Sie sind auf Seite 1von 4

Motion Reconstruction Using and Inertial Measuring Unit and Accelerometer Data

Jaime Andres Rincon Aranago e-mail: jairinar@posgrado.upv.es, jarain78@gmail.com June 22, 2012
Abstract: The motion capture using inercial measure units (IMU) and development system ARDUINO MEGA that allow the reconstruction of the movement in a 3D environment , is proposed in this paper. The 3D accelerometers output (Volts) has become in gravit units (g) for the accelerometer and degrees per second (g/s) for the gyroscope by an analogue digital converter(ADC) of 10 bits. With these values can be calculated the angles of inclination of the object, then sent them by radio frequency to the computer across a module of communication xbee and received them using the software PROCESSING. My approach relies in build up the object as 3D solid with the OPENGL tool and with the already processed information make the reconstruction of the movement. This procedure can be made in online modality in which the movements of the object are represented in real time. Keywords: Acelerometro, Gyroscopio tructure; another method is the using accelerometers accompanied of gyroscopes to calculate the angles produced by the movement of the actor. This method allows to place the sensors to the actor and when he executes the movements, the sensors capture them to be reconstructed.

MOTION CAPTURING AND REAL TIME PROCESSING DATA

INTRODUCTION

I use Two accelerometers and a gyroscope which provide the control input to reconstruction framework. The sensors were attached to the lower arm next to the wrist. The accelerometer allows me to measure the acceleration of the arm when it moves, this acceleration captured by the sensor does not come in volts, to obtain them g it is necessary to realize a process to turn this voltage in g, this conversion to g is necessary to do for the axis Xequation 1, axis Y equation 2 and axis Z equation 3 [starlino,1][Paul D,4]. ADC1 V cc 1023 ADC2 V cc V oltB = 1023 ADC3 V cc V oltC = 1023 V oltA = where: ADC1 , ADC2 , ADC3 : are the values of analogue digital convertion (ADC), of each sensor (Accelerometer and gyroscope)

The increasing availability and demand of highquality motion capture (mocap) data has become a driving force for the development of data-driven methods in computer animation and mostly due to the increasing market of the video games and movies. ; In this paper I propose the motion capture using accelerometer and gyroscopes knows as IMU (Inertial measurement unit) together with ARDUINO MEGA tries to reduce the costs but without harming the capture precision.

(1) (2) (3)

RELATED WORK

For the motion capture is commonly used technology of optical type with multi cameras [1], the prob- VoltA, VoltB, VoltC : Voltages of the senlem is its high cost and complex controlled infras- sors outputs 1

Constant 1023 : 210

Having the componentes of the force, I calculate the vector R, using the equation 10

Having the voltages corresponding to every (10) R = Rx2 + Ry 2 + Rz 2 sensor, it is necessary to reduce VzeroG,every accelerometer has VzeroG, and it can be found in the accelerometers data sheet, for my accelerometers Having the vector calculated already R now I can VzeroG = 1.65 V ; having VzeroG and the voltage calculate the angles Axr, Ayr and Azr, equations of every Axis, they are reduced to obtain the delta 11,12 and 13. of V, since is shows in the Equation 4,5,6. Rx Axr = arccos (11) R DeltaV oltsRx = V oltA1.65V (4) Ry Ayr = arccos (12) DeltaV oltsRy = V oltB 1.65V (5) R DeltaV oltsRz = voltC 1.65V (6) Azr = arccos Rz R (13)

This process is realized for both accelerometers. Having the Delta-V and knowing the sensibili- Having the angles calculated of each one now I can dad, which is also give it by the manufacturer I calculate the angle of the gyroscope. can obtain them g that the sensor is capturing, The gyroscope allows me to measure deg s , this gyusing the Equations 7,8,9. roscope of three axes captures the movement in X, Y and Z, this sensor and one of the accelerometers has been located in the joint of the arm and the Rx = DeltaV oltsRx/0.4785V /g (7) wrist. To calculate them deg s it is needed to know the Ry = DeltaV oltsRy/0.4785V /g (8) sensibility of the gyroscopio, for this I use sensor data sheet and know that the above mentioned Rz = DeltaV oltsRz/0.4785V /g (9) value is of 0.002V/deg/s now I proceed to do With this I have obtained three components the calculations. showed in the gure 1. V aref ResADC Sencivilidad (14)
ResADC 360

