JP6322984B2 - Attitude estimation apparatus, method and program - Google Patents

Description

  The present invention relates to an attitude estimation apparatus, method, and program.

  One method for supporting the independent life of elderly people, rehabilitating patients, etc. (hereinafter also referred to as “subjects”) is to detect the daily behavior patterns of the subject. The behavior pattern includes postures such as standing, sitting, lying down, walking, and falling. Information related to the health condition of the subject can be extracted by detecting the posture transition in the behavior pattern of the subject or falling from walking.

  The posture of the subject can be detected by mounting a plurality of acceleration sensors on the subject, for example. However, if acceleration sensors are installed at multiple locations on the subject's body, the configuration of the device that detects the posture becomes complicated and the load on the subject increases, which may cause problems in daily life. is there. On the other hand, if the number of acceleration sensors is reduced, it becomes difficult to identify each posture from the output of the acceleration sensor, so that the posture estimation accuracy of the subject may be lowered.

  In addition, the threshold used to detect the transition from the output of the acceleration sensor to, for example, sitting to standing differs depending on the height of the target person, so if the threshold value is not set for each target person, false detection of posture may occur. (For example, Patent Documents 1 and 2 and Non-Patent Document 1).

  On the other hand, a method for estimating a posture of a subject by creating a learning model for posture estimation and applying the output of an acceleration sensor to the learning model has been proposed (for example, Non-Patent Document 2). However, if one learning model is applied to all subjects, posture estimation accuracy may be reduced. In addition, if a dedicated learning model is created for each target person, it takes time to create a learning model and costs increase. Furthermore, with posture estimation using a learning model, it is difficult to estimate the posture of the subject in real time.

JP 2005-245709 A JP 2010-125239 A

Hiroshi Kurasawa et al., “Attitude estimation method using 3-axis acceleration sensor considering sensor mounting location”, IPSJ SIG 2006 (54), 15-22, May 2006 Naoyuki Hashida et al., “Contribution to identification accuracy of personal adaptation technology in daily behavior identification using acceleration sensor”, IEICE Technical Report 108 (138), 69-74, July 2008

  With the conventional posture estimation method, it is difficult to accurately estimate the posture of the subject person, particularly the standing state and the sitting state.

  SUMMARY OF THE INVENTION An object of the present invention is to provide a posture estimation device, method, and program capable of estimating the posture of a subject person, particularly a standing state and a sitting state with high accuracy.

According to one aspect of the present invention, a three-axis acceleration sensor that detects the acceleration of the upper body of the subject is connected to the subject in which the directions of the first axis and the second axis of the three-axis acceleration sensor are standing. A posture estimation device for estimating a posture of the subject worn so as to coincide with a front-rear direction and a vertical direction, wherein the stationary state of the subject is determined based on a DC signal component included in an output of the three-axis acceleration sensor. Based on an AC signal component included in the output of the three-axis acceleration sensor at the same time as detecting the stationary state of the subject by the stationary state detecting unit and detecting the stationary state by the stationary state detecting unit Motion detection means for detecting the motion state of the subject and storing parameters relating to the motion state; parameters relating to the previous motion state stored by the motion detection means; Of corresponding before and after motion state, the front stationary state detecting means and stored by s times and on the basis of the parameters relating to this stationary state, a state transition detection means for detecting a state transition of the subject, the State transition detection means, when the stationary state detected by the stationary state detection means before and after the movement detection means detects the movement state of the subject is an upright state including the sitting state and the standing state of the subject, The first axis at the end of the state transition from the inclination angle formed by the direction of the first axis and the direction of gravity at the start of the state transition from the upright state detected after detecting the motion state of the subject The displacement to the tilt angle formed by the direction of gravity and the direction of gravity is calculated. If the displacement is less than the first value, the transition to the standing state is estimated, and the second value greater than the first value is calculated. If over, go to the seated state Posture estimation apparatus is provided for estimating the transition.

  According to the disclosed posture estimation apparatus, method, and program, it is possible to accurately estimate the posture of the subject, particularly the standing state and the sitting state.

It is a block diagram which shows an example of the attitude | position estimation system in one Example. It is a figure explaining an example of mounting | wearing with the subject of a posture estimation apparatus. It is a block diagram which shows an example of the attitude | position estimation apparatus in one Example. It is a flowchart explaining an example of operation | movement of an attitude | position estimation apparatus. It is a flowchart explaining an example of a separation process. It is a flowchart explaining an example of mounting | wearing and a motion detection process. It is a flowchart explaining an example of a stationary state detection process. It is a figure explaining the relationship between the inclination angle of an attitude | position estimation apparatus, and a coordinate axis. It is a flowchart explaining an example of seating and standing-up determination processing. It is a flowchart explaining an example of a fall detection process. It is a flowchart explaining an example of a walk detection process. It is a flowchart explaining an example of a periodicity determination process. It is a flowchart explaining an example of a state transition detection process. It is a flowchart explaining an example of a state transition determination process. It is a flowchart explaining an example of a state transition determination process. It is a figure which shows an example of an experimental result. It is a figure which shows an example of an experimental result. It is a figure which shows an example of an experimental result. It is a figure which shows an example of an experimental result. It is a figure which shows an example of an experimental result. It is a figure which shows an example of an experimental result. It is a figure which shows an example of an experimental result. It is a figure which shows an example of an experimental result.

  In the disclosed posture estimation apparatus, method, and program, the directions of the first and second axes of the three-axis acceleration sensor that detects the acceleration of the upper body of the subject coincide with the longitudinal direction and the vertical direction of the subject in the standing state, respectively. Thus, the posture of the subject who is worn is estimated. The posture estimation device includes a stationary state detection unit that detects a stationary state of the subject based on a DC signal component included in the output of the triaxial acceleration sensor, and a target based on the AC signal component included in the output of the triaxial acceleration sensor. A motion detection unit that detects a person's motion state, a parameter related to the previous motion state stored by the motion detection unit, and a previous and current stationary state stored by the stationary state detection unit corresponding to before and after the previous motion state And a state transition detection unit that detects a state transition of the subject based on the parameters. The state transition detection unit detects the end of the motion state detected by the motion detection unit when the stationary state detected by the stationary state detection unit before and after the motion detection unit detects the motion state is an upright state including a sitting state and a standing state. Calculating the displacement of the inclination angle formed by the direction of the first axis and the direction of gravity within a certain period until the time, and estimating the transition to the standing state if the displacement is less than the first value; If the second value, which is larger than the second value, is exceeded, the transition to the seating state is estimated.

  Embodiments of the disclosed posture estimation apparatus, method, and program will be described below with reference to the drawings.

  FIG. 1 is a block diagram illustrating an example of a posture estimation system in one embodiment. The posture estimation system shown in FIG. 1 includes a posture estimation device 1 and a server 11. The posture estimation apparatus 1 has a configuration in which a CPU (Central Processing Unit) 2, which is an example of a processor, a storage unit 3, a sensor unit 4, and a communication unit 5 are connected via a bus 6. The CPU 2 controls the operation of the posture estimation apparatus 1 as a whole. The storage unit 3 stores programs executed by the CPU 2 and various data. Various types of data include intermediate results of computations executed by the CPU 2, parameters (or data) used in the computations, parameters such as an inclination angle φ calculated during posture estimation processing described later, and the stationary state or motion state of the subject. Related parameters. The storage unit 3 can form an example of a computer-readable storage medium that stores a program. The sensor unit 4 includes a triaxial acceleration sensor, and may include an analog-to-digital converter (ADC) that converts an analog output of the triaxial acceleration sensor into a digital output. The triaxial acceleration sensor detects, for example, acceleration along the x-axis, y-axis, and z-axis of the xyz coordinate system, and corresponding x-axis signals, y-axis signals, and z-axis signals (that is, x, y, z axes). A known acceleration signal). In this example, the front-rear direction of the posture estimation device 1 including the three-axis acceleration sensor is the x-axis direction, the left-right direction (or horizontal direction) of the designated estimation device 1 is the y-axis direction, and the vertical direction of the posture estimation device 1 is the z-axis. Defined as direction. The communication unit 5 has a known configuration for performing wireless communication between the posture estimation device 1 and the external server 11. The server 11 can be formed by a general-purpose computer having a well-known configuration that performs wireless communication with the posture estimation device 1. For example, the server 11 may have the same configuration as the posture estimation device 1 in which the sensor unit 4 is omitted.

  In the example shown in FIG. 1, the CPU 2, the storage unit 3, the sensor unit 4, and the communication unit 5 are connected via the bus 6, but the connection between the CPU 2 and each unit in the posture estimation device 1 is via the bus 6. It is not limited to connection. Further, the posture estimation device 1 may be formed by, for example, a smart phone. When the posture estimation device 1 is formed of a smartphone, the smartphone can realize each unit (or each function) of the posture estimation device 1 by executing application software installed in the smartphone.

  FIG. 2 is a diagram for explaining an example of wearing of the posture estimation apparatus to a subject. The posture estimation apparatus 1 is mounted at a position where the acceleration of the upper body of a user (hereinafter referred to as “subject”) 21 that is a posture estimation target can be detected by the mounting unit 1A. The upper body of the subject 21 refers to a part of the body above the part including the waist of the subject 21, for example. In the example shown in FIG. 2, the posture estimation device 1 is placed on the abdomen of the subject 21 in the front-rear direction, the left-right direction, and the up-down direction of the subject 21 in which the front-rear direction, the left-right direction, and the up-down direction of the posture estimation device 1 are standing. (Or the direction of gravity). Therefore, the front-rear direction, left-right direction (or lateral direction), and up-down direction (or gravity direction) of the subject 21 in the standing state wearing the posture estimation device 1 are the x-axis direction, the y-axis direction, and the z-axis, respectively. Corresponds to the direction. Here, when the subject 21 walks in the forward direction of his / her own, it corresponds to progress in a direction that protrudes perpendicularly to the paper surface in the example of FIG. 2. Therefore, in this example, the front-rear direction of the posture estimation device 1 corresponds to a direction perpendicular to the paper surface in FIG. The mounting unit 1A has a function of holding the posture estimation device 1 and a function of mounting the posture estimation device 1 to the subject 21. The mounting portion 1A may be formed of, for example, a belt that can be mounted on the subject 21 or a connection member that can be mounted on a belt that is already used by the subject 21.

