JP2015195913A - Posture estimation method, posture estimation program and posture estimation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for improving versatility of posture estimation.SOLUTION: A posture estimation device 100 calculates a first angle from sensor values of sensors 1 to S, and calculates a second angle by following a model for connecting smoothly each bone connected mutually in a backbone and a movable range of each bone included in the backbone. Further, the posture estimation device 100 calculates a third angle by synthesizing together the first angle and the second angle.

Description

  The present invention relates to a posture estimation method, a posture estimation program, and a posture estimation device.

  As an example of the posture of the human body, attempts have been made to measure the shape of the back surface (hereinafter referred to as the spine shape). For example, motion capture may be used for measuring the spine shape. For example, in the case of optical motion capture, a marker attached to a human body is photographed by a plurality of cameras, and the position of the marker measured using the parallax of an image captured by each camera is recorded. When these markers are attached along the spine, the position of each marker is obtained as a spine shape.

  In addition, proposals have been made to visualize the spine shape using a plurality of acceleration sensors. In this case, it aims to measure the spine shape by calculating the inclination of each sensor with respect to the gravitational acceleration from the output of each of the three-axis acceleration sensors attached along the spine. The output of the acceleration sensor includes a component of motion acceleration generated by acceleration / deceleration of the sensor in addition to a component of gravitational acceleration generated by the inclination of the sensor. In order to remove the motion component, a method using an angular velocity sensor in terms of hardware and a method using a Kalman filter for removing high frequencies through a low-pass filter have been reported in terms of software.

JP 2012-205816 A JP 2002-163646 A JP 2010-263930 A JP 2012-98254 A

  However, the above technique has one aspect lacking in versatility, as will be described below.

  For example, when the above-described optical motion capture system is used, the spine shape can be measured only in a situation where a plurality of cameras can fit the marker within the imaging range, so that the application range is limited to the imaging range of the camera. . Therefore, it is difficult to monitor the posture on a daily basis using an optical motion capture system, and its versatility is limited.

  In addition, when using an acceleration sensor, the spine shape cannot be measured under a dynamic situation in which intense exercise such as running is performed. In such a dynamic situation, the motion acceleration component and the gravitational acceleration component cannot be separated only by removing the high-frequency component, and a curve that cannot be assumed in the structure of the spine (which cannot be assumed) is a spine. In some cases, it is output as a shape. Thus, even when using an acceleration sensor, there is a limitation that measurement is difficult under dynamic conditions, so that its versatility is limited. When the angular velocity sensor is used, the power consumption increases, and the continuous operation time becomes shorter than when only the acceleration sensor is used. When applying the Kalman filter, a forward mapping function is necessary, but it is extremely difficult to obtain from every movement of daily life. Furthermore, enormous processing time is required.

  In one aspect, an object of the present invention is to provide a posture estimation method, a posture estimation program, and a posture estimation device that can improve the versatility of posture estimation.

  According to one aspect of the posture estimation method, the computer calculates the first angle from the value of the acceleration sensor and smoothly connects the bones connected to each other in the spine, and the range of motion of the bones included in the spine. A process of calculating a third angle by calculating a second angle and combining the first angle and the second angle is executed.

  The versatility of posture estimation can be improved.

FIG. 1 is a block diagram illustrating a functional configuration of the posture estimation apparatus according to the first embodiment. FIG. 2 is a diagram illustrating an example of sensor attachment. FIG. 3 is a diagram illustrating an example of a sagittal plane and a frontal plane. FIG. 4 is a diagram illustrating an example of curve visualization. FIG. 5 is a diagram illustrating an example of load data. FIG. 6 is a flowchart illustrating a procedure of parameter registration processing according to the first embodiment. FIG. 7 is a flowchart illustrating the procedure of the posture estimation process according to the first embodiment. FIG. 8 is a flowchart illustrating the procedure of the load calculation process according to the first embodiment. FIG. 9 is a diagram illustrating an example of an estimation result of a spine shape during forward bending. FIG. 10 is a diagram illustrating an example of a spine shape in a walking cycle. FIG. 11 is a diagram illustrating an example of quantification of the spine shape. FIG. 12 is a schematic diagram illustrating an example of a computer that executes a posture estimation program according to the first and second embodiments.

  A posture estimation method, a posture estimation program, and a posture estimation device according to the present application will be described below with reference to the accompanying drawings. Note that this embodiment does not limit the disclosed technology. Each embodiment can be appropriately combined within a range in which processing contents are not contradictory.

[Configuration of posture estimation device]
First, the functional configuration of the posture estimation apparatus according to the present embodiment will be described. FIG. 1 is a block diagram illustrating a functional configuration of the posture estimation apparatus according to the first embodiment. The posture estimation apparatus 100 shown in FIG. 1 estimates the posture of the human body. As an example, the posture estimation device 100 uses a plurality of sensors 1 to S attached to the back of the human body. Posture estimation processing for estimating

  As an aspect, the posture estimation apparatus 100 can be implemented by causing a desired computer to install a posture estimation program in which the posture estimation processing described above is realized as package software, online software, or an application program.

  Here, the above attitude estimation program can be installed in any device as long as it is a computer equipped with a processor and memory, but as an example, it is installed in a mobile communication terminal such as a smartphone, a mobile phone, or a PHS (Personal Handyphone System). Can be made. In addition, the posture estimation program can be installed not only on mobile communication terminals but also on all mobile terminal devices including tablet terminals and slate terminals. As described above, when the posture estimation program is installed in the mobile terminal device, the mobile terminal device can function as the posture estimation device 100. The posture estimation program is not limited to the portable terminal device, and can be installed in general information processing devices such as desktop and notebook personal computers.

  As illustrated in FIG. 1, the posture estimation apparatus 100 includes an acquisition unit 110, a first calculation unit 120, a parameter storage unit 130, a parameter registration unit 140, a parameter setting unit 150, a second calculation unit 160, It has the 3rd calculation part 170, the shape visualization part 180, and the load calculation part 190. FIG.

