KR20140086426A - Method for estimating movement pattern with multi sensor data and smartwear using the same - Google Patents

Abstract

Provided are a movement pattern estimating method using multi-sensor data and a smartwear with the same. The movement pattern estimating method using multi-sensor data according to an embodiment of the present invention identifies a movement pattern of a person wearing clothing by using acceleration data, and identifies the movement pattern of a person wearing clothing by using gyroscope data when the movement pattern is not identified by the acceleration data. Thus, a problem of information reliability arising from an error of smartwear system information may be solved, and an accurate method of estimating various movements and action patterns through multiple sensors can be provided.

Description

TECHNICAL FIELD The present invention relates to a mobile pattern inference method using multi-sensor data, and a smart-

The present invention relates to a method of deducing a movement pattern of a smart wearer and a method of correcting sensor data in the process.

In the last three years, 75.1% of the national park mountain accidents have been analyzed as the result of unreasonable hiking and lack of preparation. According to the mountain safety accident analysis by the Ministry of Environment in the year 07'-09 ', 41.6% of fatal accidents are heart diseases and 26.3% are accidents caused by exhaustion and convulsions.

According to the 2010 Mountain Accident Analysis by the Japan Metropolitan Police Agency, more than 76% of the victims are middle-aged, and more than 60% of them are out of fitness. Accidents caused by similar causes also occur in bicycle riding.

Smartwares can help you avoid such accidents by measuring your workout and providing information or warning you of dangerous situations.

SmartWare measures the momentum using sensors, and has a reliability problem in sensor information of a conventional smartware system using a single sensor. Therefore, it is necessary to apply various sensors to smartwares without using a single sensor such as gyro or acceleration, so as to have information reliability.

The gyro sensor is vulnerable to continuous operation (ex, equivalent speed). For example, if continuous operation is performed, very short error information can be displayed in a specific section on the collected data. Therefore, if the cumulative error information is reflected, the reliability of the service or the provided information may be adversely affected.

In addition, the acceleration sensor generates an error rate in accordance with a continuous gradient change. The increase / decrease ratio of the speed shows the increase / decrease of the speed with respect to the linear motion in the specific direction. However, an error occurs because the object is rotated based on the time filter by considering the linear motion.

Accordingly, there is a need for finding a way to provide smartware capable of providing reliable exercise information.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and it is an object of the present invention to solve the information reliability problem of smartware with multiple sensors and error control filters, and to accurately predict the various motions and behavior patterns of users And to provide smartware using the same.

According to an aspect of the present invention, there is provided a method for estimating a movement pattern using multi-sensor data, the method comprising: collecting acceleration data from an acceleration sensor provided in a garment; Collecting gyro data from a gyro sensor provided in the garment; A first grasping step of grasping a movement pattern of the wearer using the acceleration and temperature sensor data; And grasping a movement pattern of the wearer using the gyro data if the movement pattern is not recognized by the first grasping step.

If the absolute value of the difference between the current acceleration data and the previous acceleration data exceeds the first criterion, then the first grasp step is regarded as the occurrence of the movement if the current value is greater than the energy value of the ' Can be handled.

The first criterion may include: calculating an average of the acceleration data; Calculating a variance of the acceleration data; Removing acceleration data outside the variance from the average; And setting a variance of acceleration data not removed by the removing step as the first criterion; And comparing the energy value with the energy value of the 'slow' state.

If the absolute value of the difference between the current acceleration data and the previous acceleration data exceeds the second criterion, then the second determination step considers that the movement has occurred, and if the absolute value of the difference between the current acceleration data and the previous acceleration data exceeds the second criterion, Can be handled.

The second criterion may include calculating an average of the gyro data; Calculating a variance of the gyro data; Removing gyro data outside the variance from the average; And setting a variance of gyro data not removed by the removing step as the second criterion.

Collecting temperature data; Correcting the acceleration data based on the temperature data; And correcting the gyro data based on the temperature data.

