JP2003093566A - Method for discriminating movement and movement sensing module - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method for discriminating movements and gestures of a body by using only acceleration data detected by means of a triaxial acceleration sensor which an instrument mounted on the wrist possesses and to provide a movement sensing module miniaturized by this method and with less burden of mounting and with a low cost and a low consumption of electric power. SOLUTION: The method for discriminating the movement in which the acceleration data (low frequency components are eliminated) measured by means of the triaxial acceleration sensor mounted on the wrist part are integrated twice to obtain positional information and an average size and a direction of a positional vector within a specified time are calculated and kind of the movement performed is discriminated based on the calculated data and the movement sensing module in which the triaxial acceleration sensor and a discriminating circuit are built are provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method or apparatus for automatically identifying the type of exercise performed by a body and its intensity.

[0002]

2. Description of the Related Art In recent years, interest in health and medical treatment has been increasing, and the relationship between health and the motor function of the body has been strongly recognized. Therefore, for example, there are many proposals to detect the amount of exercise of the body with a handy device (with a built-in motion sensor and attached to a part of the body) so that the user does not feel burdened, and to utilize it for health management. On the other hand, there is also a proposal as a means of communication that does not use language, in which the gesture of the user is determined and the result is notified to others by communication or the like.

[0003]

In various body proposals, it is desirable to avoid wearing motion sensors at multiple points on the body at all times. When wearing it in one place, the waist part close to the center of gravity of the body may be selected, but it is considered that the body part performs a movement highly correlated with the movement of the whole body, and gestures can be detected. Because of this, the arm that does not feel uncomfortable is often used. However, it is not always easy to identify the motion performed by the human body or to estimate the amount of exercise based on the result of the motion sense of the arm, which is one specific part of the human body, and a simple and optimal method for that is conventionally proposed. Didn't.

Further, even if the movement of the arm portion is completed, it is necessary to measure six motion elements of acceleration in the three axis directions and angular velocity in the three axis directions in order to complete the movement. In particular, measurement of angular velocity requires a vibration gyroscope, but unlike an acceleration sensor, a gyro sensor has to excite a vibrating body for a sensor, which increases the number of measurement items as well as the structure of the sensor itself and the measurement circuit. Since it becomes complicated, it is unavoidable that the equipment will be large and expensive and power consumption will increase when trying to perform a complete measurement.

According to the present invention, a device mounted on an arm is provided with an acceleration sensor capable of measuring acceleration in three axial directions, and the acceleration data detected by this sensor can be used to identify body movements and gestures. It is an object of the present invention to provide a motion sensing module that can be downsized, thereby reducing the burden of mounting, reducing the cost, and reducing the power consumption.

[0006]

In order to achieve the above object, the motion identification method of the present invention has the following features. (1) Acceleration data measured by a triaxial acceleration sensor attached to a predetermined part of the body is integrated twice to obtain position information, and the average size and direction of the position vector within a predetermined time is calculated. Then, identify the type of exercise performed based on the calculated data.

The motion identification method of the present invention may further include any of the following features. (2) The acceleration data has low frequency components below a predetermined frequency cut. (3) The predetermined part of the body is the wrist.

To achieve the above object, the motion sensing module of the present invention has the following features. (4) A portable device that includes a three-axis acceleration sensor and a calculation device that adds a predetermined time calculation to acceleration data measured by the acceleration sensor, and is a portable device that is mounted on a predetermined part of the body, The arithmetic unit outputs acceleration data of 2
Integrating times to obtain position information, calculating the average size and direction of the position vector within the predetermined time, and identifying the type of exercise performed based on the calculated data.

The motion sensing module of the present invention may further include any of the following features. (5) The acceleration data has low frequency components below a predetermined frequency cut. (6) The predetermined part of the body is the wrist.

(7) The triaxial acceleration sensor has a structure in which a load mass whose center of gravity is separated from the plane of the plate is fixed, and the deformation of the plate due to acceleration acting on the load mass is piezoelectrically detected. .

[0011]

1 is a front view showing an example of an embodiment of a method and an apparatus according to the present invention and directions of coordinate axes. The motion sensing module 1 has a size similar to a wristwatch and a similar appearance, and is a triaxial acceleration sensor, a device for calculating acceleration data, a storage device for necessary data, a liquid crystal display device for calculation results, and a wireless communication device with external equipment. (In case of necessity), it includes an operating member, a power source, etc., and is attached to the back of the left wrist 3, for example, by the wristband 2 as in a wristwatch. X, Y, and Z are orthogonal coordinate axes fixed to the motion sensing module 1. That is, X shown in the drawing is the direction of the arm axis, Y is the width direction of the palm, and Z is the direction substantially perpendicular to the back of the hand with the finger extended.

