A WIRELESS BODY-WEARABLE SENSOR SYSTEM FOR

DESIGNING PHYSICALLY INTERACTIVE VIDEO GAMES

Bobak Mortazavi, Hagop Hagopian, Jonathan Woodbridge, Behrooz Yadegar and Majid Sarrafzadeh

Computer Science Department, University of California, Los Angeles, CA 90095, U.S.A.

Keywords:

eHealth, Exercise, Obesity, Wireless Systems, Games for health.

Abstract:

This paper presents a novel wireless body-wearable sensor system to control video games, making the user

more active while gaming. Our system uses data gathered from accelerometers and pressure sensors worn by

the player, to detect speciﬁc movements, which are used as in-game motion controls. Our system is designed

to detect different types of motions, and with proper integration into any game, will ensure that the player

will constantly remain active. To demonstrate, our system was adapted to a popular soccer game that uses

the player’s various leg motions as replacements for keyboard strokes. The result is a higher heart rate and

calorie burn for the user, making the overall experience signiﬁcantly healthier and more active than analogous

sedentary video games.

1 INTRODUCTION

Video games continue to be an ever-growing and

tremendous source of revenue for companies, and en-

tertainment for people. In the past year alone, mil-

lions of games have been sold and millions of hours

of time have been spent playing them. In January

2010, Activision Blizzard Inc. reported that Call of

Duty: Modern Warfare 2 had earned more than $1 bil-

lion in worldwide sales since its release that previous

November, and earning more than half of that in only

its ﬁrst ﬁve days of availability (Activision Blizzard

Inc., 2010). Likewise, Electronic Arts Inc.’s FIFA

Soccer 10 had already sold over 4.5 million units,

making it the fastest selling sports video game in the

world; also averaging three million games played on-

line per day(Electronic Arts Inc., 2009).

With a multitude of hours spent playing video

games and watching television, questions have

emerged with respect to the obesity level in the coun-

try, speciﬁcally children, and the contribution that

sedentary activities have. This obesity level in the

United States manifests itself in many ways, includ-

ing the increase in adult waist size from 1998 to

2004 by 40.5% (Li et al., 2007). Projections of the

year 2030 places approximately 90% of all American

adults as overweight or obese, of which at least half of

those would be classiﬁed as obese (Wang et al., 2008).

This trend is not limited, however, to adults; children

are increasingly becoming overweight and obese as

a result of further increased sedentary behavior asso-

ciated with television watching and video game play-

ing (Robinson, 1999)(Stettler et al., 2004)(Rey-Lopez

et al., 2008). Speciﬁcally with regards to video game

use, the prevalence of obesity increased from around

5% in children spending no time playing video games

to over 20% when playing three hours per day (Stet-

tler et al., 2004).

The potential to target video game use and de-

velop active and healthy video games is a ﬁeld with

great potential. The use of sensor networks worn on

the body as input devices has achieved success even

in simple pose-based game control systems (White-

head et al., 2007). The Nintendo Wii’s hand-held

accelerometer-based motion controller has led to 70.9

million units in lifetime sales(Nintendo Co. Ltd.,

2010). Wii Fit, a game that comes with a balance

board and various games and exercises targeted at im-

proving ﬁtness, has sold 21.82 million units since its

release(Nintendo Co. Ltd., 2009). This result illus-

trates the growing popularity and potential for healthy

video games, especially those that can induce exer-

cise activity (Brown, 2006). However, the Wii remote

is limited to only hand-held activity, and the balance

board can only sense weight. In fact, activity on the

Wii can be generated with little to no movement if

desired. Most games, including those on the Wii, use

gesture based movements, stance based movements

and a few use continuous activity monitoring (Stach

et al., 2009).

Mortazavi B., Hagopian H., Wodbridge J., Yadegar B. and Sarrafzadeh M..

A WIRELESS BODY-WEARABLE SENSOR SYSTEM FOR DESIGNING PHYSICALLY INTERACTIVE VIDEO GAMES.

DOI: 10.5220/0003159100620069

In Proceedings of the International Conference on Biomedical Electronics and Devices (BIODEVICES-2011), pages 62-69

ISBN: 978-989-8425-37-9

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

In this paper, we will show that the use of a body-

wearable sensor network can help achievevideo game

control to promote actual exercise activity. Using

sensors to capture gestures(Jung and Cha, 2010) or

pose-based movements (Whitehead et al., 2007) have

shown that monitoring of basic controls can be per-

formed. Our system, however, will take this a step

further and monitor not only continuous active con-

trols but the strength of these movements as well.

