycle. Therefore, the number of samples of the observation variables subject to the principal component analysis is the number of walks pertaining to the measurements for generating the model x 2.

[0036] The observation variables to be subjected to the principal component analysis are a data group consisting of averaged acceleration data of multiple axes at each measurement time within one walking cycle. FIG. 9 shows an example of the averaged acceleration data. As shown in FIG. 9, one piece of averaged acceleration data consists of averaged acceleration data of three axes (X-axis, Y-axis, and Z-axis). That is, the entire graph in FIG. 9 is one observation variable. For example, when using the measured values ​​of averaged acceleration data at 1% intervals in the time axis direction within a walking cycle, the number of components of the observation variables is 101 (0% to 100%) x 3 (X-axis, Y-axis, and Z-axis). In the present invention, the method of principal component analysis is used to convert these many components into N principal components.

[0037] Specifically, first, (1) calculate the non-standardized principal component scores of the averaged acceleration data and the eigenvectors of each principal component axis (hereinafter referred to as "acceleration eigenvectors"). Here, for the non-standardized principal component scores to be calculated, N eigenvalues ​​of each principal component axis obtained by the principal component analysis, each of which has an eigenvalue of 1 or more, are selected. (2) Next, standardization processing (average 0, variance 1) is performed on the non-standardized principal component scores of the averaged acceleration data calculated in (1) above to obtain standardized principal component scores A1, A2, ..., A1 of the averaged acceleration data. N The acceleration eigenvectors of the principal component axes calculated in (1) above are stored in the storage unit 160 as one of the analysis parameters 169.

[0038] The dependent variable B of the multiple regression model is the M-dimensional (M: natural number) principal components [B1, B2, …, B M ] T Here, since the averaged joint angle data corresponds to the averaged acceleration data, the number of samples of the observation variables that are the subject of the principal component analysis is the same as the number of samples in the principal component analysis of the averaged acceleration data.

[0039] The observation variables to be subjected to the principal component analysis are a data group consisting of averaged joint angle data of each joint as viewed from a plurality of directions for the plurality of joints at each measurement time within one walking cycle. FIG. 10 shows an example of the averaged joint angle data. As shown in FIG. 10, one averaged joint angle data is made up of averaged joint angle data for a plurality of joints. The averaged joint angle data for each joint is made up of averaged joint angle data for a plurality of directions. That is, the entirety of the plurality of graphs in FIG. 10 becomes one observation variable. For example, when using the measured values ​​of the averaged joint angle data at 1% intervals in the time axis direction within the walking cycle, the number of components of each observation variable is 101 (0% to 100%) x 3 (sagittal plane, frontal plane, horizontal plane) x 4 (pelvis, hip joint, knee joint, ankle joint). In the present invention, the method of principal component analysis is used to convert these many components into M principal components.

[0040] Specifically, first, (1) calculate the unstandardized principal component scores of the averaged joint angle data and the eigenvectors of each principal component axis (hereinafter referred to as "joint angle eigenvectors"). Here, M unstandardized principal component scores with eigenvalues ​​of 1 or more for each principal component axis obtained by the principal component analysis are selected as the calculated unstandardized principal component scores. (2) Next, standardization processing (average 0, variance 1) is performed on the unstandardized principal component scores of the averaged joint angle data calculated in (1) above to obtain standardized principal component scores B1, B2, ..., B M The joint angle eigenvector of each principal component axis calculated in (1) above is stored in the storage unit 160 as one of the analysis parameters 169.

[0041] A multiple regression model is created by performing multiple regression analysis using the explanatory variable (independent variable) A and the response variable (dependent variable) obtained by the above process. The multiple regression model can be expressed by the following formula.

