KR20190143513A - Method and apparatus of recognizing a motion using the rectangular coordinate system - Google Patents

Abstract

The present invention relates to a motion recognition method, which comprises the steps of: obtaining an angular acceleration of a motion from a gyro sensor; obtaining the magnitude of the acceleration present on a plane perpendicular to the axis of rotation of the motion from the acceleration meter; calculating a rotation radius of the motion from the obtained angular acceleration and the magnitude of the acceleration; and determining the type of the motion by using the calculated rotation radius. The axis of rotation of the motion is characterized in that any one of the x-axis, y-axis, z-axis in the rectangular coordinate system. By obtaining the rotation radius from the sensor data and using the same for motion recognition, motions having similar sensor data can be more precisely distinguished.

Description

Method and apparatus for recognizing a motion using the rectangular coordinate system

The present invention relates to a motion recognition method. More particularly, the present invention relates to a motion recognition method and apparatus for more precisely distinguishing between motions having similar sensor data by obtaining a rotation radius from the sensor data and using the same for motion recognition. .

Wearable devices, including smartphones, smart watches, and Google Glass, have a variety of built-in sensors. Built-in sensors include camera sensors, acoustic sensors, proximity sensors, illuminance sensors, gravity sensors, acceleration sensors, magnetic field sensors, acceleration sensors, and more. Among these, sensors, such as acceleration sensors and gyro sensors, whose values change in accordance with a user's movement are called motion sensors.

An acceleration sensor is a sensor that measures a change in speed. And since the velocity represents the change in travel distance, this relationship can be used to calculate the travel distance per unit time. Gyro sensor refers to a sensor that measures the average angular acceleration per unit time. At this time, the gyro sensor measures the angular acceleration value for each axis according to the control axis. Usually, three orthogonal axes are used as the control axis. The three-axis gyro sensor used in smartphones is called pitch, roll, and yaw. The gyro sensor is capable of measuring the most accurate rotational variation among a single sensor, but has a feature that continuously accumulates error values from a long time point of view.

Conventionally, statistical values obtained from the sensor data of the acceleration sensor and the sensor data of the gyro sensor are used for motion recognition, but there is a problem in that the motions having similar sensor data cannot be accurately distinguished.

Accordingly, the first problem to be solved by the present invention is to provide a motion recognition apparatus that can more accurately distinguish a motion having similar sensor data by obtaining a rotation radius from the sensor data and using it for motion recognition.

A second problem to be solved by the present invention is to provide a motion recognition method that can more accurately distinguish a motion having similar sensor data by obtaining a rotation radius from the sensor data and using it for motion recognition.

Further, the present invention provides a computer-readable recording medium having recorded thereon a program for executing the above method on a computer.

In order to achieve the first object, the present invention obtains the angular acceleration from the angular velocity of the motion obtained from the gyro sensor, and obtains the magnitude of the acceleration present on a plane perpendicular to the axis of rotation of the motion from the acceleration obtained from the acceleration meter, A feature extractor configured to calculate a motion feature including a rotation radius of the motion calculated from the obtained angular acceleration and the magnitude of the acceleration; A motion learning unit receiving the motion feature from the feature extractor and performing machine learning; A storage unit which stores a result of the motion learning unit; And a motion determination unit configured to determine a type of the new motion by comparing with a result of the motion learning unit when the feature extractor extracts a motion feature for a new motion, wherein the axis of rotation of the motion is an x-axis, y in a rectangular coordinate system. It provides a motion recognition device characterized in that any one of the axis, z axis.

The present invention comprises the steps of obtaining the angular acceleration from the angular velocity of the motion obtained from the gyro sensor to achieve the second object; Obtaining the magnitude of the acceleration present on a plane perpendicular to the axis of rotation of the motion from the acceleration obtained from the acceleration meter; Calculating a rotation radius of the motion from the obtained angular acceleration and the magnitude of the acceleration; And determining the type of the motion using the calculated rotation radius, wherein the rotation axis of the motion is any one of an x axis, a y axis, and a z axis in a Cartesian coordinate system. to provide.

According to an embodiment of the present disclosure, the method may further include calculating at least one statistical value from the acceleration or the angular velocity of the motion in addition to the rotation radius, and determining the type of the motion may include calculating the calculated statistical value and rotation. It is preferable to determine the type of the motion using a radius.

In addition, machine learning using the calculated statistical value and the rotation radius; And storing the machine learning result, wherein the determining of the type of motion comprises comparing the stored machine learning result with a machine learning result for a newly input motion to determine the type of the newly input motion. can do.