  The posture estimation apparatus 100 may include various functional units included in a known computer in addition to the functional units illustrated in FIG. For example, when the posture estimation apparatus 100 is implemented as a stationary terminal, the posture estimation apparatus 100 may further include an input device such as a keyboard and a mouse. Further, when the posture estimation apparatus 100 is implemented as a mobile communication terminal, it further includes functional units such as an antenna, a wireless communication unit connected to the mobile communication network, and a GPS (Global Positioning System) receiver. It doesn't matter.

  The acquisition unit 110 is a processing unit that acquires sensor values output by the plurality of sensors 1 to S.

  As an embodiment, the acquisition unit 110 acquires sensor values of each component of the X axis, the Y axis, and the Z axis from the plurality of three-axis acceleration sensors 1 to S. Here, an arbitrary period can be adopted as a sampling period in which the triaxial acceleration sensors 1 to S measure the triaxial acceleration, and the triaxial acceleration is transmitted from the triaxial acceleration sensors 1 to S to the acquisition unit 110. An arbitrary period can be adopted as a period in which the time series data is uploaded.

  FIG. 2 is a diagram illustrating an example of sensor attachment. FIG. 2 shows an example of sensor attachment when viewed from the sagittal plane. In the example of FIG. 2, a case where ten three-axis acceleration sensors 1 to 10 are mounted on the back of a human body is illustrated, but the embodiment is not limited to this, and the number of sensors attached to the back is one. It may be one or a plurality of numbers other than ten. FIG. 2 shows a schematic diagram of attaching the sensors 1 to 10 on the spine, but actually the sensors 1 to 10 are attached to the surface of the back.

  As shown in FIG. 2, the spine, the so-called spinal column, includes the cervical, thoracic, and lumbar, as well as the sacrum and coccyx. Of these, the cervical vertebra generally includes seven bones from the first cervical vertebra C1 to the seventh cervical vertebra C7. The thoracic vertebra includes 12 bones from the first thoracic vertebra Th1 to the twelfth thoracic vertebra Th12. The lumbar vertebra includes five bones from the first lumbar vertebra L1 to the fifth lumbar vertebra L5.

In the example shown in FIG. 2, the sensor 1 is attached to the cervical vertebra, for example, C7. In addition, six sensors 2 to 7 are attached to the thoracic vertebra, for example, Th2, Th4, Th6, Th8, Th10, and Th12. In addition, three sensors 8-10 are attached to the lumbar vertebrae, eg L2, L4 and L5. Hereinafter, the sensor value acquired at time t from the jth attached sensor may be referred to as “q _j (t)”. The q _j (t) includes acceleration values of the X, Y, and Z axis components. In this case, the sensor value Q (t) acquired from each of the sensors 1 to S at time t can be expressed as {q ₁ (t),..., Q _S (t)}.

  The first calculation unit 120 is a processing unit that calculates an inclination angle with respect to gravitational acceleration from the sensor values of the sensors 1 to S acquired by the acquisition unit 110.

As one embodiment, the first calculation unit 120 obtains the inclination angle a ₁ on the sagittal plane and the inclination angle a ₂ on the frontal plane every time the sensor values of the sensors 1 to S are acquired by the acquisition unit 110. Is calculated for each sensor. FIG. 3 is a diagram illustrating an example of a sagittal plane and a frontal plane. As shown in FIG. 3, the front face indicates a plane that divides the human body into the abdominal side and the back side, that is, the front side and the rear side. On the other hand, the sagittal plane refers to a plane that divides the human body into left and right. These frontal plane and sagittal plane are orthogonal to each other. That is, obtaining the sagittal plane-like inclination angle a ₁ means obtaining the degree of forward bending or backward bending, while obtaining the inclination angle a ₂ on the front face is the degree of lateral bending on the left and right. Will be asked. Among these, the inclination angle a ₁ on the sagittal plane can be calculated by the following equation (1). Thereby, the sagittal plane-shaped inclination angles a _{1, j} (t) for each of the sensors 1 to S are obtained. Further, the inclination angle a ₂ on the front face can be calculated by the following equation (2). As a result, a front facet-like inclination angle a _{2, j} (t) for each of the sensors 1 to S is obtained. In the following equations (1) and (2), “x” indicates acceleration in the X-axis direction, “y” indicates acceleration in the Y-axis direction, and “z” indicates acceleration in the Z-axis direction. Point to. Incidentally, the formula (1) and the inclination angle a ₂ of the inclination angle a ₁ and frontal on sagittal plane unit is [rad] by the equation (2) is calculated below.

As described above, after the inclination angle a _{1, j} (t) and the inclination angle a _{2, j} (t) are obtained for each sensor, the first calculation unit 120 corrects the rotation when the sensors 1 to S are attached. can do. That is, the sensors 1 to S are not necessarily mounted straight along the midline of the spine on the frontal plane, but may be mounted by rotating to the left or right rather than the midline. It is also possible to correct the amount of rotation attached to the left and right at the time of attachment as an offset. Specifically, as shown in the above equation (2), before the inclination angle a ₂ of the face value, the component on Y-Z plane, is calculated using the ratio of the X component. Here, the inclination angle in a reference posture, for example, an upright posture is (a ₀₁ , a ₀₂ ), the inclination angle in an arbitrary posture is (a ₁ , a ₂ ), and the inclination angle in an arbitrary posture is expressed by the following equation (3). Is redefined. In the following formula (3), it is assumed that the sensor plane and the frontal plane are parallel. With such assumptions, X-axis, so that coincides with the normal direction components of the frontal, it is adopted the inclination angle a ₁ at any position in the inclination angle on the sagittal plane (a _1).

(A ₁ , a ₂ ): = (a ₁ , a ₂ −a ₀₂ ) (3)

By the above correction, the corrected tilt angles A (t) of the sensors 1 to S are obtained for each sagittal plane and frontal plane. That is, A (t) is expressed by the inclination angles a _1,1 (t) to a _{1, S} (t) on the sagittal planes of the sensors 1 to S, and Inclination angles a _2,1 (t) to a _{2, S} (t) on the frontal plane of the sensors 1 to S are included.