Sports games, in particular, are a well-suited target

for such a system, with previous work in sports games

(Mueller et al., 2007) and general activity recognition

for such sports applications (Avci et al., 2010) having

been done. Such exercise games are not only limited

in scope, but restricted by the platform and periph-

erals needed to play. For example, some systems re-

quire an exercise bike (Mokka et al., 2003), while oth-

ers require using a soccer ball and camera to capture

only the ball movement and not the player movement

(Mueller et al., 2007). Indeed, the number of conti-

nous activity monitoring systems applicable to a wide

range of games is quite limited (Stach et al., 2009)

and our system targets this area. The system we have

designed allows for the body to become an active and

continuous motion controller and we show how such a

controller can be adapted to games, particularly sports

games; such a system can actually promote physical

health through exercise.

The rest of the paper is organized as follows. Sec-

tion 2 gives a brief description of related work in

terms of body-wearable sensor networks and the po-

tential health beneﬁts of targeting television and video

gaming. Section 3 contains our system design along

with its use as a motion controller for a popular soc-

cer game. Section 4 summarizess the health beneﬁts

of our system. Section 5 discusses future work and

Section 6 concludes our study.

2 RELATED WORK

Several research projects extend beyond the limita-

tions of gesture or stance based body control for

games and extend to body-sensor networks for health

promotion. Each project takes a unique approach to

help target a speciﬁc health goal while our system ex-

tends beyond a single goal and can work for a multi-

tude of situations.

2.1 No Pain no Game

No Pain No Game is a home automation system

developed to monitor the physical activity of chil-

dren throughout the day in order to regulate the

time spent watching television or playing electronic

games (Hsiao et al., 2010). The system uses

an accelerometer-based pedometer to calculate the

Metabolic Equivalent of Task (MET) in order to mea-

sure activity. It then uses this data to control power

outlets and manage electronic activities such as tele-

vision viewing or video gaming. No Pain No Game

attempts to address and limit the time spent playing

video games, while our work makes that spent time

healthier.

2.2 Games for Stroke Rehabilitation

A collaboration between Washington University in St.

Louis and University of California, San Diego pro-

duced a game customized to help stroke rehabilita-

tion through the use of a tracking web-camera and on-

body accelerometers built into Wii remotes (Alankus

et al., 2010). This is a prime example of a class of

games targeted to a speciﬁc subset of controls. Multi-

ple sensors were used in this project to monitor single

or multiple muscles for associated movements specif-

ically related to stroke therapy exercises. By placing

the Wii-mote in speciﬁc areas around the arm, they

are able to monitor shoulder, elbow, or wrist move-

ments and exercise, and use these inputs for games

targeted to stroke rehabilitation exercises. Our system

improves on this idea and extends the control from a

subset of exercises to a range of motions for various

activities.

2.3 Health Games

Research by Thompson et. al. shows a different di-

rection in the healthy video game domain by targeting

the game design instead of the game control, in mak-

ing a game speciﬁcally built to address Type 2 Dia-

betes (Thompson et al., 2008). Their goal was to build

a fun game whose content addressed dietary issues

and diabetes understanding. Their game is played

with a standard keyboard and their work focuses on

the story and game content. The game is intended to

educate the user, but the user remains sedentary while

playing.

2.4 Dance Dance Revolution

The best example of a popular video game being

used for physical activity is Dance Dance Revolution

(DDR), the popular dancing game by Konami. The

game requires the use of a dance pad to sense pres-

sure in speciﬁc directions and can double the heart

rate of the user to exercise levels (Brown, 2006).

Further studies on DDR have shown that promoting

A WIRELESS BODY-WEARABLE SENSOR SYSTEM FOR DESIGNING PHYSICALLY INTERACTIVE VIDEO

GAMES

DDR leads to an increase in vigorous physical activ-

ity (Maloney et al., 2008), which further indicates the

ability to promote video games that induce physical

activity without having to specially tailor the game.

Again, DDR along with Wii Fit are limited in scope

due to their need of a physical pad as the control

mechanism.