  The position where the posture estimation apparatus 1 is mounted is not particularly limited as long as the acceleration of the upper body of the subject 21 can be detected, and may be, for example, the chest or the back.

  FIG. 3 is a block diagram illustrating an example of a posture estimation apparatus according to an embodiment. The appearance state extraction force estimation apparatus 1 shown in FIG. 3 includes a separation unit 21 that is an example of a separation unit, a stationary unit 22 that is an example of a stationary state detection unit, a motion detection unit 23 that is an example of a motion detection unit, and state transition detection. A state transition detection unit 24 which is an example of the means is included.

  The separation unit 21 acquires the triaxial acceleration sensor output from the sensor unit 4 for every predetermined period including a predetermined number of samples, and corresponds to a motion component corresponding to an alternating current (AC) signal component and a direct current (DC) signal component. Separated into gravity components. The separated motion component is input to the motion detection unit 23, and the separated gravity component is input to the stationary state detection unit 22. The motion detection unit 23 detects the motion state of the subject 21 based on the motion component from the separation unit 21 and stores parameters regarding the motion state. The stationary state detection unit 22 detects the stationary state of the target person 21 based on the gravity component from the separation unit 21 and the parameter regarding the previous movement state stored by the movement detection unit 23, and stores the parameter regarding the stationary state. To do. The state transition detection unit 24 is based on the parameter related to the previous stationary state and the parameter related to the current stationary state stored by the stationary state detection unit 22 and the parameter related to the previous motion state stored in the motion detection unit 23. Is detected.

  If the stationary state detector 22 detects the upright state of the subject 21 before and after the motion detector 23 detects the motion state of the subject 21, the state transition detector 24 detects the subject 21 detected by the motion detector 23. The displacement of the tilt angle of the posture estimation device 1 within a certain period until the end of the motion state is calculated, and if the displacement is less than a first value (for example, −20 °) and less, the transition to the standing state is estimated. If it exceeds a second value (for example, 20 °) that is larger than the first value and more, a transition to the seating state is estimated. The inclination angle of the posture estimation device 1 is an angle formed by the front-rear direction of the posture estimation device 1 and the gravity direction. The upright state of the subject 21 means a stationary state in which the axis of the upper half of the subject 21 (corresponding to the z-axis) is parallel to or substantially parallel to the direction of gravity, and includes a sitting state and a standing state.

  For example, when the subject 21 sits on a chair, the subject 21 moves in the front-rear direction because the subject 21 is seated on the back of the chair within a certain period after the movement in the gravitational direction in which the subject 21 sits. There is. That is, there is a certain amount of backward movement after the subject 21 is seated. On the other hand, when the subject 21 sitting on the chair stands up, the subject 21 moves in the front-rear direction within a certain period after the movement in the direction opposite to the direction of gravity to stand, particularly, The backward movement is less than when sitting. Therefore, in the state transition detection unit 24 according to the movement of the subject 21 in the front-rear direction within a certain period after the upright state is detected before and after detecting the exercise state of the subject 21, It is possible to accurately determine whether it is a seating-standing transition (that is, a transition from a sitting state to a standing state) or a standing-sitting transition (that is, a transition from a standing state to a sitting state). Specifically, the movement in the front-rear direction of the subject 21 within a certain period after the upright state is detected before and after the motion state of the subject 21 is detected is less than the first value. If less, the state transition detection unit 24 estimates the transition to the standing state, and if it exceeds the second value larger than the first value, the state transition detection unit 24 estimates the transition to the seating state. Can do.

  FIG. 4 is a flowchart for explaining an example of the operation of the posture estimation apparatus. The process shown in FIG. 4 can be executed by the CPU 2, for example. The process shown in FIG. 4 includes a separation process ST1, a mounting and movement detection process ST2, a stationary state detection process ST3, a fall detection process ST4, a walk detection process ST5, a state transition detection process ST6, and another movement parameter extraction process ST7.

  The separation process ST1 can be executed by the separation means (or separation function) realized by the CPU 2 executing the program, and may be executed by the separation unit 21 in FIG. In step S1-1, the CPU 2 acquires the triaxial acceleration sensor output from the sensor unit 4 for each window of a certain period including, for example, a predetermined number of samples. In step S1-2, the CPU 2 extracts a motion component from the alternating current (AC) signal component of the sensor output. In step S1-3, the CPU 2 extracts a gravity component from the direct current (DC) signal component of the sensor output. As a result, the sensor output is separated into an AC signal component and a DC signal component for each window of a certain period.

  The mounting and motion detection process ST2 can be executed by mounting and motion detection means (or a mounting and motion detection function) realized by the CPU 2 executing a program. The motion detection processing is performed by the motion detection unit 23 of FIG. May be executed. In step S2-1, the CPU 2 determines whether or not the posture estimation apparatus 1 is worn on the subject 21 based on the AC signal component extracted in step S1-2. If the determination result is No, the process is performed. Returns to step S1-2. When the determination result in step S2-1 is Yes, in step S2-2, the CPU 2 determines whether the subject 21 is in an exercise state based on the AC signal component, and the process is performed when the determination result is No. The process proceeds to the stationary state detection process ST3, and if the determination result is Yes, the CPU 2 detects the movement state, and the process proceeds to step S2-3. In step S2-3, the CPU 2 extracts parameters related to the exercise state detected in step S2-2 and stores them in the storage unit 3, and the process proceeds to the fall detection process ST4. The exercise state includes states such as the subject 21 falling, walking, and state transition. Note that step S2-1 can be omitted.

  The time when the determination result in step S2-2 is No corresponds to the time when the start of the current stationary state is detected, that is, the time when the end of the current motion state is detected. The time when the determination result in step S2-2 is Yes before the current stationary state is detected corresponds to the time when the start of the previous motion state is detected. In addition, the time when the determination result of step S2-2 is No before the previous motion state is detected corresponds to the time when the start of the previous stationary state is detected. That is, the period from the end of the previous stationary state to the start of the current stationary state corresponds to the period from the start to the end of the previous motion state.

  The stationary state detection process ST3 can be executed by a stationary state detection means (or a stationary state detection function) realized by the CPU 2 executing a program, and may be executed by the stationary state detection unit 22 of FIG. In step S3-1, the CPU 2 detects the stationary state of the subject 21 based on the determination result of No in step S2-2 and the DC signal component extracted in step S1-3. The stationary state includes a seating state and a standing state of the subject 21. In step S3-2, the CPU 2 extracts the parameter related to the stationary state detected in step S3-1 and stores it in the storage unit 3, and the process proceeds to a state transition detection process ST6 described later.

  The fall detection process ST4 can be executed by a fall detection means (or a fall detection function) realized by the CPU 2 executing a program. In step S4-1, the CPU 2 determines whether or not the subject 21 is in a fallen state based on the AC signal component extracted in step S1-2. If the determination result in step S4-1 is Yes, in step S4-2, the CPU 2 extracts a parameter related to the overturning state detected in step S4-1, stores the parameter in the storage unit 3, and outputs a warning. On the other hand, if the decision result in the step S4-1 is No, the process advances to a walking detection process ST5.

  The walk detection process ST5 can be executed by a walk detection means (or a walk detection function) realized by the CPU 2 executing a program. In step S5-1, CPU2 determines whether the subject 21 is a walking state based on the AC signal component extracted by step S1-2. If the determination result in step S5-1 is Yes, in step S5-2, the CPU 2 extracts the parameter related to the walking state detected in step S5-1 and stores the parameter in the storage unit 3. On the other hand, if the decision result in the step S5-1 is No, the CPU 2 ends the process.

  The state transition detection process ST6 can be executed by state transition detection means (or a state transition detection function) realized by the CPU 2 executing a program, and may be executed by the state transition detection unit 24 of FIG. In step S6-0a, it is determined whether or not there is a sensor output to be processed by the CPU 2 (for example, a sensor output segment to be processed next as described later), and the process is separated if the determination result is Yes. Returning to step S1-1 of process ST1, the next sensor output (for example, the next segment of the sensor output) is acquired. If the decision result in the step S6-0a is No, in a step S6-0b, whether or not the CPU 2 has the previous exercise state, that is, whether or not the parameter relating to the previous exercise state is stored in the storage unit 3 If the determination result is No, the CPU 2 ends the process. On the other hand, if the decision result in the step S6-0b is Yes, the process advances to a step S6-1. In step S6-1, the CPU 2 extracts the parameters related to the previous motion state extracted by the wearing and motion detection processing ST2 (or the fall detection processing ST4 or the walking detection processing ST5) and stored in the storage unit 3, and the stationary state detection. It is determined whether or not it is a state transition of the subject 21 based on the parameters related to the previous and current stationary states extracted by the process ST3 and stored in the storage unit 3. If the determination result in step S6-1 is Yes, in step S6-2, the CPU 2 extracts parameters related to the state transition detected in step S6-1 and stores them in the storage unit 3, and the process is a stationary state detection process. Return to step S3-2 of ST3. On the other hand, if the decision result in the step S6-1 is No, the process advances to another exercise parameter extraction process ST7.

  In other exercise parameter extraction processing ST7, the CPU 2 extracts parameters related to other exercises of the subject 21 and stores them in the storage unit 3.

  After step S4-2 of the fall detection process ST4, step S5-2 of the walk detection process ST5, or other motion parameter extraction process ST7, in step S8, the sensor output to be processed by the CPU 2 (for example, the process to be processed next) It is determined whether there is a sensor output segment). If the decision result in the step S8 is Yes, the process returns to the step S1-1 of the separation process ST1 to acquire a sensor output to be processed next (for example, a sensor output segment to be processed next). On the other hand, if the decision result in the step S8 is No, the CPU 2 ends the process.

  Note that when the posture estimation process of the posture estimation apparatus 1 shown in FIG. 3 is executed, the step S2-1, the fall detection process ST4, and the walk detection process ST5 of the wearing and movement detection process ST2 shown in FIG. 4 can be omitted.