  Here, when the sensors 1 to S are not mounted corresponding to each bone included in the spine, in other words, when the number of sensors S> the number M of the spines, the first calculation unit 120 calculates the following formula (5 ), The number of row elements in the matrix is adjusted to the number of spines M. For example, the first calculation unit 120 calculates the inclination angle of the bone element existing between the sensors by linear interpolation, and interpolates the number of elements in the row from the number S of sensors to the number M of the spines. Hereinafter, the inclination angle of each bone 1 to M of the spine obtained from the acceleration measured by the sensors 1 to S may be referred to as “measurement angle”.

  In the subsequent functional units, processing is executed individually for each of the sagittal plane and the front face value, but the processing content executed between the sagittal plane and the front face value is the same. The description will be made focusing on the processing in the sagittal plane.

  The parameter storage unit 130 is a storage unit that stores various parameters used for posture estimation.

For example, the parameter storage unit 130, as an example of a constraint of the spine shape used in the calculation of the model angle by the second calculation unit 160 described later, and stores the movable range sigma ₁ of each bone. The “model angle” here refers to the angle of each bone similarly to the measurement angle described above, whereas the measurement angle is calculated from the sensor values of the sensors 1 to S, whereas the model angle is Under the initial value set by the parameter setting unit 150, the angle of each bone calculated according to a model that smoothly connects bones adjacent to each other in the spine. The above-mentioned “movable range Σ _{1 of} each bone” can be determined by the angle of the bone at the time of forward bending and backward bending, and the angle of one of the bones at the time of forward bending or backward bending is sampled. It can also be defined by estimating the angle of one bone from the angle of one bone. The “movable range Σ _{1 of} each bone” can be expressed as (σ _1,1 ,..., Σ _{1, M} ).

In addition, the parameter storage unit 130 stores the angle Θ ₁ of each bone in the upright posture as an example of an initial value used in the calculation of the model angle by the second calculation unit 150. Such an “upright angle Θ ₁ ” can be obtained from the acceleration of each component of the X, Y, and Z axes of the sensors 1 to S measured in the upright posture. The upright angle Θ ₁ can be expressed as (θ _1,1 ,..., Θ _{1, M} ).

Further, the parameter storage unit 130 stores a distance L between bones as an example of a parameter used for visualization of the spine shape. As the “distance L between bones”, a distance measured for each bone in a posture with no back curvature, for example, prone position can be adopted. Such “distance L between bones” can be expressed as (L ₁ ,..., L _M ).

The parameter registration unit 140 is a processing unit that registers parameters in the parameter storage unit 130 described above. Here, the movable range Σ ₁ of each bone, the upright angle Θ _1, and the distance L between the bones can be measured in advance, or when the human body wears the sensors 1 to S. It can also be calculated from the sensor value at the time when the posture such as upright, forward bending, backward bending or prone position is estimated.

As one embodiment, when the sensor value Q (t) is acquired from the sensors 1 to S, the parameter registration unit 140, like the first calculation unit 120 described above, uses the above formulas (1) and (2). Accordingly, the inclination angle a ₁ on the sagittal plane and the inclination angle a ₂ on the frontal plane are calculated for each sensor. And the parameter registration part 140 correct | amends the rotation at the time of attachment of each sensor 1-S according to said Formula (3). In addition, the parameter registration unit 140 calculates the inclination angle of the bone elements existing between the sensors by linear interpolation, and interpolates the number of elements in the row from the number S of sensors to the number M of the spines. Thereafter, the parameter registration unit 140 stores the measurement angle A (t) obtained in this way in an internal memory (not shown). Subsequently, the parameter registration unit 140 stores, in the internal memory, a correspondence table in which the inclination angles of the bones 1 to M that can be calculated in each posture such as upright, forward bending, backward bending, and prone position are associated with each other. The stored measurement angle A (t) is compared. At this time, when the measurement angle A (t) is within a predetermined range from the measurement angles of the postures in the correspondence table, the parameter registration unit 140 corresponds to the posture among the parameters stored in the parameter storage unit 130. The parameter to be updated is overwritten on the measurement angle A (t). Through such processing, the measurement angle at a specific moment when the user takes the above posture in daily life can be cut out and registered as a parameter.

  The parameter setting unit 150 is a processing unit that sets parameters used for calculating the model angle and the estimated angle. The “estimated angle” here refers to the angle of the bone estimated by combining the measured angle and the model angle. Although these model angles and estimated angles will be described in detail later, the model angles and estimated angles of the bones 1 to M are repeatedly calculated. For this reason, the following description will be given assuming that the number of iterations is “k”.

As an embodiment, the parameter setting unit 150 sets a constraint condition used by the second calculation unit 150 to be described later for calculating the model angle. Specifically, the parameter setting unit 150 determines whether or not the sensor value Q (t) acquired from the sensors 1 to S corresponds to a specific operation pattern. As an example of such an operation pattern, data in which an operation such as traveling or rehabilitation and a sensor value corresponding to the operation are defined can be employed. The parameter setting unit 150, if the sensor value Q (t) does not correspond to the operation of such driving and rehabilitation, sets the movable range sigma ₁ of each bone stored in the parameter storage unit 130 as a constraint condition. On the other hand, when the sensor value Q (t) corresponds to an operation such as traveling or rehabilitation, the parameter setting unit 150 sets a movable range corresponding to the operation. For example, the parameter setting unit 150, since the travel time can be estimated that the movable range of the spine than the stop becomes narrow, coefficient of less than 1 to movable range sigma ₁ stored in the parameter storage unit 130, for example, multiplying 1/2 By doing so, the movable range corresponding to the traveling time is set. In addition, since the parameter setting unit 150 can estimate that the movable range becomes wider than other states during rehabilitation, the movable range Σ ₁ stored in the parameter storage unit 130 is multiplied by one or more coefficients, for example, 1.2. Thus, the movable range corresponding to rehabilitation is set. Hereinafter, the movable range set as the constraint condition by the parameter setting unit 150 may be referred to as “movable range Σ ₂ ”. This movable range Σ ₂ is expressed as (σ _2,1 ,..., Σ _{2, M} ).