3 SYSTEM DESCRIPTION

Our body-wearable video game controller allows for,

and in fact enforces, physical activity at exercise lev-

els that can be adaptable to a wide range of video

games. While sedentary activities, such as video

games, have previously been linked to the cause of

obesity amongst children and adults, this system will

actually turn such activities into entertaining exercise,

promoting the health beneﬁts of such games by using

the human body as an active controller. Our system

incorporates tri-axial accelerometers attached to the

hands and feet to monitor movement, pressure sen-

sors to detect standing and running, and software al-

gorithms to further classify and interface with any

PC-based game. In this paper, the system has been

adapted to the FIFA Soccer 10 PC game.

3.1 Hardware

With the use of low-cost commodity hardware, we

are able to enact the desired controls necessary for

the soccer game. We use two Analog Devices 3-Axis

Accelerometers (ADXL335) (Analog Devices, 2010)

attached to an Texas Instruments’ MSP430 (Texas In-

struments, 2010) development board shown in Fig-

ure 1. The MSP430 Development board allows us

to communicate the accelerometer data wirelessly to

the PC. Also, attached to the MSP430 is a square-

shaped force sensitive resistor for pressure sensing.

Each of these sensors is strapped to a foot, allowing

for two footed control of the players within the game.

Three actions are detected for the primary foot. The

ﬁrst is a forward shooting motion in the y-axis direc-

tion of the accelerometer. The second, primarily in

the x-axis direction, is the passing motion, while the

z-axis helps determine strength and detect running.

The accelerometer on the secondary foot can deter-

mine crossing, through passing, and running. All of

these actions map directly to the actions deﬁned by

FIFA Soccer 10. A Nintendo Wii-Remote is used as a

wireless direction pad for easy turning of players.

The Wii-Remote is used only to indicate the di-

rection of the running movement generated by the ac-

celerometers and pressure sensors, and can easily be

Figure 1: Accelerometer + Pressure Sensor Hardware Unit.

replaced by a gyroscope in future designs.

3.2 Software

The software algorithm, written in the C# language,

developed in Microsoft Visual Studio, helps further

classify the continuous actions and consists of four

primary components. A ﬂowchart of the algorithm

is shown in Figure 3. This algorithm detects the

trained movements and constantly calculates health

information based upon sensor readings and consists

of ﬁve primary pieces: calibration, segmentation, sup-

port vector machine classiﬁcation, movement deci-

sion, and calculation of health statistics.

3.2.1 Calibration Phase

The top of a foot is not generally ﬂat and the angle of

the y-axis, the rotation around the x-axis (pitch), and

the rotation around the y-axis (roll) varies from person

to person and must be accounted for. Since we use a

strap, visible in Figure 2 to place the device ﬁrmly

on top of the foot, rotation around the z-axis (yaw),

is not considered a factor in affecting motion anal-

ysis, but can be accommodated in a similar fashion

as the other two angles of rotation. Figure 2 shows

the orientation of the sensor and the desired ﬂat di-

rections that the readings will be adjusted to. Since

we know the standard ﬂat position of the accelerome-

ter, we can obtain a reading and normalize its values

to that of 0 acceleration in the y-direction and 0 ac-

celeration in the x-direction with an acceleration in

the z-direction counteracting gravity. Using this as an

initial vector, our algorithm can begin by taking an

initial set of readings, while asking the user to remain

still brieﬂy, to determine the acceleration readings due

to the position of the sensor on the foot and associated

force of gravity.

BIODEVICES 2011 - International Conference on Biomedical Electronics and Devices

Figure 2: Hardware unit attached to foot. Bold lines are the sensor’s coordinate frame of reference. Dashed Lines show the

calibrated and rotate frame of reference. Angle θ is the pitch, or rotation around the x-axis, and Angle γ is the roll, or rotation

around the y-axis.

Start

Read

acceleration +

pressure

values

Filter +

Interpolate

Apply Rotation

Matrix

Apply

Segmentation

Initial Movement

Prediction + SVM

Vector Creation

user quit?

Stop

Yes

Apply SVM

Remove False

Positives (Using

Pressure Values)

Classify Speciﬁc

Movement

Determine

Strength

Update Health

Statistics

Apply

Movement

Figure 3: Flowchart of our Algorithm.

cos(θ) =

v1·

v1kk

v2k

(1)

Equation 1 is a well-known equation for ﬁnding the

angle between two vectors

v1 and

v2, where

v1 ·

is the dot product of the two vectors and k

v1k is the

magnitude of a vector.