According to another embodiment of the present invention, a smart ware using a multisensor includes an acceleration sensor for collecting acceleration data; A gyro sensor for collecting gyro data; And a control unit for recognizing a movement pattern using the acceleration data and recognizing a movement pattern using the gyro data if the movement pattern is not recognized by the acceleration data.

INDUSTRIAL APPLICABILITY As described above, according to the present invention, the problem of information reliability due to errors in smartware system information can be solved. In addition, it is possible to provide an accurate reasoning method of various motions and behavior patterns through multiple sensors. Furthermore, it is possible to improve the reliability through the multi-sensor and error control filter.

1 illustrates smartware according to an embodiment of the present invention;
FIG. 2 is a detailed block diagram of the main module shown in FIG. 1,
FIG. 3 is a flow chart for explaining a process of inferring a movement pattern by collecting / correcting sensor data;
4 is a detailed flowchart of the correction algorithm,
5 is a diagram provided in a side view of three-dimensional vector calculation,
6 is a detailed flowchart of the context recognition algorithm,
FIG. 7 is a diagram showing movement pattern factor values through acceleration data, and FIG.
8 is a diagram showing movement pattern factor values through gyro data.

Hereinafter, the present invention will be described in detail with reference to the drawings.

1 is a diagram illustrating a smart ware according to an embodiment of the present invention. As shown in FIG. 1, temperature sensors for measuring body temperature are attached to various positions in smartwares, and include a battery for power supply, an MCU for controlling the entire smartware system, and a main module.

The main module measures exercise information of the smart wearer and transmits body temperature information measured by the temperature sensors to the wearer's smartphone. In addition, if a risk situation is predicted to occur through exercise information and body temperature information, it is visually warned.

2 is a detailed block diagram of the main module. 2, the main module includes a local communication module 110, a correction unit # 1 120, a fabric display 130, a control unit 140, a cloth display control unit 150, 160, a temperature sensor 170, a gyro sensor 180, and an acceleration sensor 190.

The short distance communication module 110 is a module for performing wireless communication (Wi-Fi, Bluethood, ZigBee, etc.) with the wearer's smart phone. The first module 120 collects temperature data measured by the smart sensors / Correct.

The fabric display 130 is a display for displaying warning warnings / messages when the wearer is in a dangerous state with respect to body temperature and momentum, and is controlled by the fabric display control unit 150 provided in the control unit 140. [

The correction unit # 2 160 provided in the control unit 140 collects and corrects the data measured by the temperature sensor 170, the gyro sensor 180 and the acceleration sensor 190 and calculates the movement pattern of the wearer (walking, Stair climb, elevator movement, etc.).

When an abnormality occurs in the correction unit # 1 (120) in order to ensure continuous data collection, the role of the correction unit # 2 (160) is parallel to the role of the correction unit # (# 1) 120 in parallel.

The correction unit # 2 160 first grasps the movement pattern of the wearer using the acceleration data. If the movement pattern is not recognized by the acceleration data, the correction unit # 2 160 recognizes the movement pattern of the wearer using the gyro data.

At this time, the movement occurrence judgment based on the acceleration data is regarded as a movement when the data larger than the slow data occurs, and the movement continuously occurs when the absolute value of the difference between the current acceleration data and the previous acceleration data exceeds the variance Handled. Here, the variance is calculated after eliminating the acceleration data largely deviated from the dispersion in the section average. This is the same for gyro data.

3 shows a process in which the correction unit # 2 160 160 estimates a movement pattern by collecting / correcting data measured by the temperature sensor 170, the gyro sensor 180, and the acceleration sensor 190. FIG.

3, the correction unit # 2 160 collects data from the temperature sensor 170, the gyro sensor 180 and the acceleration sensor 190 (S210), and performs FFT (S220).

When the temperature change by the temperature sensor 170 exceeds the reference value (S230), the gyro sensor 180 and the acceleration sensor 190 are corrected by referring to the correction table (Factory Calibration) (S240) The error is compensated (S250).