  FIG. 5 is a flowchart for explaining an example of the separation process ST1. In FIG. 5, when CPU2 acquires the 3-axis acceleration sensor output from the sensor part 4 by step S1-1, CPU2 will perform the process of step S1-2 including steps S1-2a-S1-2d, and step S1-3a-. The process of step S1-3 including S1-3d is executed in parallel. In step S1-2a, the CPU 2 performs high-pass filter processing with a center frequency fc = 0.25 Hz, for example, on the sensor output. In step S1-2b, the CPU 2 performs, for example, a median filter process with a window size n = 3 on the sensor output subjected to the high-pass filter process to remove spike noise. In step S1-2c, the CPU 2 divides the sensor output subjected to the median filter processing into AC signal component segments having a window size (for example, a fixed period of 1 second) of the number of samples s. In step S1-2d, CPU2 memorize | stores the AC signal component by which CPU2 was segmented in the memory | storage part 3 with the time information which CPU2 manages as an exercise | movement component showing the exercise | movement of the subject 21. FIG. The time information managed by the CPU 2 includes, for example, date / time information including the time, and a time measured by the internal timer of the CPU 2 while the CPU 2 is executing processing.

  On the other hand, in step S1-3a, the CPU 2 performs a low-pass filter process with a center frequency fc = 0.25 Hz, for example, on the sensor output. In step S1-3b, the CPU 2 performs, for example, a median filter process with a window size n = 3 on the sensor output subjected to the low-pass filter process to remove spike noise. In step S1-3c, the CPU 2 divides the sensor output subjected to the median filter processing into segments of DC signal components having a window size (for example, a fixed period of 1 second) of the number of samples s. In step S <b> 1-3 d, the CPU 2 stores the segmented DC signal component as a gravity component applied to the subject 21 in the storage unit 3 together with time information managed by the CPU 2.

FIG. 6 is a flowchart for explaining an example of the wearing and motion detection process ST2. In FIG. 6, in step S2-1a, the CPU 2 uses the xyz coordinate system to calculate the amplitude vector ρ _i.

Is represented by the root mean square (RMS) ρ _rms of the amplitude vector ρ _i based on the AC signal component.

And is stored in the storage unit 3 together with time information managed by the CPU 2. In step S2-1b, the CPU 2 determines whether or not ρ _i > ρ _static . ρ _static represents an amplitude vector when the posture estimation device 1 is not worn by the subject 21 and is placed on a table, for example. If the decision result in the step S2-1b is No, in the step S2-1c, the CPU 2 detects that the posture estimation device 1 is not worn on the subject 21, and the data regarding the non-wearing state is obtained. The information is stored in the storage unit 3 together with time information managed by the CPU 2. On the other hand, if the determination result in step S2-1b is Yes, in step S2-2a, the CPU 2 represents a signal amplitude area (SMA: Signal Magnitude) in which the x, y, and z-axis signals are normalized by the following equation. Area) is calculated for a certain window w, and is stored in the storage unit 3 together with time information managed by the CPU 2.

  In step S2-2b, if the SMA threshold is represented by Th1, the CPU 2 determines whether or not SMA> Th1. If the determination result in step S2-2b is Yes, in step S2-2c, the CPU 2 detects that the subject 21 is in an exercise state, and stores parameters related to the exercise state in the storage unit 3 together with time information managed by the CPU 2. Remember. On the other hand, if the determination result of step S2-2b is No, in step S2-2d, the CPU 2 determines that the exercise of the subject 21 has been completed and has entered a stationary state, and the process proceeds to a stationary state detection process ST3. Needless to say, when step S2-1 in FIG. 4 is omitted, steps S2-1a and S2-1b in FIG. 6 can be omitted.

  FIG. 7 is a flowchart illustrating an example of the stationary state detection process ST3. In FIG. 7, in step S <b> 3-1 a, the CPU 2 calculates the inclination angle φ represented by the following equation based on the determination result of No in step S <b> 2-2 and the DC signal component, and stores it along with time information managed by the CPU 2. Store in part 3. As described above, the tilt angle φ is an angle formed by the front-rear direction of the posture estimation device 1 and the gravity direction. In this example, using the xyz coordinate system, the direction parallel to the upper body axis of the subject 21 is the z axis, the front-rear direction of the subject 21 viewed from the upper body of the subject 21 is the x axis, and from the upper body of the subject 21 The left-right direction (or horizontal direction) of the viewed subject 21 is the y-axis. For example, when the subject 21 is standing, the gravitational direction is parallel to the z-axis, so the tilt angle φ is 90 °. When the subject 21 is completely lying on the bed, for example, the gravitational direction Is parallel to the x-axis, the inclination angle φ = 0 °.

Here, the gravitational acceleration component φ _i is expressed by the following equation.

  FIG. 8 is a diagram for explaining the relationship between the inclination angle φ and the xyz coordinate axes of the posture estimation apparatus 1. In FIG. 8, (a) is, for example, φ = 90 ° and the subject 21 wearing the posture estimation device 1 is in a standing state, and (b) is, for example, 60 ° <φ <90 ° and the subject 21 is In the case of lying, (c) is, for example, φ ≦ 60 ° and the subject 21 is seated, and (d) is, for example, φ = 0 °, and the subject 21 is in a complete lying state. Show. Thus, the standing state, the sitting state, and the lying state of the subject 21 can be detected according to the inclination angle φ.

  In step S3-1b, the CPU 2 determines whether or not the inclination angle φ is φ> 60 °, and if the determination result is Yes, in step S3-2a, the CPU 2 determines that the subject 21 in FIG. ) And the parameters related to the lying state are stored in the storage unit 3 together with the time information managed by the CPU 2. In step S3-1b, the CPU 2 determines whether or not the tilt angle φ is φ = 0 °. If the determination result is Yes, in step S3-2a, the CPU 2 determines that the subject 21 is in FIG. It is also possible to detect the complete recumbent state shown in (d) and store the parameters regarding the complete recumbent state in the storage unit 3 together with the time information managed by the CPU 2. Further, in addition to step S3-1b, the CPU 2 may provide a step of determining whether or not the tilt angle φ is φ = 0 °.

  On the other hand, if the determination result in step S3-1b is No, in step S3-1c, the CPU 2 maintains the angular displacement T with respect to the reference inclination angle φr when the subject 21 having the inclination angle φ is in the upright state, and maintaining a stationary state. Whether or not the subject 21 is in the sitting state is determined based on the time d, the exercise state detected last time, and the like. If the determination result in step S3-1c is Yes, in step S3-2b, the CPU 2 detects that the subject 21 is in the sitting state shown in FIG. 8C, and the CPU 2 manages parameters relating to the sitting state. The time information is stored in the storage unit 3 together with the time information. On the other hand, if the decision result in the step S3-1c is No, in the step S3-2c, the CPU 2 detects that the subject 21 is in the standing state shown in (a) of FIG. Is stored in the storage unit 3 together with the time information managed by.

  FIG. 9 is a flowchart for explaining an example of the seating and standing-up determination process executed in steps S3-1c, S3-2b, and S3-2c surrounded by a broken line in FIG. In FIG. 9, in step S31, the CPU 2 determines whether or not the inclination angle φ is φ> 20 °. If the decision result in the step S31 is Yes, in a step S3-2b, the CPU 2 detects the seating state and stores parameters relating to the seating state in the storage unit 3 together with time information managed by the CPU 2.

  On the other hand, if the decision result in the step S31 is No, the process advances to steps S32, S34, S36. The processes of steps S32 and S33, steps S34 and S35, and steps S36 and S37 are executed by the CPU 2 in parallel.

In step S32, the CPU 2 calculates the angular displacement T (°) = φ−φr by subtracting the reference inclination angle φr from the current inclination angle φ, and stores it in the storage unit 3 together with the time information managed by the CPU2. The reference inclination angle φr is an inclination angle φ when the subject 21 is in an upright state, and can be measured for each subject 21 in advance, for example. In step S33, based on the angular displacement T, the CPU 2 calculates the probability P ₁ (Sit) that the subject 21 is in the seated state and the probability P ₁ (Stand) that the subject 21 is in the standing state based on the following equations. And it memorize | stores in the memory | storage part 3 with the time information which CPU2 manages. The probability P ₁ (Sit) of being in a seated state is higher as the angular displacement T is larger.

In step S34, a period d (seconds) in which the state of the subject 21 is maintained, such as a duration in which the CPU 2 is maintained at the tilt angle φ and a time in which the subject 21 is in a resting state, It is calculated based on the time measured by the timer and stored in the storage unit 3 together with time information managed by the CPU 2. The resting state is a state in which the subject 21 does not perform a large movement, for example, the sensor output (DC signal component) from the sensor unit 4 is below a certain amplitude for each of the x, y, and z axes, or the SMA. State below the value. In step S35, based on the period d, the CPU ₂ calculates a probability P ₂ (Sit) that the subject 21 is seated and a probability P ₂ (Stand) that the subject 21 is standing based on the following equation. And stored in the storage unit 3. The probability P ₂ (Sit) of being in a seated state is higher as the period d is longer.

In step S <b> 36, the CPU 2 reads out parameters related to the previous exercise state M of the subject person 21 from, for example, the storage unit 3. In step S37, CPU 2 is based on parameters relating to the previous state of motion M, the probability P ₃ subject 21 on the basis of the following expression is sitting state _(Sit), the probability P ₃ subject 21 is standing state ( Stand). The probability P ₃ (Sit) of being in a seated state is 0.2 if the previous motion state M is a seated-standing state, 0.0 if it is a standing-sitting state, 0.3 if it is a walking state, and others. It is 0.5 in the state (otherwise).

In step S38, the target person 21 determines from the weighted average of the probabilities P ₁ (Sit), P ₂ (Sit), and P ₃ (Sit) that are the seating states obtained by the CPU 2 in steps S33, S35, and S37 based on the following equation. The probability P (SIT) of being in a seated state is calculated, and the weighted average of the probabilities P ₁ (Std), P ₂ (Std), and P ₃ (Std) of being in the standing state obtained in steps S33, S35, and S37 is Based on the equation, the probability P (STD) that the subject person 21 is standing is calculated. In the following equation, ω ₁ , ω ₂ , and ω ₃ are weighting coefficients that satisfy ω ₁ + ω ₂ + ω ₃ = 1.

  In step S39, the CPU 2 determines whether or not P (SIT)> P (STD). If the determination result is Yes, in step S3-2b, the CPU 2 detects the sitting state, and parameters related to the sitting state. Are stored in the storage unit 3 together with time information managed by the CPU 2. On the other hand, if the determination result in step S39 is No, in step S3-2c, the CPU 2 detects the standing state, and stores the parameters regarding the standing state in the storage unit 3 together with time information managed by the CPU 2.