Since there are many proposals that can be already adopted for the three-axis acceleration sensor, it is not necessary to specify the form, but basically, the three-axis acceleration sensor is made of a material having a piezoelectric property, and the inertial force generated by the acting acceleration causes An element having a bendable portion, exhibiting different bending and different piezoelectric action depending on the direction of the inertial force, and having an electrode for detecting the bending is small in size and consumes less power, which is excellent.

FIG. 5 shows a concrete example of a three-axis acceleration sensor using a metal disk suitable for use in the present invention.
(A) is a top view and (b) is an AA sectional view. The metal body is made of 7 mm x 7 mm x 2 mm phosphor bronze block material, and has a central shaft 22 with a diameter of 6 mm and a thickness of 0.2 m.
The metal disk 21 of m is carved out, and the surrounding frame-shaped part left unmachined is the support 20 integrated with the metal disk 21, and the lower surface thereof is fixed to the sensor container. Central axis 22
A ring-shaped load mass 23 is press-fitted into this. Therefore, the center of gravity of the load mass 23 is eccentric to a position spaced downward from the surface of the metal disk 21. Also, in the figure, the acceleration Gx in the directions of the XYZ coordinate axes and the respective axial directions acting on the center of gravity of the load mass 23
Gy and Gz are shown.

The piezoelectric plate 24 is 7 mm × 7 mm × 0.12 m
m is a thin plate and is made of a piezoelectric porcelain material such as PZT that is polarized in the plate thickness direction, and eight axially symmetrical fan-shaped electrode films (the X electrode films 25 and 26 having the same shape, respectively) are arranged on the upper surface side. Y electrode films 27, 28 and Z electrode films 29, 30, 31, 32) having an area about half of those are formed, the lower surface is adhered to the upper surface of the metal disk 21, and the metal disk 21 is It also serves as a common electrode (reference electrode) of the piezoelectric plate 24. The entire sensor is normally housed in an airtight container, and the electrode films 25 to 32 and the metal disk 21 are connected to the acceleration detection circuit via flexible lead wires and insulating terminals, respectively. The illustration of the circuit is omitted.

5 (c) and 5 (d) are schematic diagrams showing a state of deformation due to acceleration. The illustrated orientation is the same as the sectional view of FIG. When the acceleration Gx acts, an inertial force Fx is generated in the opposite direction at the center of gravity of the load mass 23 as shown in (c),
The load mass 23 is displaced in the direction of the force, and the metal disk 21 (particularly the central portion thereof) is deformed into a wave shape. In the piezoelectric plate 24 (not shown below) attached to the upper side of the metal disk 21, the piezoelectric material in the portion covered with the X electrode film 25 is a convex side, and therefore extends.
The piezoelectric material in the portion covered by the X electrode film 26 is concave, so that the piezoelectric material shrinks, and a voltage (with respect to the potential of the reference electrode) having a reverse polarity and proportional to the amount of bending is generated in each electrode film. Inertial force F
Since the deformation of the portion covered with the Y electrode films 27 and 28 due to x is equal to the unevenness, no voltage is generated. (Not shown,
Since the acceleration Gy also acts in the direction parallel to the plate surface, the same phenomenon occurs even though the direction is vertical, and a reverse voltage is generated in the Y electrode films 27 and 28. )

When the acceleration Gz acts, FIG. 5 (d)
An inertial force Fz is generated at the center of gravity of the load mass 23 as
The metal disk 21 is deformed upward into a concave shape. In this case, the piezoelectric plate 24 contracts equally at all parts, so that voltages of the same polarity and the same amount are generated in the Z electrode films 29 to 32.
Because of the above-described actions, the outputs of the X electrode films 25 and 26 and the Y electrode films 27 and 28 are connected to the input parts of the two differential amplifiers, respectively, and the Z electrode films 29 to 32 are collectively input to the amplifier. It is possible to know the magnitude of each direction component of the acceleration by connecting to the section and measuring the change in the output thereof. As described above, the acceleration sensor of this example is small and has a simple structure, and is capable of measuring triaxial acceleration.
Suitable for incorporation in wristwatch-type devices.

The acceleration sensor (including the detection circuit) used in the example of the embodiment of the present invention is manufactured by Microstone Co., Ltd., and its characteristics are as follows. Detection range: ± 40 m / squaresec, detection sensitivity:
+50 mV / m / squaresec, external dimensions: W2
0 mm x D12.5 mm x H5 mm.

