Experimental
Motion Capture Systems
Development

Kathleen Ruiz
URP Advisor
Assistant Professor of Multidisciplinary Electronic Art and
Information Technology in the Arts

Rensselaer Polytechnic Institute
ruiz@rpi.edu
518-276-2539

Main Researchers:

Michael Kingsley
kingsm@rpi.edu
518-273-7102

Dan Berlin
berlid@rpi.edu

test data
from
The AVA Project
(4.52mb QuickTime .mov)

Abstract

Motion capture is the process of recording a live motion event and translating it into usable mathematical terms by tracking a number of key points in space over time and combining them to obtain a single 3 dimensional representation of the performance. 1 In working on The AVA Project's motion capture at U.C. Irvine's Motion Capture Studio with Lisa Naugle, Ph.D., Professor of Dance & Director in The Claire Trevor School of the Arts, and Frederic Bevilacqua, Ph.D., Post Doctoral Research Fellow at the Beekman Laser Institute, I was struck by the wonder of motion capture. It was a direct link from the real to the iimaginary. I had worked with motion tracking previously for test navigation in on line virtual environments, so motion capture was a logical next step. In discussing the process and the data structures with my students, one group (consisting of Mike Kingsley and Dan Berlin) was particularly inspired, and so much so that they are now creating a new motion capture system which, it is hoped, will ultimately surpass existing systems which have numerous flaws and difficulties.

Motion capture is used in many fields:

- Biomedical research into physiological movement and gait analysis.
- Sports movement analysis, such as correct form in a baseball or golf swing.
- Motion picture industry to provide lifelike movement to computer generated characters.
- Live and recorded artistic performances.
- Interactive virtual reality training simulations.
- Interactive computer games.

Unfortunately, most motion capture systems on today's market are prohibitively expensive for educational institutions and small businesses. Our goal is to develop a relatively low-cost competitive motion capture system for the RPI community thus opening doors to advanced levels of transdisciplinary research which could include artistic animation and biomedical three dimensional analyses. To be competitive with current technologies, this system should have a sampling rate of approximately 100 Hz and a resolution of less than 0.1 inch.

The goal of our research is to produce a system that allows real-time tracking of an unlimited number of key points with no space limitations at the highest frequency possible with the smallest margin of error. There are three methods currently used throughout the industry to accomplish performance level motion capture tasks: optical, magnetic, and mechanical. We are proposing a fourth, Range Finding using radio frequency or ultrasonic waves. Other methods considered for use in the motion capture device were mechanical potentiometers and magnetic systems but these were deemed too unreliable for use in our situation. RF and ultrasonics are very similar in theory and will both be looked at as possible system choices.

History

The history of motion capture can be traced back to the late 1800's when Etienne Jules Marey and Eadweard Muybridge performed motion studies of various animals and humans. These studies offered artists and physiologists their first glimpse at the very essence of what motion is.

Etienne Jules Marey's motion capture suit

The goal of all motion capture research is to produce a system that allows real-time tracking of an unlimited number of key points with no space limitations at the highest frequency possible with the smallest margin of error.

Muybridge's photographic studies of animals in motion

There are three methods currently used throughout the industry to accomplish performance level motion capture tasks: optical, magnetic, and mechanical. We are proposing a fourth, Range Finding using radio frequency or ultrasonic waves. Other methods considered for use in the motion capture device were mechanical potentiometers and magnetic systems but these were deemed too unreliable for use in our situation. RF and ultrasonics are very similar in theory and will both be looked at as possible system choices.

Optical -

Motion Analysis Eagle Digital System (Optical) 4

Optical motion capture employs between 4 and 32 cameras to film the actor performing the necessary movements at speeds from 30 to 1000 frames per second. The actor wears small reflective markers on their clothes to provide the contrast necessary for the processing software to "see" and record the movement of the points. When all the different camera inputs are combined and processed, 3 dimensional data representing the points is produced. Optical systems have problems when markers become occluded from the camera. In addition to this a tremendous amount of processing and time is required to clean the data enough to make it usable for motion capture applications. Optical motion capture systems can cost between $100,000 and $250,000.

Electromagnetic -

Ascension Technology Corp. Motionstar Wireless (Magnetic) 5

Electromagnetic systems produce low-frequency electromagnetic fields which are picked up by sensors on the motion capture actor. The sensors can then report their movement and orientation based on the field to a central processing unit. This allows for near real-time processing and performances. Magnetic data is usually fairly clean as compared to other systems. However, magnetic systems are adversely affected by even small amounts of electromagnetic noise as well as the presence of ferrous metals in the vicinity. Electromagnetic system prices vary from $5,000 to $150,000.