In order to solve the above technical problem, the present invention provides a computer-readable recording medium that records a program for executing the above-described motion recognition method on a computer.

According to the present invention, by obtaining a rotation radius from the sensor data to utilize in motion recognition, it is possible to more precisely distinguish the motion having similar sensor data.

1 is a block diagram of a motion recognition apparatus according to an embodiment of the present invention.
2 is a flowchart illustrating a motion recognition method according to an exemplary embodiment of the present invention.
3 is a flowchart illustrating a motion recognition method according to another embodiment of the present invention.
Figure 4 shows the result of obtaining the value of the radius when the user performs the biceps curl, Figure 5 when the front raise.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Advantages and features of the present invention and methods for achieving them will be apparent with reference to the embodiments described below in detail with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but can be embodied in various different forms, and the present embodiments only make the disclosure of the present invention complete, and those of ordinary skill in the art to which the present invention belongs. It is provided to fully inform those skilled in the art of the scope of the invention, which is defined only by the scope of the claims.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, "comprises" and / or "comprising" does not exclude the presence or addition of one or more other components in addition to the mentioned components. Like reference numerals refer to like elements throughout.

Unless otherwise defined, all terms (including technical and scientific terms) used in the present specification may be used in a sense that can be commonly understood by those skilled in the art. In addition, terms that are defined in a commonly used dictionary are not ideally or excessively interpreted unless they are specifically defined clearly.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a motion recognition apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the motion recognition apparatus according to the present embodiment includes a feature extractor 110, a motion learner 120, a storage 130, and a motion determiner 140.

The motion recognition apparatus according to the present invention may further include a gyro sensor and an acceleration meter. At this time, the gyro sensor uses the angular velocity ω as the output value and the acceleration meter uses the acceleration a as the output value. In some cases, the gyro sensor may use the angular acceleration as an output value.

The feature extractor 110 obtains the angular acceleration from the angular velocity ω of the motion obtained from the gyro sensor, and obtains the magnitude of the acceleration existing on a plane perpendicular to the axis of rotation from the acceleration a obtained from the acceleration meter. A motion characteristic is calculated that includes the radius of rotation of the motion calculated from the angular acceleration and the magnitude of the acceleration.

The motion feature refers to a feature in which sensor data input from a gyro sensor, an accelerometer, a geomagnetic sensor, or the like is distinguished from other sensor data. The motion feature may represent an entire portion of the sensor data or may represent a portion of the sensor data. For example, the motion feature may comprise at least one statistical value of sensor data. For example, the statistical value may include a maximum value, a minimum value, a median value, an average value, an interquartile range (IQR) value, a root mean square (RMS) value, and the like, of a predetermined number of sampled sensor data. Do not.

The motion learner 120 receives the motion feature from the feature extractor 110 and performs machine learning.

Examples of machine learning include a K-nearest neighbors (KNN) algorithm or Support Vector Machines (SVM). By the KNN algorithm, a plurality of motion features may be compared with previously classified or learned motion features, and classified into groups having the same or similar tendency.

The storage unit 130 stores the classification result of the motion learning unit 120. The result stored in the storage unit 130 is used to compare the motion feature extracted by the feature extractor 110 with respect to the new motion with the result of the machine learning performed by the motion learner 120.

The motion determination unit 140 determines the type of the new motion by comparing the result of the machine learning with respect to the motion feature for the new motion and the result stored in the storage unit 130.

2 is a flowchart illustrating a motion recognition method according to an exemplary embodiment of the present invention.

Referring to FIG. 2, the motion recognition method according to the present embodiment includes steps that are processed in time series in the motion recognition apparatus illustrated in FIG. 1. Therefore, even if omitted below, the above descriptions of the motion recognition apparatus shown in FIG. 1 also apply to the motion recognition method according to the present embodiment.

In operation 210, the motion recognition apparatus may acquire an angular acceleration from the angular velocity of the motion obtained from the gyro sensor or directly obtain the angular acceleration from the gyro sensor.

The angular acceleration satisfies Equation 1 below.

의 시간동안 시간 t+1에서의 각속도(

)와 시간 t에서의 각속도(

)의 차이는 시간 t에서의 각가속도(

)와 같다.

Angular velocity at time t + 1 for

) And the angular velocity at time t (

), The difference is the angular acceleration at time t (

)

가 일정하다고 가정하면, 다음의 수학식 2로 바꿔 쓸 수 있다.