As described above, when the tilt angle φ is larger than the first angle (for example, φ> 60 °), the CPU 2 detects the lying state of the subject 21, and when it is equal to or smaller than the first angle (for example, φ It is determined whether the subject 21 is seated or standing (if it is other than> 60 °). Then, the CPU 2 detects the seating state when the tilt angle φ is equal to or smaller than the first angle and larger than the second angle smaller than the first angle (for example, φ> 20 °), and the second state is detected. If the angle is equal to or smaller than (for example, φ> 20 °), the determination of whether the seating state or the standing state is continued. Further, the inclination angle phi is equal to or less than the second angle, CPU 2 is angular displacement T and _{(°) = φ-φ r} , duration inclination angle phi is sustained, resting condition of the subject 21 is continued Based on a period d (seconds) such as time and a parameter related to the previous exercise state M of the subject 21 stored in the storage unit 3, the probability P (STD) that the subject 21 is standing and the subject is seated The probability P (SIT) is calculated, and the seating state is detected when P (SIT)> P (STD), and the standing state is detected when other than P (SIT)> P (STD).

  As described above, if the change in the inclination angle φ is larger than a certain value, the period d during which the inclination angle φ is sustained is longer than the certain value, and the previous exercise state M is the sitting state, there is a possibility that the present state is the sitting state. Therefore, the CPU 2 estimates that the subject 21 is in a seated state.

  Note that the CPU 2 may estimate a plurality of types of sitting states or a plurality of types of standing states. For example, φ <5 ° in an upright seated state, φ> 20 ° in a forward tilted seat state, φ> 20 ° in a back tilted seated state, φ <5 ° in an upright standing state, and in a standing upright state By defining that 20 °> φ> 10 °, the CPU 2 estimates an upright seating state, a forward tilted seating state, a rearward tilted seating state, an upright standing state, and a forward tilted standing state. Also good.

FIG. 10 is a flowchart for explaining an example of the fall detection process ST4. In FIG. 10, in step S4-1a, the CPU 2 calculates a difference D between the current acceleration expressed by the following equation and the average acceleration in the past t _x seconds based on the AC signal component.

In step S4-1b, if the CPU 2 represents the threshold value of the difference D by Th2, whether the difference D _k (k = 1,..., M) for the next m seconds is D _k > Th2 for all k. It is determined whether or not, and if the determination result is No, the processing returns to step S4-1a. If the difference _Dk is greater than the threshold Th2 for at least the next m seconds and the determination result in step S4-1b is Yes, in step S4-1c, the CPU 2 determines whether φ> 60 °, and the determination result. If No, the subject 21 is determined to be in a lying position, and the process returns to step S4-1a. If the determination result in step S4-1c is Yes, in step S4-2a, the CPU 2 detects the fall state of the subject 21, and stores the parameters relating to the fall state in the storage unit 3 together with time information managed by the CPU 2. In step S4-2b, the CPU 2 calculates the SMA for the next 60 seconds, for example, and detects the exercise state in the same manner as in step S2-2b in FIG. In step S4-2c, the CPU 2 determines whether or not φ <60 °. If the determination result is No, in step S4-2d, the CPU 2 outputs a warning and the CPU 2 manages data related to the warning. It memorize | stores in the memory | storage part 3 with time information. On the other hand, if the determination result in step S4-2c is Yes, in step S4-2e, the CPU 2 detects the transition of the subject 21 from the falling state to the upright state, and the CPU 2 manages parameters related to the standing state. At the same time, it is stored in the storage unit 3.

As described above, when the difference _Dk is larger than the threshold Th2 for at least the next m seconds and the inclination angle φ is larger than the first angle (for example, φ> 60 °), the CPU 2 detects the fall state. To do. Further, for example, when the movement state is detected from the SMA of the next fixed time (for example, 60 seconds) after the CPU 2 detects the falling state, and the inclination angle φ is larger than the first angle (for example, φ> 60 °) If there is, a warning is output, and if the angle is smaller than the first angle (for example, φ <60 °), the transition of the subject 21 from the falling state to the standing state is detected.

  FIG. 11 is a flowchart for explaining an example of the walking detection process ST5. In FIG. 11, in step S5-1a, based on the AC signal component, the CPU 2 is in an upright state and determines whether or not the average of the inclination angle φ in a past fixed exercise period (for example, several seconds) is less than 35 °. judge. If the determination result in step S5-1a is Yes, in step S5-1b, the CPU 2 determines whether the exercise period is longer than, for example, 5 seconds. If the decision result in the step S5-1b is Yes, in a step S5-1c, the CPU 2 decides whether or not the exercise is a periodic exercise by a periodicity judgment process described later with reference to FIG. When the determination result of any of steps S5-1a, S5-1b, and S5-1c is No, in step S5-2b, the CPU 2 detects that the subject 21 is not in a walking state, and the walking state The parameters relating to the non-state are stored in the storage unit 3 together with the time information managed by the CPU 2. If the determination result in step S5-1c is Yes, in step S5-1d, the CPU 2 determines whether or not the SMA during the exercise period is greater than the average SMA during the exercise period other than the walking state. If the determination result in step S5-1d is Yes, in step S5-2a, the CPU 2 detects that the subject 21 is in a walking state, and stores parameters related to the walking state in the storage unit 3 together with time information managed by the CPU 2. Remember. On the other hand, if the determination result in step S5-1d is No, in step S5-2c, the CPU 2 has a sufficiently high possibility that the subject 21 is probably in a walking state, that is, the subject 21 is in a walking state. And the parameter relating to this state is stored in the storage unit 3 together with time information managed by the CPU 2.

FIG. 12 is a flowchart illustrating an example of the periodicity determination process in step S5-1c surrounded by a broken line in FIG. In FIG. 12, in step S51, the CPU 2 performs a Fourier transform on the z-axis signal in the AC signal component to detect a peak value, and randomly selects the peak value of the z-axis signal. In step S52, for example, a sample signal S _ref for 1 second is acquired with the peak value selected by the CPU 2 as the center. At step S53, sequentially applying a window of 1 second, for example 10 seconds period after the CPU2 sample signal _{S i.} In step S54, the CPU 2 calculates a correlation coefficient γ _i between the sample signal S _ref and the ten sample signals S _i after the sample signal S _ref . In step S55, the CPU 2 sets i to i = 1 and j to j = 1. In step S56, if the threshold value of the correlation coefficient γ _i is represented by Th3, the CPU 2 determines whether γ _i > Th3. The threshold value Th3 is, for example, 0.85.

  If the determination result in step S56 is No, in step S57, the CPU 2 determines whether i> 10. If the determination result is No, in step S58, the CPU 2 increments i to i = i + 1. Then, the process returns to step S56. On the other hand, if the determination result in step S57 is Yes, in step S59, the CPU 2 determines that the exercise of the subject 21 is a non-periodic exercise, and the CPU 2 manages data relating to the non-periodic exercise. It memorize | stores in the memory | storage part 3 with time information.

If the decision result in step S56 is Yes, in step S60, the CPU 2 records the sample signal S _i in the storage unit 3 as a candidate for the walking template T _j . In step S61, the CPU 2 increments j to j = j + 1. In step S62, the CPU 2 determines whether j> 3. If the decision result in the step S62 is No, the process returns to the step S57. On the other hand, if the determination result in step S62 is Yes, in step S63, the CPU 2 determines that the exercise of the subject 21 is a periodic exercise, and the CPU 2 manages data relating to the periodic exercise. It memorize | stores in the memory | storage part 3 with time information. In this example, the above process is repeatedly executed until, for example, three candidates for the walking template _Tj are acquired.

  Thus, when the inclination angle φ is smaller than the third angle (for example, φ <35 °) and the exercise period is longer than a certain time (for example, 5 seconds), there is a high possibility of being in a walking state. . The third angle is smaller than the first angle and larger than the second angle. Therefore, if the exercise has periodicity and the SMA during the exercise period is larger than the average SMA during the exercise period other than the walking state, the CPU 2 detects the walking state. If the inclination angle φ is greater than or equal to the third angle, or if the exercise period is less than a certain time, or if the exercise has no periodicity, the CPU 2 determines that it is not in a walking state. Further, if the SMA during the exercise period is equal to or less than the average SMA during the exercise period other than the walking state, the CPU 2 determines that it is probably in the walking state. Note that whether or not the exercise has periodicity can be determined by whether or not a certain number (for example, three) of walking templates as described with reference to FIG. 12 is acquired. Judge that the movement is periodic.

  FIG. 13 is a flowchart for explaining an example of the state transition detection process ST6. In FIG. 13, in step S6-1a, the CPU 2 determines whether or not the previous stationary state is a recumbent state based on the parameter relating to the previous stationary state detected by the stationary state detection process ST3 and stored in the storage unit 3. judge. If the determination result in step S6-1a is Yes, in step S6-1b, the CPU 2 detects the current stationary state based on the parameters related to the stationary state detected by the stationary state detection process ST3 and stored in the storage unit 2. It is determined whether or not is in a recumbent state. Whether the previous or current stationary state is a recumbent state is, for example, whether or not the previous or current inclination angle φ is larger than the first angle (φ> 60 °) as in step S3-1b of FIG. Can be determined based on If the determination result in step S6-1b is Yes, in step S6-2a, the CPU 2 determines the recumbent-recumbent transition of the subject 21, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. Remember. If the determination result in step S6-1b is No, in step S6-2b, the CPU 2 determines whether the subject 21 is lying down and standing upright, and data related to the determination result is stored in the storage unit 3 together with time information managed by the CPU 2. Remember.

  On the other hand, if the determination result in step S6-1a is No, in step S6-1c, the CPU 2 detects the current stationary state based on the parameters related to the current stationary state detected by the stationary state detection process ST3 and stored in the storage unit 3. It is determined whether the stationary state is an upright state. Whether or not the current stationary state is an upright state is determined, for example, based on whether the current inclination angle φ satisfies φ = 90 ° (corresponding to a standing state) or φ <60 ° (corresponding to a sitting state). Is possible. If the determination result in step S6-1c is Yes, in step S6-1d, the CPU 2 determines whether the sitting-standing transition or the standing-sitting transition is based on, for example, the displacement between the previous tilt angle φ and the current tilt angle φ. It is determined whether or not. If the determination result in step S6-1d is Yes, in step S6-1e, the CPU 2 determines whether it is a seating-standing transition based on, for example, the previous and current displacements of the inclination angle φ. If the determination result in step S6-1e is Yes, in step S6-2c, the CPU 2 determines the sitting-standing transition, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. If the determination result in step S6-1e is No, in step S6-2d, the CPU 2 determines the standing-sitting transition, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. If the determination result in step S6-1d is No, in step S6-2e, the CPU 2 determines another exercise state, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. If the determination result in step S6-1c is No, in step S6-2f, the CPU 2 determines the upright-backward transition, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. .