As another embodiment, the parameter setting unit 150 uses the above model angle as an initial value of a parameter used for calculating a weighting factor used when the third calculation unit 170 described later synthesizes the measurement angle and the model angle. Set the initial value of the bone angle. For example, the parameter setting unit 150 uses the bone angle one hour before as the initial value so that the angle of each bone can be estimated accurately in a short time. This is because if the angle change of the bone is smooth and the angular velocity is not so large, the estimated angle Φ ₂ (t) at time (t) becomes the estimated angle Φ ₂ (t−1) at time (t−1). This is because it can be expected to be close. From this, when the time (t) is after the first time, that is, when time t> 1, the parameter setting unit 150 calculates the estimated angle Φ ₂ (t−1) one time before. The angle is set as the initial value Φ ₂ (t) ⁽⁰⁾ . On the other hand, when it is the first time after starting the posture estimation apparatus 100, that is, when time t = 1, the estimated angle is not estimated one time ago. In this case, the parameter setting unit 150 sets the upright angle Θ ₁ stored in the parameter storage unit 130 as the initial angle value Φ ₂ (t) ⁽⁰⁾ . In addition, the parameter setting unit 150 may perform an estimated angle Φ estimated one iteration before the iteration count k when the iteration count k is incremented by one or more. ₂ (t) ^(K-1) is set as Φ ₂ (t) ^(K) instead of the initial angle value Φ ₂ (t) ⁽⁰⁾ .

  The second calculation unit 160 sets the model angle according to the model that smoothly connects the bones adjacent to each other in the spine under the initial value set by the parameter setting unit 150, and the movable range set by the parameter setting unit 150. A processing unit to calculate.

As one embodiment, the second calculation unit 160 calculates the model angle Φ _{1, n} (t) ^(k) for each bone of the spine using the following equation (6) in which the above model is realized. That is, the range in which each bone included in the spine can move is limited due to the structure of the spine. For this reason, in the following equation (6), a model is constructed so that an angle that is impossible (not conceivable) in the structure of the spine can be corrected by using a relative angle with the adjacent bone. “Δθ _{1, n} ” in the following equation (6) is assumed to be “θ _{1, n} −θ _{1, n−1} ”. Note that, here, as an example, a case where a relative angle with an adjacent bone is adopted as a model is illustrated, but the embodiment is not limited thereto. For example, a relative angular velocity or an angular velocity can be adopted instead of the relative angle, or a relative angle, a relative angular velocity, and an angular velocity can all be adopted.

  The third calculation unit 170 is a processing unit that calculates an estimated angle by combining the measurement angle calculated by the first calculation unit 120 and the model angle calculated by the second calculation unit 160.

As one embodiment, the third calculation unit 170 is a coefficient that determines the weight of the measurement angle A (t) and the model angle Φ _{1, n} (t) ^(k) , that is, the weight coefficient η _n in the following equation (7). ^(K) is obtained for each bone 1 to M. At this time, the third calculation unit 170 uses the measurement angle A (t) calculated by the first calculation unit 120 and the movable range Σ ₂ set by the parameter setting unit 150 to use the weight coefficient η _n described above. ^(K) is determined. At this time, the third calculation unit 170 uses the relative angle of each bone estimated once before the number of iterations and the movable range Σ ₂ set by the parameter setting unit 150 each time an iterative calculation is executed. Then, the weight coefficient η _n ^(k) is recalculated. For example, the third calculation unit 170 determines the weighting coefficient η _n ^(k) depending on whether or not the following formula (8) is satisfied. For example, when the following equation (8) is satisfied, the third calculation unit 170 sets “1” to the weighting coefficient η _n ^(k) , while when the following equation (8) is not satisfied: , “0” is set to the weighting coefficient η _n ^(k) . Here, the case where either the measurement angle or the model angle is adopted by setting “1” or “0” as the weighting factor is illustrated, but the weighting factor is not necessarily “1” or “0”. It is not necessary to select either one alternatively. Since the range of movement of human bones varies depending on daily changes, the boundary of the range of movement is uncertain. For this reason, the weighting factor may be determined using a Gaussian function or the like instead of determining the above weight η by a step function, that is, “1” or “0”. It is also possible to set a smaller value the larger the discrepancy of the movable range sigma ₂ the weight eta.

−σ_２，ｎ≦ａ_１，ｎ（ｔ）^（ｋ）−Φ_{２，ｎ−１}（ｔ）^（ｋ）≦σ_２，ｎ・・・（８）

−σ _{2, n} ≦ a _{1, n} (t) ^(k) −Φ _{2, n−1} (t) ^(k) ≦ σ _{2, n} (8)

After determining the weight coefficient η _n ^(k) , the third calculation unit 170 sets the estimated angle φ _{2, n} (t) ^(k) that minimizes the function L of the following equation (9) for each bone 1 to M. Ask for. In this equation (9), the problem of maximizing “P” is replaced by the problem of minimizing “L” by taking the logarithm of “P” in the above equation (7) and inverting the sign. . As an example, Φ _{2, n} (t) ^(k) that minimizes “L” in the following equation (9) can be obtained by using an iterative solution method called steepest descent method. Note that “λ” in the following equation (9) is an arbitrary constant. Here, the same constant value is set for “λn”, but the value of λn may be determined according to the reliability of each measurement angle. That is, the third calculation unit 170 substitutes the update amount (δL / δφ _{2, n} (t) ^(k) ) calculated by the following equation (10) into the following equation (11) to obtain φ _{2, n} (t) ^(k) is updated. “Β” in the equation (11) is also an arbitrary constant. By repeating such updating of φ _{2, n} (t) ^(k) for every bone 1 to M while incrementing the number of bones n, φ ₂ (t) ^(k) (= φ _{2 , 1} (t) ^(k) ,..., Φ _{2, M} (t) ^(k) ). Thereafter, the third calculation unit 170 updates the amount (δL / δ) of the update amount (δL / δφ _{2, n} (t) ^(k) ) calculated by the following equation (10) calculated one time before the number of iterations. δφ _{2, n} (t) ^(k−1) ) or the update amount (δL / δφ _{2, n} (t) ^(k) ) is smaller than a predetermined threshold, for example, ^1.0−10. Until then, φ _{2, n} (t) ^(k) is repeatedly updated. Although the case where the amount of update amount is used as the update end condition is illustrated here, the update may be repeatedly executed a certain number of times, for example, 1000 times.