If is assumed to be constant, it can be rewritten as Equation 2 below.

는 1개 축의 각가속도, ω는 각속도, t는 시간, c는 상수이다. 상기 1개 축은 x축, y축, z축 중 어느 하나가 될 수 있으며, 각 축을 중심으로 한 회전은 피치(Pitch), 롤(Roll), 요(Yaw)로 정의할 수 있다.here,

Is the angular acceleration of one axis, ω is the angular velocity, t is the time, and c is the constant. The one axis may be any one of an x axis, a y axis, and a z axis, and a rotation about each axis may be defined as a pitch, a roll, or a yaw.

의 시간동안 각속도의 차이에 비례한다.Therefore, the angular acceleration of the motion

It is proportional to the difference in angular velocity over time.

)와 시간 t에서의 각속도(

)의 상대적 크기에 따라 계속 변화한다. 상기 상대적 크기에 따라 각가속도의 값은 양수, 음수, 또는 0이 될 수 있다. The angular acceleration according to Equation 2 is the angular velocity at time t + 1 (

) And the angular velocity at time t (

And continue to change with the relative size of the According to the relative magnitude, the value of the angular acceleration may be positive, negative, or zero.

In operation 220, the motion recognition apparatus obtains the magnitude of the acceleration existing on the plane perpendicular to the axis of rotation of the motion from the acceleration obtained from the acceleration meter.

At this time, the acceleration of the rotation of the motion can be obtained after removing the influence of the acceleration in the gravity direction to obtain the magnitude of the acceleration. It is preferable that the acceleration of the three axes from which the influence of the gravity direction is removed is used as the input of the sensor fusion algorithm. The sensor fusion algorithm may be a Kalman filter, a madgwick filter, or a complementary filter.

On the other hand, among the three axes in the Cartesian coordinate system, the magnitude of the acceleration in the other two axes (axis1, axis2) other than the axis in which the motion rotates is as follows.

Where a is the linear acceleration.

If the axis on which the motion rotates is the X axis, two axes other than the axis on which the motion rotates may be the Y axis and the Z axis.

In operation 230, the motion recognition apparatus calculates a motion feature including a rotation radius of the motion calculated from the obtained angular acceleration and the magnitude of the acceleration.

(a는 선가속도,

는 각가속도)이 도출된다.Differentiate both sides of v = rω (v is linear velocity, ω is angular velocity)

(a is the linear acceleration,

Angular acceleration) is derived.

Therefore, the radius r is derived from Equations 2 and 3 as shown in Equation 4 below.

가 0이거나 매우 작은 값을 가질 때, 무한대와 같이 큰 값을 갖게 된다. 또한, 시간 t+1에서의 각속도(

)와 시간 t에서의 각속도(

)의 상대적 크기에 따라서 반지름 r은 양의 값과 음의 값을 번갈아 가면서 나타나게 된다.Radius r is

Is 0 or has a very small value, it has a large value such as infinity. In addition, the angular velocity at time t + 1 (

) And the angular velocity at time t (

Depending on the relative magnitude of), the radius r appears alternately between positive and negative values.

Therefore, in order to calculate the radius r as a positive value, it can be rewritten as Equation 5 below.

A radius exceeding an arbitrary threshold value is classified as an extreme value and excluded, and the radius r is converted into a positive value as shown in FIGS. 4 and 5.

FIG. 4 shows the result of obtaining the radius value when the user performs the biceps curl and FIG. 5 the front raise.

4 and 5 show the results of obtaining a radius value every 1/50 second (20 ms) according to Equation 5. FIG.

4 and 5, it can be seen that the average of the radius r value of the front raise derived through the motion recognition method according to the present invention is larger than the average of the radius r value of the biceps curl. The average radius of the biceps curl is about 0.03, and the radius of the front raise is about 0.07, which is about twice.

In more detail, any one of the X, Y, and Z axes may be determined as the rotation axis of the motion. The angular velocity of the axis determined by the rotation axis of the motion can be obtained from the gyro sensor.

For example, if the Y axis is determined as the axis of rotation of motion, the acceleration uses the acceleration component on the X-Z plane. In this case, Equation 5 may be rewritten as follows.

a _X and a _Z are the X-axis component and the Z-axis component of the acceleration, respectively, and the angular velocity ω is the angular velocity of the rotating angle about the Y-axis.

The motion feature may further include at least one statistical value from the acceleration and the angular velocity of the motion in addition to the rotation radius.