14 and 15 are flowcharts for explaining an example of the state transition determination process executed in steps S6-1d, S6-1e, and S6-2c to S6-2e surrounded by a broken line in FIG. In FIG. 14, in step S601, the CPU 2 models the x-axis signal using, for example, a spline curve based on the AC signal component. In step S602, the CPU 2 calculates the number of peaks of the modeled x-axis signal. The number of peaks of the modeled x-axis signal represents the number of acceleration peaks in the front-rear direction of the subject 21 in the case of the subject 21 standing up. In step S603, the CPU 2 determines whether the number of peaks of the modeled x-axis signal is smaller than the threshold value. If the determination result is Yes, the process proceeds to step S604. If the determination result is No, the process is performed. The process proceeds to step S606. At step S604, the inclination at the start CPU2 is the previous state of motion (or, at the end of the previous stationary state) at the end of the previous movement status from the inclination angle phi ₁ (or, at the beginning of this stationary state) The displacement Δφ = φ ₂ −φ ₁ to the angle φ ₂ is calculated.

  In step S605, the CPU 2 defines the parameter t as in the following equation, sets x to x = 1, and the process proceeds to step 607 described later. In this example, the fact that the value of Δφ is a positive value indicates that the subject 21 has tilted backward after sitting or standing. Further, when t = 1, the subject 21 stands up, when t = 2, the sitting state, and when t = 0, the other state other than the upright state.

  If the determination result in step S603 is No, the CPU 2 determines that the number of peaks of the modeled x-axis signal is too large to be regarded as a transition between sitting and standing. In step S606, the CPU 2 , X are set to t = 0, x = 0, and the process proceeds to step S607 described later. By providing step S603, if the number of peaks of the modeled x-axis signal is large and equal to or greater than the threshold value, the movement of the subject 21 in the front-rear direction is greater than that in the seating operation or standing operation, and the CPU 2 is seated. It can be determined that the state transition is not between standing.

In step S607, the CPU 2 models the y-axis signal using, for example, a spline curve, and models the z-axis signal using, for example, a spline curve, based on the AC signal component. In step S608, the CPU 2 calculates the modeled y-axis signal, the number of peaks, and the number of peaks of the z-axis signal. In the case of the standing subject 21, the number of peaks of the modeled y-axis signal represents the number of acceleration peaks in the left-right direction (or the lateral direction) with respect to the front-rear direction of the subject 21. The number of peaks of the modeled z-axis signal represents the number of acceleration peaks in the direction of gravity and the direction opposite to the direction of gravity in the case of the standing subject 21. Note that steps S607 and 608 may be omitted, and the modeling of the y and z axis signals and the calculation of the number of peaks of the y and z axis signals may be further performed in steps S601 and S602.
In step S609, the CPU 2 determines whether or not the number of peaks of the z-axis signal is smaller than the threshold value. If the determination result is Yes, the process proceeds to step S610. In step S610, the CPU 2 detects the peak value, peak shape, valley maximum value, valley shape, and peak time of the modeled z-axis signal waveform. The maximum peak value (or peak) of the modeled z-axis signal waveform corresponds to, for example, the maximum acceleration in the gravitational direction of the standing subject 21, and the peak shape is, for example, the standing subject 21. This corresponds to a change in acceleration in the direction of gravity. Further, the maximum value (or peak) of the valley of the waveform of the modeled z-axis signal corresponds to, for example, the maximum acceleration in the direction opposite to the direction of gravity of the subject 21 in the standing state, and the shape of the valley is, for example, This corresponds to a change in acceleration in a direction opposite to the direction of gravity of the subject 21 in the standing state. At the time of the peak of the waveform of the z-axis signal, the standing subject 21 performs a motion in the gravitational direction (ie, a seating motion) and sits on a chair or the like to finish the motion in the gravitational direction (eg, SMA ≦ Th1 And the determination result in step S2-2b in FIG. 6 is No), which corresponds to the time when the peak (in this example, a mountain) on the z-axis direction is obtained. After step S610, the process proceeds to step S611 and step S614. On the other hand, if the determination result in step S609 is No, it is determined that the number of peaks of the z-axis signal modeled by the CPU 2 is too large to be regarded as a transition between sitting and standing. In step S612, the CPU 2 determines that z , Y are set to z = 0, y = 0, and the process proceeds to step S621 in FIG.

  In step S611, the CPU 2 defines z as follows:

  In step S611, z = 1 when it is a sitting-standing transition, and in this case, the waveform of the modeled z-axis signal has one peak and one valley, and when the subject 21 stands up, Because the acceleration of occurs first, the mountain precedes the valley. Further, z = 2 when it is a standing-sitting transition, and in this case, the waveform of the modeled z-axis signal has one valley and one mountain, and when the subject 21 is seated, the acceleration is downward. Occurs first, so the valley precedes the mountain. Further, in an otherwise state, z = 0. In this case, unlike the case of the sitting-standing transition and the standing-sitting transition, the waveform of the modeled z-axis signal includes, for example, valley-mountain- There are valleys or mountains-valleys-mountains.

In step S613, the CPU 2 determines whether or not the number of peaks of the y-axis signal is smaller than the threshold value. If the determination result is Yes, the process proceeds to step S614 described later. On the other hand, if the determination result in step S613 is No, it is determined that the number of peaks of the y-axis signal modeled by the CPU 2 is too large to be regarded as a transition between sitting and standing. In step S617, the CPU 2 determines that y Is set to y = 0, and the process proceeds to step S621 in FIG.
In step S614, the CPU 2 calculates a displacement Δφw representing the displacement Δφ of the inclination angle φ within a certain period until the end of the motion state of the subject 21. The certain period until the end of the exercise state of the subject 21 is, for example, a time corresponding to the last segment (or window w) of the sensor output processed by the CPU 2 until the end of the exercise state of the subject 21 (for example, 1 second). At the end of the exercise state of the subject 21, it can be understood from the determination result of Step 2-2 in FIG. Further, it can be understood that the last segment of the sensor output processed by the CPU 2 is No in the determination result in step S6-0a in FIG. In step S615, the movement of the CPU 2 is finished within a certain threshold time from the peak time of the waveform of the z-axis signal detected in step S610 when the displacement Δφw is greater than 20 °, for example (step S2- in FIG. 4). Whether the determination result of 2 is No) is determined. If the decision result in the step S615 is No, the process advances to a step S621 in FIG. If the decision result in the step S615 is Yes, in a step S616, the CPU 2 defines y as y = 2 based on the following expression, and the process advances to a step S621 in FIG. The threshold time may be, for example, a time corresponding to the window w for the predetermined period (for example, 1 second), but is not particularly limited.

  In step S615, the CPU 2 determines whether or not the exercise of the subject 21 is completed within a certain threshold time from the peak time of the waveform of the z-axis signal detected in step S610. It is possible to distinguish between the case where the part tilts backward and the case where the subject tilts backward at the end part of the standing motion. In other words, when the subject 21 tilts backward at the end of the seating operation, the backward tilt of the subject 21 stops within the threshold time, for example, by touching the back of the chair. On the other hand, when the subject tilts backward at the end of the standing motion, there is no particular element that stops the backward tilt of the subject 21 within the threshold time. Therefore, if the movement of the subject 21 is completed within a certain threshold time from the time of the peak of the waveform of the z-axis signal detected by the CPU 2 in step S610, the transition of the subject 21 from the standing state to the sitting state is If the exercise of the subject 21 is not completed within the threshold time, it can be determined that the subject 21 is not transitioning from the standing state to the sitting state.

  In FIG. 15, in step S621, the CPU 2 defines a transition transition, a standing state std, and a seating state std as follows.

Here, sgn (z) and sgn (t) are sign functions such as the following equations.

  In step S622, the CPU 2 determines whether or not transition> 0. If the determination result is Yes, in step S623, the CPU 2 determines whether or not sit> std. If the determination result in step S623 is Yes, in step S6-2d, the CPU 2 determines the standing-sitting transition, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. If the decision result in the step S623 is No, in a step S624, the CPU 2 decides whether or not sit <std. If the determination result in step S624 is Yes, in step S6-2c, the CPU 2 determines the seating-standing transition, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. If the determination result in step S624 is No, in step S6-2g, the CPU 2 determines that the intermediate state is in the middle of the transition, and stores data related to the determination result in the storage unit 3 together with time information managed by the CPU 2. Remember. On the other hand, if the determination result in step S622 is No, in step S6-2e, the CPU 2 determines another exercise state and stores data related to the determination result in the storage unit 3.

In this way, the CPU 2 determines the transition between the lying state, the standing state, and the sitting state from the previous and current stationary states. If both the previous and current stationary states are in the upright state, the CPU 2 determines whether the CPU 2 is in the sitting-standing transition state or in the standing-sitting transition state. For the x-axis signal, the displacement Δφ = φ ₂ −φ ₁ to the inclination angle φ ₁ at the start of the state transition and the inclination angle φ ₂ at the end of the state transition is smaller than the first value (for example, −20 °). When the CPU ₂ determines that the transition is a sitting-standing transition and the displacement Δφ = φ ₂ −φ ₁ is larger than a second value (for example, 20 °) larger than the first value, the CPU 2 is in a standing-sitting transition. Judge that there is. On the other hand, with respect to the z-axis signal, the CPU 2 is in the sitting-standing transition or in the standing-sitting transition based on the modeled peaks and valleys of the z-axis signal and the shapes of the peaks and valleys. Determine if there is. The determination result of the state transition for the x-axis signal and the determination result of the state transition for the z-axis signal are combined by the CPU 2 according to the rules shown in FIG. 15 to obtain a final determination result of the state transition.