When the update of the estimated angle Φ ₂ (t) ^(k) is repeated in this way, the third calculation unit 170 increments the number of repetitions k. Accordingly, the parameter setting unit 150 uses Φ ₂ (t) ^(k−1) one iteration before the value Φ ₂ used for calculation of the model angle and the weighting factor instead of the initial value of the angle. (T) Set as ^(k) .

In the above equation (7), Φ ₁ considers the smoothness of the shape, but as shown in the following equation (12), the relationship between any bones not adjacent to the three items (Φ ₃ ) It is also possible to estimate a more accurate angle by adding a term to be considered and adding a term that takes temporal smoothness into four items (Φ ₄ ).

P = η ₁ φ ₁ + η ₂ φ ₂ + η ₃ φ ₃ + η ₄ φ ₄ (12)

  The shape visualization unit 180 is a processing unit that visualizes the spine shape using the estimated angle calculated by the third calculation unit 170.

As one embodiment, the shape visualization unit 180 uses the estimated angle Φ ₂ (t) calculated by the third calculation unit 170 and the distance L between the bones stored in the parameter storage unit 130, and Coordinates are calculated, and the coordinates of each bone are connected by broken lines or curves. As a result, a higher-order curve of the spine can be approximated and visualized.

FIG. 4 is a diagram illustrating an example of curve visualization. The unit of the vertical axis and the horizontal axis shown in FIG. 4 is [cm]. FIG. 4 shows six graphs representing the spine shape when the user wearing the sensors 1 to S is traveling. Among the six graphs shown in FIG. 4, the left three graphs show the inclination angles α _{1 to} α _S obtained from the sensor values of the sensors _{1 to S} and the distances d _{1 to} d _S between the sensors. The coordinates of each sensor are plotted using, and the coordinates of each sensor are connected by a curve. On the other hand, in the three graphs on the right side, the coordinates of the bones _{1 to M} are plotted using the estimated angles φ _{1 to} φ _M of the bones _{1 to M} and the distances L _{1 to} L _M between the bones. The coordinates of bones 1 to M are shown connected by curves. In the example of FIG. 4, the graphs arranged on the left and right are of the same time, and the time has passed in the order of the upper two graphs, the middle two graphs, and the lower two graphs. To do.

  As shown in FIG. 4, the upper left and right graphs each show a normal spine shape, that is, an S-shaped spine. While the normal spine shape is shown in the graph, the left graph outputs a shape that is impossible (cannot be assumed) due to the structure of the spine. Similarly, in the lower left and right graphs after more time, the right graph shows the normal spine shape, while the left graph shows a shape that is impossible (unthinkable) due to the structure of the spine. Is output. From these facts, when the inclination angle with respect to the gravitational acceleration obtained from the sensor values of the sensors 1 to S is used as it is, the spine shape cannot be estimated under a dynamic situation in which intense exercise such as running is performed. I understand that. On the other hand, when the estimated angle obtained by the third calculation unit 170 is used, it is understood that the spine shape can be estimated even under a dynamic situation where intense exercise such as running is performed. .

  The load calculation unit 190 is a processing unit that calculates an index related to a load applied to a specific portion of the lumbar spine or spine using the estimated angle calculated by the third calculation unit 170.

As one embodiment, the load calculation unit 190 determines whether or not the variance values of the estimated angles Φ ₂ (t−5) to Φ ₂ (t) over a predetermined period, for example, a period of up to 5 hours before, are equal to or larger than a predetermined threshold. Determine whether. At this time, if the variance value is less than the threshold value, it is determined that the situation is not a dynamic situation, that is, a static situation. On the other hand, if the variance value is greater than or equal to the threshold value, it is determined that the situation is dynamic.

  In a static situation, the load calculator 190 calculates the estimated angle of each bone corresponding to the first lumbar vertebra L1 to the fifth lumbar vertebra L5 among the estimated angles of the bones 1 to M calculated by the third calculator 170. Find the angle of the lumbar spine. Then, the load calculation unit 190 searches for the load corresponding to the previously determined lumbar angle from the load data in which the correspondence relationship between the lumbar angle and the load is defined, and outputs it to a predetermined output destination. As such load data, it is possible to employ data in which changes in the intervertebral disc pressure with respect to the upright position are defined for each angle of the lumbar vertebra in each posture. FIG. 5 is a diagram illustrating an example of load data. FIG. 5 shows that the intervertebral disc pressure changes due to a static posture change. Furthermore, FIG. 5 shows the change in the intervertebral disc pressure with “1” when standing upright for an experimental participant weighing 70 kg. The unfilled graph indicates the research result of “Nachemson”. When using the research results of “Wilke et al.” Shown in FIG. 5 as load data, the load “1” is output when the lumbar angle corresponds to the upright position, while the lumbar angle corresponds to the supine position The load “0.2” is output, the load “0.25” is output when the lumbar angle corresponds to the recumbent position, and the load is output when the lumbar angle corresponds to the forward bending of 20 °. “2.25” is output. Further, when the research result of “Nachemson” shown in FIG. 5 is used as load data, when the lumbar angle corresponds to the upright position, the load “1” is output, while the lumbar angle corresponds to the supine position. In this case, the load “0.25” is output, the load “0.8” is output when the lumbar angle corresponds to the recumbent position, and the lumbar angle corresponds to the forward bending of 20 °. The load “1.5” is output.