In operation 240, the motion recognition apparatus determines the type of the motion by using the calculated rotation radius.

In addition to the rotation radius, it is preferable to further determine the type of the motion by using a statistical value included in the motion feature.

3 is a flowchart illustrating a motion recognition method according to another embodiment of the present invention.

Referring to FIG. 3, the motion recognition method according to the present embodiment includes steps that are processed in time series in the motion recognition apparatus illustrated in FIG. 1. Therefore, even if omitted below, the above descriptions of the motion recognition apparatus shown in FIG. 1 also apply to the motion recognition method according to the present embodiment.

Steps 310 to 330 are the same as steps 210 to 230 of FIG. 2 and will not be described.

In operation 340, the motion recognition apparatus calculates at least one statistical value from the acceleration or the angular velocity of the motion in addition to the rotation radius.

In operation 350, the motion recognition apparatus performs machine learning using the rotation radius and the statistical value. The machine learning result performed at this time is stored in a separate memory and then used to be compared with the machine learning result of the new motion.

In operation 360, the motion recognition apparatus determines the type of the newly input motion by comparing the machine learning result stored in operation 350 with the machine learning result for the newly input motion.

Among the movements performed by the user, in particular, the rotational movement having two or more joints is irregular in the center and the linear acceleration of the rotational movement is also irregular. Therefore, the present invention can be well applied to rotational movements such as biceps curl or front raise in which the center of rotation does not change significantly and the linear acceleration of the rotational movement does not change significantly.

The term '~ part' used in the present embodiment refers to software or a hardware component such as a field-programmable gate array (FPGA) or an ASIC, and '~ part' performs certain roles. However, '~' is not meant to be limited to software or hardware. '~ Portion' may be configured to be in an addressable storage medium or may be configured to play one or more processors. Thus, as an example, '~' means components such as software components, object-oriented software components, class components, and task components, and processes, functions, properties, procedures, and the like. Subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. The functionality provided within the components and the 'parts' may be combined into a smaller number of components and the 'parts' or further separated into additional components and the 'parts'. In addition, the components and '~' may be implemented to play one or more CPUs in the device or secure multimedia card.

All of the above functions may be performed by a processor such as a microprocessor, a controller, a microcontroller, an application specific integrated circuit (ASIC), or the like according to software or program code coded to perform the function. The design, development and implementation of the code will be apparent to those skilled in the art based on the description of the present invention.

Although the present invention has been described above with reference to the embodiments, it will be understood by those skilled in the art that the present invention may be variously modified and changed without departing from the spirit and scope of the present invention. I can understand. Therefore, the present invention is not limited to the above-described embodiment, and the present invention will include all embodiments within the scope of the following claims.

Claims (5)

Obtaining angular acceleration of motion from a gyro sensor;
Obtaining from the acceleration meter the magnitude of the acceleration present on a plane perpendicular to the axis of rotation of the motion;
Calculating a rotation radius of the motion from the obtained angular acceleration and the magnitude of the acceleration; And
Determining the type of the motion by using the calculated rotation radius;
The rotation axis of the motion is any one of the x-axis, y-axis, z-axis in the Cartesian coordinate system. According to claim 1,
Calculating at least one statistical value from the acceleration or the angular velocity of the motion in addition to the rotation radius,
Determining the type of motion,
And determining the type of the motion by using the calculated statistical value and the rotation radius. The method of claim 2,
Machine learning using the calculated statistical value and the rotation radius; And
Storing the machine learning results;
Determining the type of motion,
And comparing the stored machine learning result with the machine learning result for the newly input motion to determine the type of the newly input motion. Acquiring the angular acceleration of the motion from the gyro sensor, obtaining the magnitude of the acceleration existing on the plane perpendicular to the axis of rotation of the motion from the acceleration meter, and calculating the rotation radius of the motion calculated from the obtained angular acceleration and the magnitude of the acceleration. A feature extractor for calculating a motion feature comprising;
A motion learning unit receiving the motion feature from the feature extractor and performing machine learning;
A storage unit which stores a result of the motion learning unit; And
When the feature extractor extracts a motion feature for a new motion, and comprises a motion determination unit for determining the type of the new motion compared to the result of the motion learning unit,
And a rotation axis of the motion is any one of an x axis, a y axis, and a z axis in a Cartesian coordinate system. A computer-readable recording medium having recorded thereon a program for executing the method of any one of claims 1 to 3.

Images