θ = arccos

0∗y + g∗z

+ g

+ z

(2)

γ = arccos

0∗x + g∗z

+ g

√

+ z

(3)

Given we have normalized our standard coordi-

nates in the ﬂat direction, namely 0 in the y direction,

0 in the x direction, and g, the effect of gravity, in

the z direction, we can ﬁnd the pitch angle and the

roll angle by equations 2 and 3, respectively. Please

note in 2 and 3 that x y and z refer to the x, y, and z

components of the angled vector, while our base vec-

tor is of the form < 0,0,g >. We then use a simple

rotation matrix from linear algebra to convert every

3-point < x, y, z > vector back into our normalized

space < x

′

> to classify a set of movements with

better accuracy by no longer being dependent on the

person and the exact position of the accelerometers.

The rotation of the coordinate system can be solved

via equations 4 and 5, where 4 allows us to calcu-

late rotation in the opposite direction, as was deemed

simpler in this case:

θ = 2π−θ ,

γ = 2π −γ (4)

A =

cos(

γ) 0 −sin(

γ)

−sin(

θ)sin(

γ) cos(

θ) −sin(

θ)cos(

γ)

sin(

γ) cos(

θ)sin(

γ) cos(

θ)cos(

γ)

A WIRELESS BODY-WEARABLE SENSOR SYSTEM FOR DESIGNING PHYSICALLY INTERACTIVE VIDEO

GAMES

Figure 4: Y-Acceleration Samples plotted against normal-

ized amplitude of 8-bit ADC Channel. 1, 2, 3, and 4 corre-

spond to the segments created by the segment state machine.

A·













′





(5)

Once all points are rotated back into the standard

plain, the algorithm can begin its initial movement

classiﬁcation.

3.2.2 Segmentation + SVM

A segment state machine is built to identify changes

in slope of the acceleration values. These segments

help identify the potential for the standard kick, pass,

cross, through pass, and running movements. Each

movement corresponds with a standard slope pattern

in the acceleration curve. The movements segment

into four speciﬁc sectors of a sinusoid similar to that

in Figure 4. The sensors broadcast at 100 Hz and ad-

just the slope according to a sliding window of twenty

(20) < x, y, z, pressure > samples read. This allows

for some history to indicate a trend, but is not pro-

hibitively large enough to introduce a delay in the ac-

tions of the user. This segmentation, adapted from

an initial segmenting system presented in (Hagopian,

2010), can accurately determine the slopes of accel-

eration curves and the associated state machine can

predict potential movements based on a pattern of ac-

celeration and deceleration. Figure 4 shows a kick

and the state machine’s corresponding segmentation.

The state machine itself can classify a large set of ac-

tions but for movements that are not entirely dominant

in any particular direction, a more precise pattern-

matcher can be used for the boundary cases, like a

support vector machine.

A support vector machine (SVM), such as the

open source libsvm (Chang and Lin, 2001) that we

used, allows for multi-class classiﬁcation based on ex-

tracted features from the segment state machine. The

SVM, of course, is trained with the movements in a

prior phase, but the calibration reduces the need to re-

train the segment state machine or the SVM for each

user. The test vectors used for the SVM in our al-

gorithm consists of ﬁfty (50) points, each containing

four values: x, y, z, and pressure.

3.2.3 Movement Decision and Health Features

Once the classiﬁcation is accomplished, a ﬁnal ﬁlter is

run based upon the pressure proﬁles and health statis-

tics are updated along with the movement being gen-

erated and delivered to the game. The health features

are based upon the idea of the Metabolic Equivalent

of Task (MET), in order to determine the activity level

of the user. Different sports have different MET num-

bers resulting in different calorie burning. MET is

expressed versus the cost of resting metabolic rates;

therefore, it measures speciﬁcally based upon the in-

crease in activity, where 1 METs is considered be-

ing at rest (sedentary), while 3 METs is walking and

above 6 is more vigorous activity (Ainsworth et al.,

1993). Based upon the mass the user must input

to start the system, and the level of activity of the

accelerometer+pedometer, equation 6(Hsiao et al.,

2010) helps calculate the speciﬁc number of calories

burned by the user.