In a dynamic situation, the load calculation unit 190 calculates the difference between the measurement angle A (t) and the estimated angle Φ ₂ (t) for each bone 1 to M. In addition, the load calculation unit 190 outputs a portion where the difference between the measurement angle A (t) and the estimated angle Φ ₂ (t) is equal to or greater than a predetermined threshold to a predetermined output destination. That is, when a large acceleration is detected in a very short time at a specific location, it can be seen that an impact has been applied to that location. At this time, it can be estimated that the load to the location is large. As described above, by comparing the difference between the angle corrected in consideration of the structure of the spine and the angle before correction, it is possible to estimate which part is loaded.

  These pieces of information can be output to an arbitrary output destination including a display unit (not shown) included in the posture estimation apparatus 100. For example, when a diagnostic program for executing various diagnoses from the spine shape is installed in the posture estimation apparatus 100, the diagnostic program can be the output destination. In addition, a server device that provides a diagnostic program as a Web service can be used as an output destination. Furthermore, a terminal device used by a person concerned of a user who uses the posture estimation apparatus 100, for example, a medical person such as a caregiver or a doctor, can be used as an output destination. This also enables monitoring services outside the hospital, for example, at home or at home. Needless to say, the diagnostic result of the diagnostic program can also be displayed on the terminal device of the person concerned, including the display unit of the posture estimation apparatus 100.

  1 can be realized by causing a CPU (Central Processing Unit) or MPU (Micro Processing Unit) to execute an attitude estimation program. Each functional unit described above can also be realized by a hard wired logic such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). The functional units include the acquisition unit 110, the first calculation unit 120, the parameter registration unit 140, the parameter setting unit 150, the second calculation unit 160, the third calculation unit 170, the shape visualization unit 180, the load calculation unit 190, and the like. included.

  Further, as an example, a semiconductor memory element can be adopted for the parameter storage unit 130 and the internal memory. For example, examples of the semiconductor memory device include a video random access memory (VRAM), a dynamic random access memory (DRAM), a static random access memory (SRAM), and a flash memory. Further, an auxiliary storage device such as a hard disk or an optical disk may be substituted for a memory device such as a semiconductor memory element.

[Process flow]
Subsequently, the flow of processing of the posture estimation apparatus according to the present embodiment will be described. Here, after describing (1) parameter registration processing executed by posture estimation apparatus 100, (2) posture estimation processing will be described, and (3) load calculation processing will be described thereafter.

(1) Parameter Registration Processing FIG. 6 is a flowchart illustrating the procedure of parameter registration processing according to the first embodiment. This parameter registration processing can be executed in advance before the posture estimation processing is executed, or can be executed in parallel with the posture estimation processing. The parameter registration process can be repeatedly executed as long as sensor values are acquired from the sensors 1 to S.

As shown in FIG. 6, when the sensor value Q (t) is acquired from the sensors 1 to S (step S101), the parameter registration unit 140 performs the sagittal plane according to the above equations (1) and (2). the inclination angle _{a 1} of the upper, front and an inclined angle _{a 2} of the face value is calculated for each sensor (step S102). And the parameter registration part 140 correct | amends the rotation at the time of attachment of each sensor 1-S according to said Formula (3) (step S103).

  Here, when the number of sensors S <the number M of the spines (step S104 Yes), the parameter registration unit 140 linearly interpolates the inclination angles of the bone elements existing between the sensors, thereby measuring the measurement angles of the spines 1 to M. A (t) is calculated (step S105). When the number of sensors S = the number of spines M (No in step S104), the inclination angles of the sensors 1 to S are set as the measurement angles of the bones 1 to M as they are, and the process proceeds to the above step S106.

  After that, the parameter registration unit 140 calculates in step S105 a correspondence table in which the inclination angles of the bones 1 to M that can be calculated in the posture are associated with each posture such as upright, forward bending, backward bending, and prone position. The measured angle A (t) is compared (step S106).

  At this time, when the measurement angle A (t) is within a predetermined range from the inclination angle of the posture in the correspondence table (Yes in step S107), the parameter registration unit 140 includes the parameters stored in the parameter storage unit 130. The parameter corresponding to the posture is overwritten and updated on the measurement angle A (t) (step S108), and the process ends.

(2) Posture Estimation Processing FIG. 7 is a flowchart illustrating the procedure of posture estimation processing according to the first embodiment. This posture estimation process can be repeatedly executed as long as sensor values are continuously acquired from the sensors 1 to S.

As shown in FIG. 7, when the sensor value Q (t) is acquired from the sensors 1 to S (step S301), the first calculation unit 120 is sagittal according to the above equations (1) and (2). the inclination angle _{a 1} on the surface, before the inclination angle _{a 2} of the face value is calculated for each sensor (step S302). And the 1st calculation part 120 correct | amends the rotation at the time of attachment of each sensor 1-S according to said Formula (3) (step S303).

  Here, when the number of sensors S <the number of spines M (step S304 Yes), the first calculation unit 120 measures the spines 1 to M by linearly interpolating the inclination angles of the bone elements existing between the sensors. An angle A (t) is calculated (step S305). When the number of sensors S = the number of spines M (No in step S304), the inclination angles of the sensors 1 to S are set as the measurement angles of the bones 1 to M as they are, and the process proceeds to the above step S306.

  Then, the parameter setting unit 150 sets a movable range as a constraint condition used for calculation of the model angle, or sets an initial value of the bone angle used for calculation of the weight coefficient (step S306). Subsequently, the parameter setting unit 150 sets the values of the counter n that counts the number of bones and the counter k that counts the number of iterations to initial values, for example, zero (step S307). Thereafter, the parameter setting unit 150 increments the bone number counter n by one (step S308).

  Subsequently, the second calculation unit 160 smoothly connects the bones adjacent to each other in the spine under the initial value set in step S306 or the estimated angle set at the time of repetition one time before the repetition, A model angle is calculated according to the movable range set in step S306 (step S309).

Then, the third calculation unit 170 determines the weighting coefficient η _n ^(k) using the measurement angle A (t) calculated in step S305 and the movable range Σ ₂ set in step S306 (step S306 ^). S310).