  That is, if the stationary state detection process ST3 detects the upright state of the subject 21 including the sitting state and the standing state of the subject 21 before and after the motion detection process ST2 detects the motion state of the subject 21, the state transition detection process ST6. Calculates the displacement Δφw of the inclination angle φ within a certain period until the end of the movement of the subject 21, and if the displacement Δφw is less than a first value (for example, −20 °), the standing state of the subject 21 is calculated. The transition to is estimated, and if the second value (for example, 20 °) larger than the first value is exceeded, the transition of the subject 21 to the seating state is estimated.

For example, when the subject 21 sits on a chair, the subject 21 moves in the front-rear direction because the subject 21 is seated on the back of the chair within a certain period after the movement in the seating direction ends. is there. That is, there is a certain amount of backward movement after the subject 21 is seated. On the other hand, when the subject person 21 sitting on the chair stands up, the subject person 21 moves in the front-rear direction within a certain period after the movement in the standing direction ends, particularly in the backward direction. There is less movement compared to sitting. Therefore, after detecting the standing state before and after detecting the motion state of the subject person 21, the CPU 2 is seated in accordance with the motion of the subject person 21 in the front-rear direction within a certain period until the end of the motion of the subject person 21. -It is possible to accurately determine whether it is a standing transition or a standing-sitting transition. That is, after detecting the upright state before and after detecting the motion state of the subject 21, the motion of the subject 21 in the front-rear direction within a certain period until the end of the motion of the subject 21 is less than the first value. If not, the CPU 2 estimates the transition to the standing state, and if it exceeds the second value larger than the first value, the CPU 2 can estimate the transition to the seating state.
Further, when the subject 21 tilts backward at the end of the seating operation, the subject 21 is prevented from tilting backward within a threshold time, for example, by touching the back of the chair. On the other hand, when the subject tilts backward at the end of the standing motion, there is no particular element that stops the subject 21 from tilting backward within the threshold time. Therefore, if the movement of the subject 21 is completed within a certain threshold time from the time of the peak of the waveform of the z-axis signal detected by the CPU 2 in step S610, the transition of the subject 21 from the standing state to the sitting state is If the exercise of the subject 21 is not completed within the threshold time, it can be determined that the subject 21 is not transitioning from the standing state to the sitting state.
In the above example, the movement of the subject 21 in the x-axis direction is compared by comparing the inclination angle φ, but is not limited to this. In the above example, the threshold time from the end of the exercise along the z-axis direction of the subject 21 is a fixed value, but the threshold time can be appropriately set for each subject 21, for example.

  According to the above embodiment, since the CPU 2 stores the data related to the state or state transition detected in the above-described processes ST1 to ST7 in the storage unit 3 together with the time information, it relates to the stationary state and the exercise state of the subject 21 in daily life. The history can be stored in the storage unit 3 in a certain time unit such as a half day unit, a day unit, a week unit, or a month unit. Such a history may be transmitted to the server 11 periodically or at an arbitrary timing via the communication unit 5 to analyze the health state of the target person 21 in the server 11 based on the history. Further, when each estimated stationary state, each motion state, and each state transition is stored in the storage unit 3 together with time information, detailed history information is stored in the data storage area required to store the history in the storage unit 3 It can be reduced compared to the case of doing.

  In addition, since the CPU 2 stores parameters related to the wearing or movement state detected in the wearing and movement detection process ST2 in the storage unit 3 together with time information, information such as the time, the number of times, the period, etc. when the subject 21 wears the posture estimation device 1. And information, such as the time, the number of times, and the period when the subject 21 was in an exercise state, can be stored in the storage unit 3 in a certain time unit. From the history, the CPU 2 or the server 11 may obtain information such as a ratio of a period during which the subject 21 is in an exercise state to a certain period. When the CPU 2 obtains information such as the ratio of the period during which the subject 21 is in an exercise state to the certain period, the obtained information may be included in the history in the storage unit 3.

  Furthermore, since the CPU 2 stores the parameters related to the stationary state detected in the stationary state detection process ST3 in the storage unit 3 together with the time information, the time, number of times, the period, the inclination angle φ in the sitting state, etc. when the subject 21 is stationary. Can be stored in the storage unit 3 in a certain time unit. Therefore, the CPU 2 or the server 11 may obtain information such as an average, a maximum value, and a ratio with respect to a certain period of the period in which the target person 21 is in a stationary state such as a standing state, a sitting state, and a lying state from the history. When the CPU 2 obtains information such as the average, maximum value, ratio for a certain period of time during which the subject 21 is in a stationary state such as a standing state, a sitting state, and a lying state, the obtained information is stored in the storage unit 3. It may be included in the history.

  In addition, since the CPU 2 stores the parameters regarding the fall state detected in the fall detection process ST4 in the storage unit 3 together with the time information, the time, the number of times, the period, the number of times the warning was output, etc. Can be stored in the storage unit 3 in a certain time unit. Further, from the history, the CPU 2 or the server 11 may obtain information such as an average, a maximum value, a ratio with respect to a certain period of the period when the target person 21 was in a fall state. When the CPU 2 obtains information such as the average, maximum value, and ratio for a certain period during which the subject 21 has fallen, the obtained information may be included in the history in the storage unit 3. Furthermore, the warning level that the CPU 2 includes in the history may be determined in accordance with information such as the time, number of times, period, and number of times the warning has been output. For example, when the number of falls exceeds a threshold, when the period of the fall state exceeds the threshold, or when the number of times the warning is output exceeds the threshold, the CPU 2 determines the warning level to be one level higher. You may do it.

  Furthermore, in order to memorize | store the parameter regarding the walking state which CPU2 detected in walk detection process ST5 with the time information in the memory | storage part 3, information, such as the time, the frequency | count, and the period when the target person 21 was the walking state, is said unit of time. Can be stored in the storage unit 3. As long as the subject 21 is in the walking state, the CPU 2 may store each period in which the walking state is maintained, or may store the total of the walking state within a certain time. Further, from the history, the CPU 2 or the server 11 may obtain information such as an average, a maximum value, and a ratio with respect to a certain period during which the target person 21 is in a walking state. When the CPU 2 obtains information such as the average, maximum value, and ratio for a certain period during which the subject 21 is in a walking state, the obtained information may be included in the history in the storage unit 3. Note that information such as the number of steps and walking speed calculated by the CPU 2 by a known method based on the output of the triaxial acceleration sensor from the sensor unit 4 may be included in the history.

  In addition, since the CPU 2 stores the data regarding the state transition detected in the state transition detection process ST6 in the storage unit 3 together with the time information, the information such as the time, the number of times, the period, etc. when the target person 21 performs the state transition is as described above. It can be stored in the storage unit 3 in time units. Further, from the history, the CPU 2 or the server 11 may obtain information such as an average, a maximum value, and a ratio with respect to a certain period during which the target person 21 has made a state transition. When the CPU 2 obtains information such as the average, maximum value, and ratio for a certain period during which the target person 21 has made a state transition, the obtained information may be included in the history in the storage unit 3.

  Next, an example of detection results of the wearing and motion detection process ST2, the stationary state detection process ST3, and the state transition detection process ST6 will be described with reference to FIGS.

FIG. 16 is a diagram illustrating an example of experimental results of the mounting and motion detection process ST2 and the stationary state detection process ST3 in the above embodiment. The experimental results shown in FIG. 16 show that the RMS value, the standard deviation, and the minimum value in a stationary state with the posture estimation device 1 placed on the table, and the subject 21 wears the posture estimation device 1 and stands up. The RMS value, the standard deviation, and the minimum value are shown in the case of maintaining for 30 seconds, then maintaining the sitting state for 30 seconds, and then maintaining the recumbent state for 30 seconds. FIG. 16 shows RMS values, standard deviations, and minimum values in a stationary state, a standing state, a sitting state, and a lying state with respect to the x-axis, y-axis, z-axis, amplitude vector ρ _i, and SMA. In FIG. 16, the RMS value of the SMA indicated by being surrounded by a thick solid line E1 is not different from the values indicating the stationary state, the standing state, the sitting state, and the lying state, and is particularly stationary. It is difficult to identify the state, the sitting state and the lying state. On the other hand, in the RMS value of the amplitude vector ρ _i shown by being surrounded by the thick solid line E2 in FIG. 16, the values indicating the standing state indicated by the thick solid line ellipse indicate the stationary state, the sitting state, and the lying state. It was confirmed that the values indicating the static state, the standing state, the sitting state, and the lying state are different from each other so that they can be distinguished from each other, and similar results have been confirmed in a plurality of experiments. It was. Therefore, according to the said Example, it was confirmed that the stationary state of the subject 21, a standing state, a sitting state, and a lying state can be estimated with high precision.

  FIG. 17 is a diagram illustrating an example of an experimental result of the stationary state detection process ST3 in the embodiment. The experiment results shown in FIG. 17 show that the subject 21 wears the posture estimation device 1 and lies on the bed lying down, lying on the bed with his back facing down, lying on the bed in the right direction, lying on the bed in the left direction. A state where a state, a sitting state on a lounge chair, a sitting state on a chair, a sitting state tilted forward, and a standing state are each maintained for 1 minute and the same operation is repeated three times is shown. In FIG. 17, the vertical axis indicates the acceleration (g) in the z-axis direction, and the broken line indicates the boundary B where z = 0.5, that is, φ = 60 °. It is confirmed that the subject 21 is lying in the region where the acceleration in the z-axis direction is lower than the boundary B, and the subject 21 is seated in the region where the acceleration in the z-axis direction is higher than the boundary. Was also confirmed in several experiments. Therefore, according to the said Example, while being able to identify the subject's 21 sitting state and recumbent state, it was confirmed that a seating state and recumbent state can be estimated with high precision.

  FIG. 18 is a diagram illustrating an example of an experimental result of the stationary state detection process ST3 in the embodiment. The experimental result shown in FIG. 18 shows a case where the subject 21 maintains the sitting state and the standing state for 1 minute each and repeats the same operation 20 times. FIG. 18 shows seating, standing, and the correct answer rate of sitting and standing, the incorrect answer rate, and other probabilities that cannot be said to be correct or incorrect for the three cases. In the first case, the stationary state of the subject 21 is estimated using only the angular displacement Δφ and compared with the actual stationary state, and in the second case, the stationary state of the subject 21 using only the previous motion. Was estimated and compared with the actual stationary state. On the other hand, in the third case, as in the above embodiment, the stationary state of the subject 21 is estimated and compared with the actual stationary state based on the angular displacement Δφ, the period during which the stationary state is maintained, and the previous movement. . As a result of the comparison, according to the above-described embodiment, the correct rate of sitting, standing up, and the total of sitting and standing up, incorrect rate, and other probabilities that cannot be said to be correct or incorrect are compared to the first and second cases. It was confirmed to improve. Therefore, according to the said Example, while being able to identify the seating state and standing state of the subject 21, it was confirmed that the seating state and standing state can be estimated with high precision.