FIG. 2A is a waveform diagram showing an example of the waveforms of the accelerations Gx, Gy, Gz in the respective axial directions thus obtained (the horizontal axis is time, and the vertical axis is the detected voltage). This is the case of arm swing in a walking motion. Acceleration data is 5
The data is sampled at 0 Hz and stored as digital data for a predetermined time (it is arbitrary, for example, several seconds to several minutes. It is considered that the identification accuracy improves as the number of repetitions of the exercise increases during the period). Further, FIG. 2B is a perspective view of an estimated trajectory obtained by recombining the position information of the sensor obtained by integrating each acceleration data twice on three-dimensional coordinates, and the wrist is along the body side. Mainly in the Y direction. 4 is the start point of one measurement and 5 is the end point. (It is not a precise figure but a conceptual diagram.)

In the present invention, since the motion is finally analyzed by the repetitive motion of the arm, the offset component generated at the time of integration becomes a position integration error. Therefore, low frequency components (for example, about 1 Hz or less) must be removed from each acceleration waveform. Since the acceleration sensor itself also has a charge leak, DC components and ultra-low frequency components are automatically removed, but a high-pass filter is inserted in the sensor circuit to reliably cut frequency components below the target frequency. There is.

FIG. 3 is a perspective view defining the magnitude and direction of the position vector. V is a position vector at a certain time point obtained by integrating the acceleration twice, and the starting point of the vector is the center of gravity of the locus of positions within the integration period (the average position of all position vectors). Set as the origin. The tip of the position vector V represents the magnitude of displacement and the direction of displacement of the sensor (motion sensing module) at that moment. [Therefore, FIG. 2B shows the locus of the tip of the position vector V]. ] The azimuth of displacement is X− of the position vector V
The inclination angle θ from the Z plane and the projection of the position vector V on the XZ plane are represented by the angle φ formed with the Z axis. However, θ and φ are −90 ° <θ ≦ + 90 regardless of the direction of the position vector V.
, -180 ° <φ ≤ + 180 °.

A large number of position vectors are obtained within a predetermined period by numerically integrating a large number of acceleration data, and the motion of the user of the motion sensing module is identified from them. First, the average value of the magnitudes of many position vectors is related to the so-called amplitude of repetitive movements of the arm (wrist), and represents the strength of the movements. Can also be used. The average directions (azimuth angles θ, φ) of a large number of position vectors are considered to be peculiar to some extent depending on the type of motion. (For example, extend your elbow,
Or comparing the case of bending and shaking the arm, the direction of movement of the sensor naturally changes and the direction of the position vector also differs. )

FIG. 4 is a histogram showing various motions identified by the method of the present invention. The average magnitude of the position vector and the average azimuth θ and the position vector during one type of motion are shown. φ was graphed. The φ axis and the θ axis of the azimuth angle were set orthogonal to each other on the reference plane, and the average size of the position vector was taken as the height of the graph. A 40-year-old male healthy person was selected as a test subject, and a motion sensing module having a built-in acceleration sensor was wrapped around his left wrist and wearing athletic shoes. Sampling frequency is 50Hz, A / D
The conversion accuracy was 10 bits.

There were a total of nine types of exercises, and both general exercises and gesturing exercises of almost only the arms were taken up. 11
Is walking (walking while conscious of waving), 12
Is running (running), 13 is jogging, 14 is clapping, 15 is bye-bye (waving gesture), 16 is slow walking, 17 is normal walking, 18 is fast walking, 19 is pocket (both hands in pockets) Walking in the state).

The characteristics of the various motion patterns shown in FIG. 4 will be considered a little. Regarding walking, as the speed increases, the swing of the arm increases, the average θ also rises, and the Y-direction component increases. Walking, running, and jogging are actions in which the elbow is bent compared to walking (normal), so the component in the X-axis direction increases (the average φ approaches 90 °). Walking is stronger than running or jogging because it is conscious of swinging arms. In addition, since a large amount of impulsive harmonic components are included in the acceleration information of the clap, the method of directly extracting the motion pattern from the acceleration information, which has been often used, has a large variation and the identification accuracy is deteriorated. The same applies to the method using cross-correlation of acceleration waveforms. It was found that the method of using the double integration of the present invention is superior in the identification accuracy.

From FIG. 4, each type of motion has its own strength and directionality, and despite the fact that many types of motion are performed, the strength and direction pattern of the motion trajectory calculated from the acceleration causes the motion It is clear that can be clearly distinguished. Therefore, at least for some individuals, it is well possible to remember his data and compare them to identify the type of exercise if he does any of the same types of exercise again. Recognize. Furthermore, it is expected that there are some types of exercise that can be identified without registering personal characteristics. Further, the method of the present invention is likely to be applied and developed to an auxiliary tool that conveniently performs communication such as sign language (for example, enables interpretation).