For example, the gyro sensor is normally distributed at 25-27 ° C, but at 30 ° C there is an error of ± 1% in the measured value. Therefore, it is necessary to modify the gyro sensor by steps S240 and S250.

Thereafter, the data of the gyro sensor 180 and the acceleration sensor 190 are corrected through the correction algorithm (S260), the movement pattern is inferred through the context recognition algorithm (S270), and is exposed to the wearer (S280).

Hereinafter, the correction algorithm used in step S260 and the context recognition algorithm used in step S270 will be described in detail.

4 is a detailed flowchart of step S260 of FIG. As shown in FIG. 4, the correction algorithm calculates the average and variance of the sensor information S261 corrected in step S250 (S262), determines the mobility using the calculated average and variance S263, Dimensional space comparison (S264), a window algorithm (S265), and a three-dimensional vector and a rotation axis are calculated (S266).

Gyro data and acceleration data are defined as follows.

In this case, if the errors of the gyro data and the acceleration data according to the temperature are defined as (g _ei , g _ej , g _ek ) and (a _ei , a _ej , a _ek ), the gyro data and the acceleration data are represented as follows .

On the other hand, since the discrimination based on the movement and the vibration is performed in units of several tens seconds, the data in the short interval is calculated as the average data per second. Moreover, the body temperature of the wearer does not change in seconds, but gradually changes over time. Therefore, the deviation of the error axis is based on 1 second. The sensor data is received at a cycle of 10ms-100ms. Gyro data and acceleration data can be expressed as follows because the error variation according to the temperature provided by the sensor manufacturer provides an error according to the unit of 1-2 ° C.

At this time, the change of (g _ei , g _ej , g _ek ) and (a _ei , a _ej , a _ek ) in the garment may uniformly and uniformly appear during the sensor confidence time (G _t , A _t ).

The period of the sensor mean (m _G, m _A) according to various experiments (x, y, z) is the error correction _{Gt = (g 'x, g} ' y, g 'z) and At = (a' _x, a ' _y , a' _z ). This is defined as the interval, and the discrimination method for each interval serves to minimize the cumulative error of the draft data in the two kinds of sensors.

The determination of acceleration is equal to the gravitational acceleration of the output of the three-axis accelerometer with the gravitational acceleration component. Therefore, only the component of gravitational acceleration is included. The following equation is satisfied.

In addition, not only the load due to the calculation is reduced, but also the errors generated due to the high noise and sensitivity are satisfied by comparing the variances.

The range of error to be generated can be described as a threshold value for determining the acceleration provided by the manufacturer.

If the error range is exceeded, the individual axis data is calculated on the X, Y, and Z axes, assuming movement.

,

,

) 보다 크다면, 그 축의 방향으로 이동했다고 가정한다.At present, according to the received time data to the individual axes _{(a 'xt, a' yt} , a 'zt) after receiving the data _{(a' xt +1, a '} yt +1, a' zt +1) the value obtained by subtracting the dispersed value(

,

,

), It is assumed that it has moved in the direction of the axis.

Here, the interval data is a time-based windows value. The size of windows according to time can be defined as follows.

However, the data reception for the x-axis is as follows.

Here, windows are acceleration data to the t ~ t + 1 is a 'can be represented by _xi, a' _xi ∈ a 'and _xt, j is a data array in the x accelerometer.

를 범위 내에서 반복적으로 비교하여 분산 보다 작은 경우 같은 windows에 속하고 벗어날 경우 다른 windows로 만든다. 윈도우 알고리즘에서는 구간 (t-1, t, t+1)간의 데이터 변화를 비교한다.That is, | a ' _xi -a' _{xi +1} |

In the range, and if it is smaller than the variance, it belongs to the same windows, and if it deviates, it makes it to other windows. The window algorithm compares the data changes between the intervals (t-1, t, t + 1).