  FIG. 19 is a diagram illustrating an example of an experimental result of the stationary state detection process ST3 in the embodiment. The experimental results shown in FIG. 19 show that the subject 21 maintains the upright sitting state, the forward tilted seating state, the rearward tilted seating state, the upright standing state and the forward tilted standing state for 1 minute, respectively, Indicates the case of repeated times. Φ <5 ° in an upright seated state, φ> 20 ° in a forward tilted seat state, φ> 20 ° in a rearward tilted seat state, φ <5 ° in an upright standing state, and 20 in a forward tilted standing state. °> φ> 10 °. In FIG. 19, the correct rate of the upright seating state, the forward tilted seating state, the rearward tilted seating state, the upright standing state, and the forward tilted standing state is 1.00 (ie, the correct answer rate is 100%). Comparative Example 1 corresponding to a case where a stationary state is detected by a method similar to Patent Document 1, Comparative Example 2 corresponding to a case where a stationary state is detected by a method similar to Patent Document 2, and the above Example Example 1 corresponding to is shown. According to Example 1, all the correct answer rates of the subject 21 in the upright sitting state, the forward tilted seating state, the rearward tilted seating state, the upright standing state, and the forward tilted standing state are in Comparative Examples 1 and 2. It was confirmed that it was improved. Therefore, according to the above-described embodiment, the subject 21 can be identified from the upright sitting state, the forward tilted seating state, the rearward tilted seating state, the upright standing state, and the forward tilted standing state, It was confirmed that the seating state tilted forward, the seating state tilted backward, the standing upright state, and the standing state tilted forward can be estimated with high accuracy.

  FIG. 20 is a diagram illustrating an example of an experimental result of the state transition detection process ST6 in the embodiment. FIG. 20 shows a case where the subject 21 maintains the standing state for 5 seconds, maintains the sitting state for 5 seconds through the standing-sitting transition, and repeats the operation for maintaining the standing state for 5 seconds through the sitting-standing transition 30 times. Some of the experimental results are shown. In addition, the time of each transition of the standing-sitting transition and the sitting-sitting transition was changed in the range of 1 to 5 seconds, and the forward tilt was randomly inserted 10 times in total. In FIG. 20, the vertical axis represents the acceleration x (g) in the x-axis direction, the acceleration y (g) in the y-axis direction, and the acceleration z (g) in the z-axis direction, and the horizontal axis represents time (number of samples). In FIG. 20, Q1 shows a case where there is one valley and one mountain in the z-axis signal, and the valley is a standing-sitting transition preceding the mountain, and Q2 shows one mountain and one valley in the z-axis signal. There is a case where the mountain is a sitting-standing transition that precedes the valley. In addition, among the broken lines extending in the vertical direction shown in FIG. 20, two adjacent broken lines indicating the section including the peak and valley of the waveform are detected before and after the movement state, for example, this movement state is detected. This corresponds to the end point of the stationary state before the moving state and the starting point of the stationary state after the movement state is detected.

  FIG. 21 is a diagram illustrating an example of an experimental result of the state transition detection process ST6 in Example 1 corresponding to the above example. FIG. 21 shows the results obtained for the transition as shown in FIG. 20. Of the 30 standing-sitting transitions, 25 standing-sitting transitions are correctly estimated, and 27 out of 30 sitting-sitting transitions. The seating-standing transition was correctly estimated, and 6 forward tilts out of 10 forward tilts were erroneously estimated as the standing-sitting transition. In FIG. 21 and FIGS. 22 and 23 described later, the correct rate of state transition is shown as 1.00 with the maximum value (that is, the correct rate is 100%).

  FIG. 22 is a diagram showing an example of an experimental result obtained for the transition as shown in FIG. 20 in Comparative Example 1 corresponding to the case where the state transition is detected by a method similar to Patent Document 1. FIG. 23 is a diagram showing an example of an experimental result obtained for the transition as shown in FIG. 20 in the comparative example 2 corresponding to the case where the state transition is detected by a method similar to the above-mentioned Patent Document 2.

  As is clear from a comparison between FIG. 21, FIG. 22, and FIG. 23, according to Example 1, the correct rate of the standing-sitting transition and the sitting-standing transition of the subject 21 is higher than that of Comparative Examples 1 and 2. It was confirmed to improve. Therefore, according to the above-described embodiment, it was confirmed that the standing-sitting transition and the sitting-standing transition of the subject 21 can be identified, and the standing-sitting transition and the sitting-standing transition can be estimated with high accuracy.

  By the way, in the said Example, although CPU2 of the posture estimation apparatus 1 performs a posture estimation process, you may perform a posture estimation process by the distributed process by CPU2 and the server 11 of the posture estimation apparatus 1. FIG.

  When the CPU 2 of the posture estimation apparatus 1 executes all of the posture estimation processing or most of the posture estimation processing, the posture can be estimated in real time. Further, when each estimated stationary state, each motion state, and each state transition is stored in the storage unit 3 together with time information, detailed history information is stored in the data storage area required to store the history in the storage unit 3 It can be reduced compared to the case of doing.

  On the other hand, when the server 11 executes all of the posture estimation processing or most of the posture estimation processing, the load on the CPU 2 on the posture estimation device 1 side can be reduced.

  In the above embodiment, when the inclination angle φ is compared with the comparison object, a certain allowable range is provided for the comparison object, and for example, detection of coincidence within the allowable range is allowed as detection of coincidence with the comparison object. Needless to say, you can do it.

  According to the disclosed posture estimation apparatus, method, and program, the posture of the subject 21, particularly the standing state and the sitting state, and the transition state between the standing state and the sitting state can be estimated with high accuracy. In addition, the posture of the subject 21 can be estimated in real time with a relatively simple configuration using a single three-axis acceleration sensor that does not interfere with daily life when worn on the subject 21. .