Electromechanical -

Metamotion Gypsy (Mechanical)6

Electromechanical suits resemble large exoskeletons composed of variable potentiometers which measure joint angles and extension. Because of the relatively quick analog to digital post-processing, this system can be used in real time performances. These systems free the actor of electromagnetic interference, but are often heavy and encumbering which is very problematic in performance applications. Potentiometers also have a finite lifetime, and become electronically "noisy" as they get older, resulting in worse and worse data. Pricing for electromechanical systems is approximately $5,000 to $10,000.

Radio Frequency (proposed) - The system we propose is to use radio frequency range finding. The radio frequency method would work in a similar manner as an E-ZPass tag or Mobil Speedpass wand. Each of those units carries in it a Radio Frequency transponder which, when interrogated by a transmitter or interrogator, returns certain information.
As a motion capture device, a simple RF chip could be interrogated to just return a unique set of pulses, ideally, its identity within the skeleton (i.e. left ankle, right hand, etc.). The time that it takes for the interrogator to receive the ping is twice the time it took the original signal to get to the chip. The RF speed in air is known, so the distance from the interrogator to the RF chip can be calculated. If the distance is known along 3 axes, a definitive point in 3d space can be found. In this image, each of the three spheres represents the arbitrary distance found using the RF range finding X, Y, and Z. They intersect at a single point in space, shown as an asterix, thus producing a 3 dimensional coordinate for the point. This should
enable a fairly simple algorithm to produce near real-time 3 dimensional results for position and movement.

Anticipated advantages of the RF system:
* Low cost relative to other systems. A small transponder is approx. $10.
* High predicted sampling rate
* High predicted resolution.
* Small size and portability
* Minimal calibration
* RF penetrates most materials, preventing wave occlusion problems.

Anticipated problems of the RF system:
* Electromagnetic interference from outside sources.
* Transmission speed too fast for calculation.
* Hardware unknowns.
* Doppler and phase shift irregularities in data.

Sonic / Ultrasonic (proposed) - This system is very similar to the RF system, only instead of using the electromagnetic end of the spectrum, we're using sonic and ultrasonic pulses in the range finding phase. Again, information would be transmitted by a string of pulses.

Anticipated advantages of the Sonic/Ultrasonic system:
* Another low-cost alternative. An ultrasonic transducer is about $10.
* Sonic waves are slow enough to easily work with in electronics, and are commonly used as rangefinders.
* Complete elimination of the electromagnetic noise problem. Ultrasonics are high pitched enough that normal background noise does no enter that end of the spectrum.
* Minimal calibration.
* Small size and portability.

Anticipated problems of the Sonic/Ultrasonic system:
* Sonics and ultrasonics are not omni-directional, which leads to possible wave occlusion and aiming problems.
* Sonic waves are much more prone to reflecting in a studio setting.
* Because the wave is slower, Doppler and phase shift will be a much more pronounced problem.
* Ultrasonics may be too slow a medium to achieve the sampling and resolution numbers we are looking for.

Methodology:
Initially the team will try to construct a 1 dimensional prototype to prove that the system works. This prototype would compose of a transponder, an interrogator and the other hardware and software necessary to interpret the output.
The first goal with this prototype is to be able to accurately measure non moving distances between the transponder and receiver on the X axis. The next step is the recording of movement along this fixed axis. Once we get the system to work in this capacity, theoretically, the system is proven however more testing is needed. Next, following along with this progression, another interrogator axis Y is added and used to identify a single point in a 2D plane. Again, once this is accomplished, movement in that plane would be tested. Finally after these tests, a third interrogator axis Z would be added and immobile placement in 3 dimensions tested. Once this checks out okay, we test movement of a point in 3D space. Once a point can be followed in 3D space, we systematically add more points to see where the limits of the system lie.

Related Project Research Media

Conclusion:

Research in the field of motion capture devices could allow members of the RPI Biomedical community to conduct motion and physiological studies, which could lead to advances in ergonomics, design, and safety. Members of the artistic community can utilize the technology to create animations, and live artistic presentations.

Related Links and Research

1. Menache, A., Understanding Motion Capture for Computer Animation and Video Games, Morgan Kaufman, © 1995.
2.Image credit - "kimface2," Siggraph 2002. 6 September 2002, http://www.motionanalysis.com/about_mac/siggraph2002.html
3.Image credit - "kick2," MotionStar Wireless in Action!, 6 September 2002,
4.http://www.ascension-tech.com/products/mswireless/
5.Image credit - "A_sm," Gypsy 3 motion capture system, 6 September 2002, http://www.metamotion.com/gypsy/gypsy-motion-capture-system.htm
6.Image credit – “A_sm,” Gypsy 3 motion capture system, 6 September 2002, http://www.metamotion.com/gypsy/gypsy-motion-capture-system.htm