RateGyro = V alueA DC

In the gure 2 the vector corresponding to the gyroscope appears. Now already I have calculated the angles of the accelerometer and of the gyroscope. Now I proceed to integrate them to obtain the estimation of the angle in which nds the arm, to do this estimation I use a lter Kalman [4, p ag. 269], this allows to determine the angle and eliminates the noise. In the gure 3, we can observe the sign obtained of the movement of the arm without the Kalman lter and sign peaks which provide wrong movements to the 3D model. Figure 1: Component of the Inercial Vector Force, In the gure 4 the movement appears but now [starlino,3] with a lter kalman, notice that there are not peaks and the sign is alike a senosoidal sign. 2

movements was done and as 3D bookshop I use Opengl 2 , to do the reconstruction. The values of angles were sent by serial port communication to the computer. Using processing the serial port was readed and the data separation since ARDUINO MEGA sends the information the following form: actAngleAnteBrazoX !actAngleAnteBrazoY [actAngleAnteBrazoZ @actAngleBrazoX /actAngleB razoY |actAngleB razoZ (interval] Where the special characters serve as dividers between data. Having already the separated angles they are saved in a text le to be reproduced then. For the reproduction of the movements it is necessary to have two variables: Angle and Time; with this I reproduce the capture as a movie. I design an arm with opengl and using the funcFigure 2: Component of the Inercial Vector Force, tion gluCylinder I built a cylinder, then is made [starlino,3] an adjournment of coordinates with the end that the top part of the arm (shoulder) stays in a xed point and the forearm moves to the low part of the arm (elbow); having already the arm and forearm located and with the angles captured by the sensors it get in the counterfoil of 3D rotation, with the function of openg glRotatef (XCodo, 1.0, 0.0, 0.0), this function rotates on the axis X the angle that the variable Xcodo has, other two axes are not used, since biomechanically the elbow turns in one axis. Figure 3: Sign without kalman lter

RESULTS

The reconstruction obtained by the system appears later, in the gure 5 and 6.

Figure 4: Sign with kalman lter

MOTION RECONSTRUCTION
2 http:

Figure 5: Sign with kalman lter

For the motion reconstruction, I use the program Processing 1 in which the reconstruction of the
1 http:

// www.opengl.org/

// processing.org/

Figure 6: Motion Reconstruction

acceleration data of motions contains less information than positional data I have shown that motion reconstruction is possible in many cases using the data of surprisingly few accelerometers. My current implementation that is based on single threaded openGL code is not fast enough (on current standard PCs) to meet this goal, but also not too far away: I presume that code optimizations (and porting code to C++) combined with parallelizations exploiting the trend to multi-core CPUs will achieve this goal in the near future.

In the results obtained in Motion Recon- References struction Using Sparse Accelerometer Data [JOCHEN, 1], and the use of accelerometers Xsens [1] JOCHEN TAUTGES,ARNO ZINKE,GfaR mbH,BJORN KRUGER,JAN BAUMANN and system 3 , is clear to say that these sensors award a ANDREAS WEBER,THOMAS major presicion in the motion capture, due to the HELTEN,MEINARD high quality instrumentation. But in this work I MULLER,HANS-PETER SEIDEL,BERND use low cost sensors with easy acquisition, this does EBERHARDT, Motion Reconstruction Using not mean that the sensors are poor, its just necesSparse Accelerometer Data. 2010 sary to implement an instrumentation between the sensors and the stage of acquisition as shows in the [2] Riccardo Antonello, Ilaria Nogarole and gure 6. Roberto Oboe Department of Management and Engineering University of Padova, Motion reconstruction with a low-cost MEMS IMU for the automation of human operated specimen manipulation. 2007 [3] starlino, A Guide To using IMU (Accelerometer and Gyroscope Devices) in Embedded Applications. December 29, 2009 [4] Simon Haykin, ADAPTATIVE FILTER THEORY. 1986 by Prentice-Hall Figure 7: Diagram Blocks [5] Walter Mora F.,Alex ander Borb on A, DE TEXTOS CIENTIFICOS. 2da EDICION edici on, 2012. [6] Paul D. Groves, Principles of GNSS, Inertial, and Multi-Sensor Integrated Navigation Systems (GNSS Technology and Applications).

CONCLUSION AND FUTURE WORK

The approach of this paper is related to the way [7] ing-Xun Lai , Lei Shu , Athanasios Vasilakos , Joel J. P. C. Rodrigues , of making the motion, using sensors and systems Chin-Feng Lai and Yueh-Min Huang, 3D of aquisicion of signs of low costs. It is necessary Adaptive Reconstruction of Human Motion to clarify that the work is based on the 3D reconFrom Multi-Sensors. struction of the arm. I only focus in the recontruccion of the forearm because I only have a IMU and an Acceleromete. According to consulted papers implementation of accelerometer and gyroscope is necessary to the reconstruction of the movement of the arm and the forearm [Xun Lai,7]. Although
3 http:

// www.xsens.com/

Das könnte Ihnen auch gefallen