The following additional notes are further disclosed with respect to the embodiment including the above examples.
(Appendix 1)
A three-axis acceleration sensor that detects the acceleration of the upper body of the subject so that the directions of the first axis and the second axis of the three-axis acceleration sensor coincide with the front-rear direction and the vertical direction of the subject in the standing state, respectively. A posture estimation device for estimating the posture of the subject person attached to the device,
A stationary state detecting means for detecting a stationary state of the subject based on a DC signal component included in an output of the three-axis acceleration sensor and storing a parameter relating to the stationary state;
Motion detection means for detecting a motion state of the subject based on an AC signal component included in an output of the three-axis acceleration sensor, and storing a parameter relating to the motion state;
Based on the parameters related to the previous motion state stored by the motion detection means and the parameters related to the previous and current stationary states stored by the stationary state detection means corresponding to before and after the previous motion state, the subject State transition detecting means for detecting the state transition of
In the state transition detection unit, the stationary state detected by the stationary state detection unit before and after the movement detection unit detects the movement state of the subject is an upright state including the sitting state and the standing state of the subject. , Calculating a displacement of an inclination angle formed by the direction of the first axis and the direction of gravity within a certain period until the end of the motion state of the subject detected by the motion detection means, and the displacement is the first The posture estimation device, wherein the posture estimation device estimates the transition to the standing state if the value is less than the value of, and estimates the transition to the sitting state if the value exceeds the second value greater than the first value.
(Appendix 2)
The state transition detection unit is configured to detect the subject within a threshold time from a time at which a peak of the acceleration signal along the direction of the second axis obtained by modeling the output of the triaxial acceleration sensor is obtained by the motion detection unit. The posture estimation apparatus according to appendix 1, wherein a transition from the standing state to the seating state is detected when the end of the movement of the body is detected.
(Appendix 3)
The posture estimation apparatus according to appendix 1 or 2, wherein the first value is -20 ° and the second value is 20 °.
(Appendix 4)
When the state transition detection unit determines that the previous stationary state is not a recumbent state and the current stationary state is an upright state, the state transition detecting unit determines whether the state is a sitting-standing transition or a standing-sitting transition. 4. The posture estimation apparatus according to any one of appendices 1 to 3, wherein determination is made and data relating to the determination result is stored together with time information.
(Appendix 5)
The state transition detection means has a peak number of acceleration signals along the direction of the first axis or the direction of the second axis that models the output of the three-axis acceleration sensor equal to or greater than a threshold value. The posture estimation apparatus according to any one of appendices 1 to 4, wherein if there is, it is determined that the state transition is not between sitting and standing.
(Appendix 6)
The stationary state detection means, if the tilt angle is not larger than the first value, an angular displacement represented by the difference between the tilt angle and a reference tilt angle when the subject is in the upright state, The probability P (STD) that the subject is in a standing state and the subject is in a sitting state based on a period in which the state of the subject including the tilt angle is sustained and data on the previous exercise state of the subject. The probability P (SIT) is calculated, the seating state is detected when P (SIT)> P (STD), and the standing state is detected when other than P (SIT)> P (STD). The posture estimation apparatus according to any one of appendices 1 to 5, which is a feature.
(Appendix 7)
A three-axis acceleration sensor that detects the acceleration of the upper body of the subject so that the directions of the first axis and the second axis of the three-axis acceleration sensor coincide with the front-rear direction and the vertical direction of the subject in the standing state, respectively. A posture estimation method for estimating the posture of the subject worn on the body,
A processor detects a stationary state of the subject based on a DC signal component included in the output of the three-axis acceleration sensor and stores a parameter related to the stationary state;
The processor detects a motion state of the subject based on an AC signal component included in an output of the three-axis acceleration sensor and stores a parameter related to the motion state;
The processor detects the state transition of the subject based on the stored parameters relating to the previous exercise state and the stored parameters relating to the previous and current stationary states corresponding to before and after the previous exercise state,
In the detection of the state transition, the stationary state detected by the stationary state detection before and after the movement detection detects the movement state of the subject is an upright state including the sitting state and the standing state of the subject. The displacement of the tilt angle formed by the direction of the first axis and the direction of gravity within a certain period until the end of the motion state of the subject detected by the motion detection is calculated, and the displacement is the first A posture estimation method characterized by estimating a transition to a standing state if it is less than the value of, and estimating a transition to a seating state if it exceeds a second value that is greater than the first value.
(Appendix 8)
The state transition is detected within a certain threshold time from the time at which the peak of the acceleration signal is obtained along the direction of the second axis in which the detection of the motion models the output of the three-axis acceleration sensor. 8. The posture estimation method according to appendix 7, wherein a transition from the standing state to the seating state is detected when the end of the movement is detected.
(Appendix 9)
9. The posture estimation method according to appendix 7 or 8, wherein the first value is −20 ° and the second value is 20 °.
(Appendix 10)
The state transition is detected when the previous stationary state is not a recumbent state and the current stationary state is an upright state, whether it is a sitting-standing transition or a standing-sitting transition. 10. The posture estimation method according to any one of appendices 7 to 9, wherein determination is made and data relating to the determination result is stored together with time information.
(Appendix 11)
The state transition is detected by detecting whether the number of acceleration signals along the direction of the first axis or the acceleration signal along the direction of the second axis that models the output of the three-axis acceleration sensor is greater than or equal to a threshold value. The posture estimation method according to any one of appendices 7 to 10, wherein if there is, it is determined that the state transition is not between sitting and standing.
(Appendix 12)
In the detection of the stationary state, if the tilt angle is not larger than the first value, an angular displacement represented by a difference between the tilt angle and a reference tilt angle when the subject is in the upright state, The probability P (STD) that the subject is in a standing state and the subject is in a sitting state based on a period in which the state of the subject including the tilt angle is sustained and data on the previous exercise state of the subject. The probability P (SIT) is calculated, the seating state is detected when P (SIT)> P (STD), and the standing state is detected when other than P (SIT)> P (STD). 12. The posture estimation method according to any one of appendices 7 to 11, which is a feature.
(Appendix 13)
A computer is provided with a three-axis acceleration sensor for detecting the acceleration of the upper body of the subject, the front-rear direction and the vertical direction of the subject in which the directions of the first axis and the second axis of the three-axis acceleration sensor are standing, respectively. A program for estimating the posture of the subject that is worn so as to match,
A stationary state detection procedure for detecting a stationary state of the subject based on a DC signal component included in the output of the three-axis acceleration sensor and storing a parameter related to the stationary state in a storage unit;
A motion detection procedure for detecting a motion state of the subject based on an AC signal component included in an output of the three-axis acceleration sensor and storing a parameter relating to the motion state in the storage unit;
Based on the parameters related to the previous motion state stored by the motion detection procedure and the parameters related to the previous and current still states stored by the stationary state detection procedure corresponding to before and after the previous motion state, the subject Causing the computer to execute a state transition detection procedure step for detecting a state transition of
In the state transition detection procedure, the stationary state detected by the stationary state detection procedure before and after the motion detection procedure detects the motion state of the subject is an upright state including the sitting state and the standing state of the subject. , Calculating a displacement of an inclination angle formed by the direction of the first axis and the direction of gravity within a certain period until the end of the motion state of the subject detected by the motion detection procedure, and the displacement is the first A program for estimating a transition to a standing state if it is less than a value of, and estimating a transition to a seating state if it exceeds a second value greater than the first value.
(Appendix 14)
In the state transition detection procedure, the subject is within a threshold time from a time at which a peak of an acceleration signal along the direction of the second axis obtained by modeling the output of the triaxial acceleration sensor is obtained by the motion detection procedure. The program according to appendix 13, wherein a transition from the standing state to the seating state is detected when the end of the movement of the body is detected.
(Appendix 15)
The program according to appendix 13 or 14, wherein the first value is -20 ° and the second value is 20 °.
(Appendix 16)
If it is determined that the previous stationary state is not a recumbent state and the current stationary state is an upright state, the state transition detection procedure determines whether it is a sitting-standing transition or a standing-sitting transition. The program according to any one of appendices 13 to 15, wherein the program is determined and data regarding the determination result is stored together with time information.
(Appendix 17)
In the state transition detection procedure, the peak number of acceleration signals along the direction of the first axis or the direction of the second axis obtained by modeling the output of the three-axis acceleration sensor is greater than or equal to a threshold value. The program according to any one of appendices 13 to 16, wherein if there is, it is determined that the state transition is not between sitting and standing.
(Appendix 18)
In the stationary state detection procedure, if the tilt angle is not larger than the first value, an angular displacement represented by a difference between the tilt angle and a reference tilt angle when the subject is in the upright state; The probability P (STD) that the subject is in a standing state and the subject is in a sitting state based on a period in which the state of the subject including the tilt angle is sustained and data on the previous exercise state of the subject. The probability P (SIT) is calculated, the seating state is detected when P (SIT)> P (STD), and the standing state is detected when other than P (SIT)> P (STD). 18. The program according to any one of appendices 13 to 17, which is a feature.

  As described above, the disclosed posture estimation apparatus, method, and program have been described by way of examples. However, the present invention is not limited to the above examples, and various modifications and improvements can be made within the scope of the present invention. Needless to say.

1 posture estimation device 1A mounting unit 2 CPU
3 storage unit 4 sensor unit 5 communication unit 6 bus 11 server 21 separation unit 22 stationary state detection unit 23 motion detection unit 24 state transition detection unit

Claims (5)

A three-axis acceleration sensor that detects the acceleration of the upper body of the subject so that the directions of the first axis and the second axis of the three-axis acceleration sensor coincide with the front-rear direction and the vertical direction of the subject in the standing state, respectively. A posture estimation device for estimating the posture of the subject person attached to the device,
A stationary state detecting means for detecting a stationary state of the subject based on a DC signal component included in an output of the three-axis acceleration sensor and storing a parameter relating to the stationary state;
Simultaneously with the detection of the stationary state of the subject by the stationary state detecting means , the motion state of the subject is detected based on an AC signal component included in the output of the three-axis acceleration sensor, and parameters relating to the motion state are determined. A motion detection means for storing;
And parameters relating to the motion state of the last time the motion detecting means is stored, corresponding to before and after the last motion state, the stationary state detecting means on the basis of the parameters relating to people gyrus and this quiescent before storing, the A state transition detecting means for detecting the state transition of the target person,
In the state transition detection unit, the stationary state detected by the stationary state detection unit before and after the movement detection unit detects the movement state of the subject is an upright state including the sitting state and the standing state of the subject. The first angle at the end of the state transition is determined from the inclination angle formed by the direction of the first axis and the direction of gravity at the start of the state transition from the upright state detected after detecting the motion state of the subject. A displacement to an inclination angle formed by the direction of the axis and the direction of gravity is calculated. If the displacement is less than a first value, a transition to a standing state is estimated, and a second value greater than the first value. A posture estimation device that estimates a transition to a seated state if the distance exceeds.   The state transition detection unit is configured to detect the subject within a threshold time from a time at which a peak of the acceleration signal along the direction of the second axis obtained by modeling the output of the triaxial acceleration sensor is obtained by the motion detection unit. The posture estimation device according to claim 1, wherein a transition from the standing state to the seating state is detected when the end of the movement state of the vehicle is detected.   The posture estimation apparatus according to claim 1, wherein the first value is −20 ° and the second value is 20 °. A three-axis acceleration sensor that detects the acceleration of the upper body of the subject so that the directions of the first axis and the second axis of the three-axis acceleration sensor coincide with the front-rear direction and the vertical direction of the subject in the standing state, respectively. A posture estimation method for estimating the posture of the subject worn on the body,
A stationary state detection process in which a processor detects a stationary state of the subject based on a DC signal component included in an output of the three-axis acceleration sensor and stores a parameter related to the stationary state;
The processor detects the motion state of the subject based on an AC signal component included in the output of the three-axis acceleration sensor simultaneously with the detection of the still state of the subject by the rest state detection processing, and the motion Motion detection processing for storing parameters relating to the state;
Wherein the processor, and a parameter relating to previous motion state stored, corresponding to before and after the last motion state, before storing s times and on the basis of the parameters relating to this stationary state, detecting a state transition of the subject State transition detection processing to
Run
In the state transition detection process, the stationary state detected by the stationary state detection process before and after the movement detection process detects the movement state of the subject is an upright state including the sitting state and the standing state of the subject. The first angle at the end of the state transition is determined from the inclination angle formed by the direction of the first axis and the direction of gravity at the start of the state transition from the upright state detected after detecting the motion state of the subject. A displacement to an inclination angle formed by the direction of the axis and the direction of gravity is calculated. If the displacement is less than a first value, a transition to a standing state is estimated, and a second value greater than the first value. A posture estimation method characterized by estimating a transition to a seated state if exceeding. A computer is provided with a three-axis acceleration sensor for detecting the acceleration of the upper body of the subject, the front-rear direction and the vertical direction of the subject in which the directions of the first axis and the second axis of the three-axis acceleration sensor are standing, respectively. A program for estimating the posture of the subject that is worn so as to match,
A stationary state detection procedure for detecting a stationary state of the subject based on a DC signal component included in the output of the three-axis acceleration sensor and storing a parameter related to the stationary state in a storage unit;
Simultaneously with the detection of the stationary state of the subject by the stationary state detection procedure , the motion state of the subject is detected based on an AC signal component included in the output of the three-axis acceleration sensor, and the parameters relating to the motion state are determined. A motion detection procedure stored in the storage unit;
And parameters relating to the motion state of the last time the motion detection procedure is stored, corresponding to before and after the last motion state, the stationary state detecting procedure based on the parameters related to stationary state before storing s times and time, the A state transition detection procedure for detecting a target person's state transition;
To the computer,
In the state transition detection procedure, the stationary state detected by the stationary state detection procedure before and after the motion detection procedure detects the motion state of the subject is an upright state including the sitting state and the standing state of the subject. The first angle at the end of the state transition is determined from the inclination angle formed by the direction of the first axis and the direction of gravity at the start of the state transition from the upright state detected after detecting the motion state of the subject. A displacement to an inclination angle formed by the direction of the axis and the direction of gravity is calculated. If the displacement is less than a first value, a transition to a standing state is estimated, and a second value greater than the first value. A program characterized by estimating a transition to a seated state if exceeding.