  Thereafter, the third calculation unit 170 calculates an estimated angle by combining the measurement angle calculated in step S305 and the model angle calculated in step S309 (step S311).

  At this time, when the value of the bone number counter n is smaller than the number of bones M (step S312 Yes), the processing returns to the above step S308, and the above steps S309 to S311 are performed for the next bone. Execute the process.

On the other hand, when the value of the bone number counter n is equal to the bone number M (No in step S312), the third calculation unit 170 determines that the update amount calculated by the following equation (10) satisfies a predetermined end condition. It is determined whether or not it is satisfied (step S313). For example, the third calculation unit 170 may determine whether the update amount is larger than the update amount calculated one iteration before or the update amount is smaller than a predetermined threshold, for example, ^1.0-10 . It is determined whether it corresponds to.

  If the update amount does not satisfy the predetermined end condition (No in step S313), the third calculation unit 170 increments the iteration count counter k by 1 (step S314), and the parameter setting unit 315 In addition to setting the estimated angle one time before instead of the initial value, the initial value is set in the counter n of the number of bones (step S315), and the process returns to step S308.

On the other hand, when the update amount satisfies the predetermined end condition (step S313 Yes), the shape visualization unit 180 calculates the estimated angle Φ ₂ (t) calculated in step S311 and the bone distance stored in the parameter storage unit 130. The distance L is used to calculate the coordinates of each bone, and the coordinates of each bone are connected by a broken line or a curve (step S316). In this way, the higher order curve of the spine is approximated and visualized. Thereafter, the load calculation unit 190 executes a load calculation process for calculating an index relating to a load applied to a specific portion of the lumbar spine or the spine using the estimated angle calculated in step S311 (step S317).

  Then, after the time t is incremented by one (step S318), the process returns to step S301, and then the posture estimation apparatus 100 repeatedly executes the processes from step S301 to step S317 described above.

(3) Load Calculation Process FIG. 8 is a flowchart illustrating the procedure of the load calculation process according to the first embodiment. This load calculation process is a process corresponding to step S317 shown in FIG. 7, and can be started at an arbitrary timing after step S313 Yes.

As shown in FIG. 8, the load calculation unit 190 has a variance value of the estimated angles Φ ₂ (t−5) to Φ ₂ (t) over a predetermined period, for example, a period of up to 5 hours before, a predetermined threshold value or more. Is determined (step S501).

  At this time, if the variance value is less than the threshold (No in step S501), it is determined that the situation is not a dynamic situation, that is, a static situation. In this case, the load calculation unit 190 obtains the angle of the lumbar vertebra from the estimated angle of each bone corresponding to the first lumbar vertebra L1 to the fifth lumbar vertebra L5 among the estimated angles of the bones 1 to M calculated by the third calculation unit 170. (Step S502).

  Then, the load calculation unit 190 searches for the load corresponding to the previously determined lumbar angle from the load data in which the correspondence relationship between the lumbar angle and the load is defined, and outputs it to a predetermined output destination (step S503). And step S504), a process is complete | finished.

On the other hand, if the variance value is greater than or equal to the threshold value (step S501 Yes), it is determined that the situation is dynamic. In this case, the load calculation unit 190 calculates the difference between the measurement angle A (t) and the estimated angle Φ ₂ (t) for each of the bones 1 to M (step S505). In addition, the load calculation unit 190 outputs a portion where the difference between the measurement angle A (t) and the estimated angle Φ ₂ (t) is equal to or greater than a predetermined threshold to a predetermined output destination (step S506), and ends the process. To do.

[Effect of Example 1]
As described above, the posture estimation apparatus 100 according to the present embodiment combines the posture component calculated from the sensor value of the acceleration sensor and the posture component calculated from the spine structure and kinematic characteristics. Calculate the spine shape. For this reason, unlike the above-described optical motion capture system, the application scene is not limited to a specific scene. Further, even under dynamic conditions, it is possible to suppress a case where a curve that cannot be assumed (cannot be assumed) due to the structure of the spine is output as the spine shape. Therefore, according to the posture estimation apparatus 100 according to the present embodiment, versatility of posture estimation can be improved.

  Although the embodiments related to the disclosed apparatus have been described above, the present invention may be implemented in various different forms other than the above-described embodiments. Therefore, another embodiment included in the present invention will be described below.

[Scope of application]
In the first embodiment, the case where the measurement angle of each bone is calculated by linearly interpolating the inclination angle of each sensor when the number of sensors S <the number M of bones is illustrated, but the estimation is not necessarily performed for each bone. The angle may not be calculated, and the estimated angle may be calculated for each sensor by using the inclination angle of each sensor as the measurement angle.

[Walking observation]
FIG. 9 is a diagram illustrating an example of an estimation result of a spine shape during forward bending. FIG. 10 is a diagram illustrating an example of a spine shape in a walking cycle. The posture estimation apparatus 10 can notify an excessive forward tilt or the like during walking by performing walking observation of the spine shape shown in FIGS. 9 and 10. For example, when the similarity between the spine shape estimated in the walking cycle shown in FIG. 10 and the spine shape being bent forward shown in FIG. 9 is equal to or greater than a predetermined value, walking is performed in a forward leaning posture. There is a high risk that the spine is overloaded. In this case, the posture estimation apparatus 10 can notify that effect.

[Numericalization of spine shape]
For example, the posture estimation apparatus 10 can also digitize the spine shape in the same manner as the visualization of the spine shape and the calculation of the load. Specifically, posture estimation apparatus 10 can digitize the spine shape by adding all the estimated angles at time t. FIG. 11 is a diagram illustrating an example of quantification of the spine shape. The vertical axis of the graph shown in FIG. 11 indicates the total value of the estimated angles, and the horizontal axis of the graph shown in FIG. 11 indicates time. As shown in FIG. 11, in the case of a C-shaped curve, the value increases when all estimated angles are added, and the S-shaped curve is canceled and approaches zero. The three curves placed below the graph are representative curves in each of three states (C-shaped, transient state, S-shaped). For example, the posture estimation apparatus 10 can instruct ideal luggage lifting by comparing the C-shaped bending degree, the smoothness of the transient state, and the S-shaped bending degree with ideal values.