Next, some modifications of the embodiment of the present invention will be described. First, how to take the average of the position vector data (size, azimuth angle) is not limited to the commonly used arithmetic average. As long as the aim is to improve the identification accuracy of various types of movements, a geometric mean, a weighted average with some weighting, a harmonic mean, or the adoption of a median, the processing after removal of outliers, and other data conversion to a function A generalized method of averaging afterwards can be used. It is also permissible to appropriately correct the calculation result.

The wrist, especially the left wrist, should be considered first as the body part to which the motion sensing module is attached, but other parts may also be considered depending on the conditions of the user and the type of motion to be identified. Can be used. In addition, there is a case where sensors are attached to multiple parts of the body and communicate with each other, and various motion data of different parts are integrated to identify more various or advanced motions without compromising the complexity of the device. obtain.

[0028]

In the motion identifying method and motion sensing module of the present invention, the motion is identified using the position vector obtained by integrating the acceleration data of three axes twice, so that the motion can be easily and reliably obtained. Identifies many types of exercise,
Moreover, it became possible to know the strength. Further, since only the acceleration sensor and the detection circuit having a simple structure can be used, the identification method is easy, and it is possible to contribute to downsizing, cost reduction, and power consumption reduction of the motion sensing module.

Further, when the low frequency component is removed from the acceleration data, the integration error can be reduced and the repetitive motion identification accuracy can be improved. Moreover, by attaching the acceleration sensor to the wrist, it was possible to carry out the identification measurement without disturbing the movement and without any discomfort. Further, it is preferable to use a plate-shaped piezoelectric sensor with an eccentric weight because a single compact vibrating body is a triaxial acceleration sensor, so that the motion sensing module can be simplified and downsized.

[Brief description of drawings]

FIG. 1 is a front view showing an example of an embodiment of a method and an apparatus of the present invention and directions of coordinate axes.

FIG. 2A is a waveform diagram showing an example of an acceleration waveform,
(B) is a perspective view showing position information obtained by integrating the acceleration twice.

FIG. 3 is a perspective view that defines the magnitude and direction of a position vector.

FIG. 4 is a histogram showing various movements identified by the method of the present invention.

FIG. 5 shows a specific example of a triaxial acceleration sensor used in the present invention, (a) is a plan view, (b) is a sectional view taken along line AA,
(C) and (d) are schematic views of a cross section showing a deformed state due to acceleration.

[Explanation of symbols] 1 Motion sensing module 2 armband 3 left wrist 4 starting point 5 end point 11 walking 12 run 13 Jogging 14 applause 15 bye 16 walking (slow) 17 walking (normal) 18 walking (fast) 19 pockets 20 Support 21 metal disk 22 central axis 23 Load mass 24 Piezoelectric plate 25, 26 X electrode film 27, 28 Y electrode film 29, 30, 31, 32 Z electrode film Fx X inertial force Fz Z inertial force Gx X direction acceleration Gy Y direction acceleration Gz Z acceleration V position vector X, Y, Z coordinate axes φ, θ Position vector azimuth

Claims (7)

[Claims] 1. An average magnitude and direction of a position vector within a predetermined time by integrating acceleration data measured by a triaxial acceleration sensor attached to a predetermined part of the body twice to obtain position information. And identifying the type of exercise performed based on the calculated data. 2. The motion identification method according to claim 1, wherein the acceleration data has low-frequency components below a predetermined frequency cut. 3. The movement identifying method according to claim 1, wherein the predetermined part of the body is a wrist. 4. A portable device that is equipped with a triaxial acceleration sensor and an arithmetic unit that adds a predetermined time to the acceleration data measured by the acceleration sensor and is worn on a predetermined part of the body. Then, the arithmetic device integrates the acceleration data twice to obtain position information, calculates the average magnitude and direction of the position vector within the predetermined time, and the calculation is performed based on the calculated data. A motion sensing module characterized by identifying the type of motion. 5. The motion sensing module according to claim 4, wherein the acceleration data has low-frequency components below a predetermined frequency cut. 6. The motion sensing module according to claim 4, wherein the predetermined part of the body is a wrist. 7. The triaxial acceleration sensor has a configuration in which a load mass whose center of gravity is far from the plane of the plate is fixed, and a deformation of the plate due to acceleration acting on the load mass is piezoelectrically detected. 7. A motion sensing module according to claim 4, 5 or 6.