[0042] B1=α 11 ·A1+α 21 ·A2+α 31 ·A3+ … +α N1 A N +β1 B2=α 12 ·A1+α22 ·A2+α 32 ·A3+ … +α N2 A N +β2 B3=α 13 ·A1+α 23 ·A2+α 33 ·A3+ … +α N3 A N +β3 … B M = α 1M ·A1+α 2M ·A2+α 3M ·A3+ … +α NM A N +β M

[0043] The above formula can be expressed as a vector as follows: B = αA + β

[0044] Here, α is an N×M-dimensional coefficient parameter, and β is an M-dimensional coefficient parameter. These α and β are stored in the storage unit 160 as one of the analysis parameters 169.

[0045] 11, the walking posture analysis unit 140 calculates right side joint angle data 163a from the right average acceleration data 162a, and calculates left side joint angle data 163a from the left average acceleration data 162b. Hereinafter, the analysis process of the walking posture in the walking posture analysis unit 140 will be described in detail.

[0046] First, the walking posture analysis unit 140 converts the right foot averaged acceleration data 162a stored in the storage unit 160 into N principal components using a principal component analysis technique (step S8 in FIG. 2). Specifically, the right averaged acceleration data 162a of each axis is multiplied by the acceleration eigenvector included in the analysis parameters 169 to calculate a non-standardized principal component score of the right averaged acceleration data, and further performs standardization (average 0, variance 1) processing to calculate a standardized principal component score of the right averaged acceleration data. This standardized principal component score of the right averaged acceleration data is the explanatory variable (independent variable) A of a multiple regression model consisting of N-dimensional principal components.

[0047] Next, the walking posture analysis unit 140 uses the coefficient parameters of the multiple regression model included in the analysis parameters 169, i.e., the above-mentioned multiple regression model, to calculate the standardized principal component scores of the joint angle data, which is the objective variable (dependent variable) B, from the explanatory variable (independent variable) A (step S9 in FIG. 2). Here, the objective variable (dependent variable) B is M-dimensional data.

[0048] Next, the walking posture analysis unit 140 uses the standardized principal component scores of the calculated M pieces of joint angle data as principal components and inversely converts the principal components into joint angle data using a method of principal component analysis (step S10 in FIG. 2). Specifically, the walking posture analysis unit 140 calculates the joint angle data of each joint (pelvis, hip joint, knee joint, ankle joint) of the right half of the body in each direction (sagittal plane, frontal plane, horizontal plane) in one right walking cycle using the standardized principal component scores of the calculated joint angle data and the joint angle eigenvector included in the analysis parameters 169. The walking posture analysis unit 140 stores the calculated joint angle data of each joint of the right half of the body in the storage unit 160 as right joint angle data 163a.

[0049] The walking posture analysis unit 140 performs a similar analysis process to calculate left side joint angle data 163b based on averaged acceleration data 162b related to the left foot, and stores the data in the storage unit 160. Note that the analysis parameters 169 used in this analysis process are common to the left and right.

[0050] The display control unit 150 performs display control for visualizing the right joint angle data 163a and the left joint angle data 163b stored in the storage unit 160 (step S11 in FIG. 2). Specifically, the display control unit 150 is made up of well-known three-dimensional motion analysis software, and displays each of the joint angle data 163a and 163b as a skeletal movement within one walking cycle on a predetermined display device 300 in three dimensions as a repeated moving image. There is no restriction on the form in which the skeleton is displayed by the display control unit 150. FIG. 12 is an example of an image displaying a skeleton 301 at the start of walking and a skeleton 302 at the end of walking.

[0051] According to the walking posture analysis device 100 of this embodiment, the posture of the subject while walking is analyzed based on the acceleration data while the subject is walking. This acceleration data can be easily and inexpensively acquired by the acceleration sensor 200. Therefore, the posture of the subject while walking can be analyzed simply and inexpensively.

[0052] Although one embodiment of the present invention has been described in detail above, the present invention is not limited to the above embodiment, and various improvements and modifications may be made without departing from the spirit and scope of the present invention.