Calories =

MET ∗3.5∗m

200

∗t (6)

In equation 6, m is the mass in kilograms and t is the

total duration in minutes. This calorie information,

along with the pedometer information based upon the

pressure sensor, give a basic set of health information

to the user as he/she is using our system. While those

health statistics are updated, the movements are de-

cided. The strength of these movements can be calcu-

lated by various heuristics. In this case, the heuristic

is the magnitude of the absolute sum of activity ac-

cording to:

k ~movek=

∑

i=0

|move

| (7)

, where ~move is the movement vector, l is the length

of ~move, and move

is the i

value of the vector. FIFA

Soccer 10 allows a wide range of strengths for kicks,

crosses, and a differentiation between standard run-

ning and sprinting. No movement of the player will

occur if the user is not at least stepping in place. All

movements are based upon actual human movements.

For example, running in place faster will allow the

player to sprint, soft swings of the leg will gener-

ate soft kicks while stronger swings generate harder

shots, and wild kicks by the user cause the player in

FIFA Soccer 10 to lose control of the ball by over-

powering the shot. Our system is able to continu-

ously monitor not only the movements generated by

the body but the strength of those movements in or-

der to more properly determine the activity level MET

and the actions within an associated game.

BIODEVICES 2011 - International Conference on Biomedical Electronics and Devices

Figure 5: Pressure value and Y-acceleration samples plotted against normalized amplitude of 8-Bit ADC Channel. Points A

and B correspond to the decision point of the system on an action.

3.3 Accelerometer Vibration Issues and

Enforcing Activity

Running in place is of particular concern as it

can cause sensor vibration, which even the pattern-

matching algorithms like an SVM can mistake for

movement. As a result, the pressure proﬁle of the user

becomes a signiﬁcant portion of the decision mak-

ing process for movements in order to eliminate false

positives as a result of acceleration vibration. The

pressure proﬁle must work in association with the ac-

celerometers in order to more accurately detect move-

ment as shown in Figure 5. In Figure 5, the left

graph at point A detects a proper kick. When pressure

is lifted, the sensor peaks to its maximum amplitude,

showing the foot is in the air while the kick occurs.

The right graph, at point B, shows a step misrepre-

sented as a kick. With the addition of the pressure

sensor, however, the system can now determine that

at the classiﬁcation point, the foot is pressed on the

ground, and thus, appropriately designates this only

as a step and not as a kick. This sensor fusion is

a quick and effective way of dealing with a signiﬁ-

cant problem with accelerometers when strapped to

the body. Our system, with the addition of the pres-

sure sensors under the feet, is able to eliminate what

might otherwise be mistaken as movements in simple

accelerometer-only based systems.

The important contribution to this system is the

ability to control video games with the body in an

active and continuous fashion in order to promote

healthier game play. As a result, a primary component

to our system is, in fact, to enforce that the user be ac-

tively playing the game. As the algorithm runs, if at

any point the pressure proﬁle decreases from what is

expected as a standing person, it will pause and alert

the user that he/she is no longer standing and actively

playing. The algorithm allows for the pressure to be

lifted but these intervals of reduced pressure are mon-

itored in order to guarantee actions are occurring and

the user is not, instead, attempting to circumvent the

necessary activity levels of the system. As a result, we

have a simple but effective manner to ensure the user

is not sedentary and is instead up and about actively

moving and, hence, exercising.

4 HEALTH BENEFITS

FIFA Soccer 10 was a suitable choice for our system

because it allows for varying speeds of running and

different leg movements for different actions; it has

supported over 113 million total online games played

(Electronic Arts Inc., 2009) and serves as a good ex-

ample of a popular game that can be made active and

healthy with our system. Our FIFA Soccer 10 system

allows the user to run, sprint, kick and pass with one

foot, and cross and through pass with the other. One

can use this system to control any action in the game

and can even take it online to challenge others. Ta-

ble 1, which shows the average of ﬁve (5) users play-

ing with our system over a ﬁfteen (15) minute period,

compares the level of exercise of various activities.

Heart rate is based on the speciﬁc user, calorie burn

is based on equation 6 for a 160 lb individual, MET

for walking and soccer is a known value (Ainsworth

et al., 1993), and MET for FIFA Soccer 10 PC and

our system is based on activity level based on system

equation 7. As indicated in the table, FIFA Soccer

10 as played with a standard keyboard is completely

sedentary, whereas our system performs on par with

moderate exercise. As a point of reference, actual soc-

cer is also listed.