)을 아래와 같이 나타낼 수 있다.Assuming that the received data aggregation is _xdk during the time t to t + 1 in the value of windows, a ' _xi is one of the received x data from t to t + 1, and there are a total of k , The occurrence averages (m _At ) and variance (t) during the period (t ~ t + 1)

) Can be expressed as follows.

If five values (100, 105, 110, 95, 80) are received for one second, as shown in Figure 5, and the variance is 10, the average is 98. Therefore, according to the above equation, only four values (100, 105, 110, 95) are collected and an average of 102.5 is x-axis data of the three-dimensional vector.

The three-dimensional vector is as follows.

At this time, the vector represents the magnitude of the motion by the amount of momentum, and there is a great possibility that only one axis and two axes are represented as a graph depending on the motility. Therefore, it is necessary to deduce the angle on the corresponding two axes. The method of obtaining arbitrary rotation angles (ρ, φ, θ) of three axes can be defined as follows using Euler's offense.

ρ is the pitch along the x-axis, φ is the roll, and θ is theta.

As a result, the direction and speed of the wearer can be determined through the rotation angle.

Hereinafter, the context recognition algorithm used in step S270 will be described in detail. 6 is a detailed flowchart of step S270 of FIG. As shown in FIG. 6, the situation recognition algorithm determines the movement pattern of the smart wearer (S272) by using the calculated acceleration data (S271) and deduces the movement pattern (S275).

However, as shown in FIG. 7, a period where energy overlaps is generated. In this case, it is impossible to determine the movement pattern by the acceleration data (S272-N).

For example, an elevator can be analyzed because only the z-axis changes significantly. However, since it is difficult to distinguish the rotation component, it is desirable to provide a linear component with the acceleration sensor and classify the behavior through the rotation component and the position component of the gyro sensor.

Therefore, if the movement pattern can not be determined by the acceleration data (S272-N), the movement pattern of the smart wearer is determined (S274), and the movement pattern is inferred (S275). The movement pattern using the gyro data can be judged by referring to the table of FIG.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, but, on the contrary, It should be understood that various modifications may be made by those skilled in the art without departing from the spirit and scope of the present invention.

110: Local area communication module 120: Correction block # 1
130: Fabric display 140:
150: fabric display control unit 160: correction unit # 2
170: Temperature sensor 180: Gyro sensor
190: Accelerometer

Claims (7)

Collecting acceleration data from an acceleration sensor provided in the garment;
Collecting gyro data from a gyro sensor provided in the garment;
A first grasping step of grasping a movement pattern of the wearer using the acceleration data; And
And determining the movement pattern of the wearer using the gyro data if the movement pattern is not recognized by the first grasping step.
The method according to claim 1,
Wherein the first grasping step comprises:
Wherein movement is judged to be continuous when the absolute value of the difference between the current acceleration data and the previous acceleration data exceeds the first criterion and movement is continuously generated.
3. The method of claim 2,
The first criterion may include:
Calculating an average of the acceleration data;
Calculating a variance of the acceleration data;
Removing acceleration data outside the variance from the average; And
And setting a variance of acceleration data not removed by the removing step as the first criterion.
The method according to claim 1,
Wherein the second grasping step comprises:
Wherein movement is judged to be continuous when the absolute value of the difference between the current gyro data and the previous gyro data exceeds the second criterion.
5. The method of claim 4,
The second criterion may include:
Calculating an average of the gyro data;
Calculating a variance of the gyro data;
Removing gyro data outside the variance from the average; And
And setting a variance of the gyro data not removed by the removing step as the second criterion.
The method according to claim 1,
Collecting temperature data;
Correcting the acceleration data based on the temperature data; And
And correcting the gyro data based on the temperature data based on the temperature data.
An acceleration sensor for collecting acceleration data;
A gyro sensor for collecting gyro data; And
And a control unit for recognizing a movement pattern using the acceleration data and recognizing a movement pattern using the gyro data if the movement pattern is not recognized by the acceleration data, Wear.

Images