[0053] For example, in the above embodiment, the analysis parameters 169 used by the walking posture analysis unit 140 are stored in advance in the memory unit 160, but it is also possible to access a specified management server or storage via a network for each analysis process or periodically to obtain the latest analysis parameters 169 from the management server, etc.

[0054] In the above embodiment, the walking posture analysis device 100 is implemented as a single device including the acceleration sensor 200, the acceleration data acquisition unit 110, the data range setting unit 120, the walking cycle determination unit 130, the walking posture analysis unit 140, the display control unit 150, the storage unit 160, and the display device 300, but these may be implemented in any combination and distributed across multiple devices. For example, a device for acquiring acceleration data, a device for performing analysis processing, and a device for displaying the analysis results may each be implemented as separate devices. [Explanation of symbols]

[0055] 100…Walking posture analysis device 110...Acceleration data acquisition unit 120...Data range setting section 130...Walking cycle determination unit 140…Walking posture analysis section 150...Display control unit 160...Storage section

Claims

1. an acceleration data acquisition unit that acquires acceleration data of a walking subject measured by an acceleration sensor; a walking cycle determination unit that divides the acquired acceleration data into a first walking cycle, which is a walking cycle from when one foot of the subject touches the ground to when the other foot touches the ground next, and a second walking cycle, which is a walking cycle from when the other foot of the subject touches the ground to when the other foot touches the ground next, and determines whether the first walking cycle and the second walking cycle are walking cycles related to the right foot or walking cycles related to the left foot, respectively; and a walking posture analysis unit that calculates right-side joint angle data, which is angle data of the joints of the right half of the body in a walking cycle, from acceleration data in the walking cycle related to the right foot, using a multiple regression model calculated in advance, and calculates left-side joint angle data, which is angle data of the joints of the left half of the body in the walking cycle, from acceleration data in the walking cycle related to the left foot. A walking posture analysis device characterized by the above.

2. The multiple regression model is calculated in advance using N-dimensional (N: natural number) principal components obtained by principal component analysis of the acceleration data for model calculation acquired in advance as explanatory variables, and M-dimensional (M: natural number) principal components obtained by principal component analysis of the right joint angle data or the left joint angle data for model calculation corresponding to the acceleration data for model calculation as objective variables.

2. The walking posture analysis device according to claim 1.

3. The principal component analysis of the acceleration data is performed by using a data group consisting of acceleration data for each of a plurality of axes at each measurement time within one walking cycle as an observation variable, The principal component analysis of the acceleration data uses a data group consisting of angle data of each joint as viewed from multiple directions for each joint at each measurement time within one walking cycle as an observation variable.

3. The walking posture analysis device according to claim 2.

4. The walking cycle determination unit determines whether the first walking cycle or the second walking cycle is a walking cycle related to the right leg or a walking cycle related to the left leg based on an average value of acceleration data in a left-right direction with respect to a walking direction within a predetermined time section in the first walking cycle or the second walking cycle.

4. The walking posture analysis device according to claim 1, wherein the walking posture analysis device is a walking posture analysis device.

5. A method for analyzing walking posture using a computer, comprising: acquiring acceleration data of a walking subject measured by an acceleration sensor; a step of dividing the acquired acceleration data into a first walking cycle, which is a walking cycle from when one foot of the subject touches the ground to when the other foot touches the ground next, and a second walking cycle, which is a walking cycle from when the other foot of the subject touches the ground to when the other foot touches the ground next, and determining whether the first walking cycle and the second walking cycle are walking cycles related to the right foot or the left foot, respectively; and calculating right-side joint angle data, which is angle data of the joints of the right half of the body in a walking cycle, from the acceleration data in the walking cycle related to the right foot, using a multiple regression model calculated in advance, and calculating left-side joint angle data, which is angle data of the joints of the left half of the body in the walking cycle, from the acceleration data in the walking cycle related to the left foot. A walking posture analysis method comprising the steps of:

6. A computer is caused to function as the walking posture analysis device according to claim 1. A walking posture analysis program characterized by:

Images