A WIRELESS BODY-WEARABLE SENSOR SYSTEM FOR DESIGNING PHYSICALLY INTERACTIVE VIDEO

GAMES

Table 1: Table showing approximate calorie burn and heart

rate (beats per minute) associated with 5 users playing FIFA

Soccer 10 on PC with keyboard, with our system, and a

comparison to walking 4.5 miles per hour and playing ca-

sual soccer (Ainsworth et al., 1993) over a 15 minute period.

Activity Heart Rate MET Calories

FIFA 10 (PC) 73 bpm 1.0 19

Walking – 4.5 85.7

FIFA 10 (Active) 144 bpm 4.5 85.7

Soccer – 7.0 133.3

5 FUTURE DIRECTIONS

AND DISCUSSION

In the current form of our controller system, we can

monitor movements and enforce physical activity and

adapt it to a wide range of games. Although we have

shown how, with the devices strapped to the user’s

feet, one can play a popular and widely available soc-

cer game, our system is not solely limited to these

actions. One can attach the devices to other parts of

the body, such as the hands, and generate movements

from there as well. The general structure of the sys-

tem remains the same with a re-training of the clas-

siﬁcation portion of the algorithm and an adjustment

to the calibration phase being the only pieces that re-

quire attention. Indeed, our system is adaptable and

ﬂexible to many different end applications and lends

itself to further user studies on human activity moni-

toring.

Additional sensors can be added to make the sys-

tem more ﬂexible if needed. As we further monitor

the health status of users with this system, it becomes

apparent that the addition of human vital signs, such

as heart rate and respiration are features that can be

added not only to monitor health but to affect future

games. For example, one may become fatigued in

real life, and by monitoring this, our system can tell

the game to make their characters also show fatigue.

Vibration and sounds may be added to enhance feed-

back to the body to allow for better control. More

complex movements, such as rotations measured by a

gyroscope, can also be monitored by the addition of

hardware into our system. In fact, with the announce-

ment of Microsoft Corporation’s Kinect (Microsoft,

2010) camera-based movement system, these addi-

tions would keep our system at the forefront of body-

controlled games. In cases where camera systems

may only visualize moves and approximate health in-

formation, our system will be able to continue gaining

accurate results and provide feedback to the user for

more realistic behavior and feel.

6 CONCLUSIONS

The obese and overweight population of the United

States, and most of the industrialized nations, is in-

creasing. As the population grows and the costs be-

come more prohibitive to health-care, new and alter-

native ways to attack sedentary behavior must be de-

veloped. Television viewing and video game playing

are time-consuming processes in which users are rel-

atively, if not completely, inactive. Research projects

have attempted to address this inactivity by limiting

the television time, or by restricting movements to a

small subset, but we have developed a system with

which players of video games may use their bodies as

input devices to the games they are already playing.

This system will ensure that the video game player

will no longer be in a prolonged sedentary state, but

instead, has to exercise to a point where the game

becomes beneﬁcial. Through the use of accelerom-

eters and pressure sensors, we can enable active and

continuous motion control and activity level monitor-

ing. With a clever classiﬁcation algorithm we can

mimic the input behaviors necessary to perform re-

alistic actions as a body-controller for a video game,

and at the same time, measure and monitor activity

to ensure a physically motivated game playing expe-

rience in which calories are burned and the heart rate

climbs. This system is adaptable to current state-of-

the-art games by simply replacing any of the given in-

puts and can be used to keep users entertained while

making them active.

REFERENCES

Activision Blizzard Inc. (2010). Call of duty(r): Mod-

ern warfare(r) 2 surpasses $1 billion in retail sales

worldwide. http://investor.activision.com/ releasede-

tail.cfm?ReleaseID=437383.

Ainsworth, B. E., Haskell, W. L., Leon, A. S., Jacobs, D. R.,

Montoye, H. J., Sallis, J. F., and Paffenbarger, R. S.

(1993). Compendium of physical activities: classi-

ﬁcation of energy costs of human physical activities.

Medicine & Science in Sports & Exercise, 25(1).

Alankus, G., Lazar, A., May, M., and Kelleher, C. (2010).

Towards customizable games for stroke rehabilitation.

In CHI ’10: Proceedings of the 28th international

conference on Human factors in computing systems,

pages 2113–2122, New York, NY, USA. ACM.

Analog Devices (2010). Adxl335: Small, low power,

3-axis 3 g accelerometer. http://www.analog.com/

en/sensors/inertial-sensors/adxl335/products/

product.html.