[Other implementation examples]
In the first embodiment, the case where the posture estimation apparatus 100 executes the posture estimation processing in a stand-alone manner is exemplified, but the posture estimation device 100 may be implemented as a client server system. For example, the posture estimation apparatus 100 may be implemented as a Web server that provides a posture estimation service, or may be implemented as a cloud that provides a posture estimation service by outsourcing. As described above, when the posture estimation device 100 operates as a server device, a mobile terminal device such as a smartphone or a mobile phone or an information processing device such as a personal computer can be accommodated as a client terminal. When the sensor values of the sensors 1 to S are acquired from these client terminals via the network, the above-described parameter registration processing, posture estimation processing, load calculation processing, and the like are executed, and posture estimation results and estimation results are obtained. It is also possible to provide a posture estimation service by responding to the client terminal with a diagnosis result that has been diagnosed by using it.

[Attitude estimation program]
The various processes described in the above embodiments can be realized by executing a prepared program on a computer such as a personal computer or a workstation. In the following, an example of a computer that executes a posture estimation program having the same function as that of the above-described embodiment will be described with reference to FIG.

  FIG. 12 is a schematic diagram illustrating an example of a computer that executes a posture estimation program according to the first and second embodiments. As illustrated in FIG. 12, the computer 100 includes an operation unit 110a, a speaker 110b, a camera 110c, a display 120, and a communication unit 130. Further, the computer 100 includes a CPU 150, a ROM 160, an HDD 170, and a RAM 180. These units 110 to 180 are connected via a bus 140.

  As shown in FIG. 12, the HDD 170 includes the acquisition unit 110, the first calculation unit 120, the parameter registration unit 140, the parameter setting unit 150, the second calculation unit 160, and the third calculation unit 170 described in the first embodiment. A posture estimation program 170a that exhibits the same functions as the shape visualization unit 180 and the load calculation unit 190 is stored in advance. About this posture estimation program 170a, each acquisition unit 110, first calculation unit 120, parameter registration unit 140, parameter setting unit 150, second calculation unit 160, third calculation unit 170, shape visualization unit illustrated in FIG. Similarly to each component of 180 and the load calculation unit 190, they may be appropriately integrated or separated. In other words, all data stored in the HDD 170 need not always be stored in the HDD 170, and only data necessary for processing may be stored in the HDD 170.

  Then, the CPU 150 reads the posture estimation program 170 a from the HDD 170 and expands it in the RAM 180. Thus, as shown in FIG. 12, the posture estimation program 170a functions as a posture estimation process 180a. The posture estimation process 180a expands various data read from the HDD 170 in an area allocated to itself on the RAM 180 as appropriate, and executes various processes based on the expanded various data. The posture estimation process 180a includes the acquisition unit 110, the first calculation unit 120, the parameter registration unit 140, the parameter setting unit 150, the second calculation unit 160, the third calculation unit 170, the shape visualization unit 180, and the like illustrated in FIG. The process performed in the load calculation part 190, for example, the process shown in FIGS. In addition, each processing unit virtually realized on the CPU 150 does not always require that all processing units operate on the CPU 150, and only a processing unit necessary for the processing needs to be virtually realized.

  Note that the posture estimation program 170a is not necessarily stored in the HDD 170 or the ROM 160 from the beginning. For example, each program is stored in a “portable physical medium” such as a flexible disk inserted into the computer 100, so-called FD, CD-ROM, DVD disk, magneto-optical disk, or IC card. Then, the computer 100 may acquire and execute each program from these portable physical media. In addition, each program is stored in another computer or server device connected to the computer 100 via a public line, the Internet, a LAN, a WAN, etc., and the computer 100 acquires and executes each program from these. It may be.

1, 2,..., S sensor 100 posture estimation device 110 acquisition unit 120 first calculation unit 130 parameter storage unit 140 parameter registration unit 150 parameter setting unit 160 second calculation unit 170 third calculation unit 180 shape visualization unit 190 load Calculation unit

Claims (9)

Computer
The first angle is calculated from the sensor value of the acceleration sensor,
Calculating a second angle according to a model that smoothly connects bones connected to each other in the spine and a range of motion of the bones included in the spine;
A posture estimation method comprising: executing a process of calculating a third angle by combining the first angle and the second angle.   The posture estimation method according to claim 1, wherein the model is defined by at least one of a relative angle, a relative angular velocity, and an angular velocity of any two bones included in the spine. The process of calculating the third angle includes
The posture according to claim 1 or 2, wherein a weighting factor for combining the first angle and the second angle is determined based on an absolute value of the first and second angle differences. Estimation method. The computer is
4. The posture estimation method according to claim 1, wherein the movable range is updated using the first angle when the first angle corresponds to a predetermined posture. 5. The computer is
5. The process according to claim 1, further comprising: setting a movable range corresponding to the specific operation pattern when the sensor value corresponds to the specific operation pattern. 6. Posture estimation method. The computer is
The posture estimation method according to claim 1, further comprising executing a process of visualizing a shape of the spine using the third angle. The computer is
The posture estimation method according to any one of claims 1 to 5, further comprising a step of calculating an index relating to a load applied to a specific portion of the lumbar spine or the spine using the third angle. On the computer,
The first angle is calculated from the sensor value of the acceleration sensor,
Calculating a second angle according to a model that smoothly connects bones connected to each other in the spine and a range of motion of the bones included in the spine;
A posture estimation program for executing a process of calculating a third angle by combining the first angle and the second angle. A first calculation unit for calculating a first angle from a sensor value of the acceleration sensor;
A second calculation unit that calculates a second angle according to a model that smoothly connects bones connected to each other in the spine, and a movable range of bones included in the spine;
And a third calculation unit that calculates a third angle by combining the first angle and the second angle.

Images