Avci, A., Bosch, S., Marin-Perianu, M., Marin-Perianu,

R. S., and Havinga, P. J. M. (2010). Activity recog-

nition using inertial sensing for healthcare, wellbeing

BIODEVICES 2011 - International Conference on Biomedical Electronics and Devices

and sports applications: A survey. In 23th Interna-

tional Conference on Architecture of Computing Sys-

tems, ARCS 2010, Hannover, Germany, pages 167–

176, Berlin. VDE Verlag.

Brown, D. (2006). Playing to win: Video games and the

ﬁght against obesity. Journal of the American Dietetic

Association, 106(2):188 – 189.

Chang, C.-C. and Lin, C.-J. (2001). LIBSVM: a library

for support vector machines. Software available at

http://www.csie.ntu.edu.tw/∼cjlin/libsvm.

Electronic Arts Inc. (2009). Fifa 10 continues record

sales pace; ﬁfa 10 ultimate team announced.

http://investor.ea.com/releasedetail.cfm?ReleaseID=

427414.

Hagopian, H. (2010). Health and gaming: Making the body

more active. Master’s thesis, University of California,

Los Angeles.

Hsiao, M.-C., Amini, N., and Sarrafzadeh, M. (2010). A

wireless home automation system for childhood obe-

sity prevention. In Proceedings of the International

Conference on Biomedical Electronics and Devices.

Jung, Y. and Cha, B. (2010). Gesture Recognition Based

on Motion Inertial Sensors for Ubiquitous Interactive

Game Contents. IETE Technical Review.

Li, C., Ford, E. S., McGuire, L. C., and Mokdad, A. H.

(2007). Increasing trends in waist circumference and

abdominal obesity among u.s. adults[ast]. Obesity,

15(1):216–216.

Maloney, A. E., Carter Bethea, T., Kelsey, K. S., Marks,

J. T., Paez, S., Rosenberg, A. M., Catellier, D. J.,

Hamer, R. M., and Sikich, L. (2008). A pilot of a video

game (ddr) to promote physical activity and decrease

sedentary screen time. Obesity, 16(9):2074–2080.

Microsoft (2010). kinect for xbox 360 is ofﬁcial

name of microsofts controller-free game device.

http://www.microsoft.com/presspass/features/2010/

jun10/06-13kinectintroduced.mspx.

Mokka, S., V¨a¨at¨anen, A., Heinil¨a, J., and V¨alkkynen, P.

(2003). Fitness computer game with a bodily user in-

terface. In ICEC ’03: Proceedings of the second in-

ternational conference on Entertainment computing,

pages 1–3, Pittsburgh, PA, USA. Carnegie Mellon

University.

Mueller, F., Stevens, G., Thorogood, A., O’Brien, S., and

Wulf, V. (2007). Sports over a distance. Personal and

Ubiquitous Computing, 11:633–645.

Nintendo Co. Ltd. (2009). Financial results brieﬁng for

the three-month period ending june 2009. http://

www.nintendo.co.jp/ir/pdf/2009/090731e.pdf.

Nintendo Co. Ltd. (2010). Consolidated ﬁnancial

statements. http://www.nintendo.co.jp/ir/pdf/2010/

100506e.pdf.

Rey-Lopez, J. P., Vicente-Rodrguez, G., Biosca, M., and

Moreno, L. A. (2008). Sedentary behaviour and obe-

sity development in children and adolescents. Nu-

trition, Metabolism and Cardiovascular Diseases,

18(3):242 – 251.

Robinson, T. N. (1999). Reducing Children’s Television

Viewing to Prevent Obesity: A Randomized Con-

trolled Trial. JAMA, 282(16):1561–1567.

Stach, T., Graham, T. C. N., Brehmer, M., and Hollatz,

A. (2009). Classifying input for active games. In

ACE ’09: Proceedings of the International Confer-

ence on Advances in Computer Enterntainment Tech-

nology, pages 379–382, New York, NY, USA. ACM.

Stettler, N., Signer, T. M., and Suter, P. M. (2004). Elec-

tronic games and environmental factors associated

with childhood obesity in switzerland[ast][ast]. Obe-

sity, 12(6):896–903.

Texas Instruments (2010). Msp43016-bit ultra-low power

mcus. http://focus.ti.com/mcu/docs/mcuprodover

view.tsp?sectionId=95&tabId=140&familyId=342&

DCMP=MCU other&HQS=Other+IL+msp430.

Thompson, D., Baranowski, T., Buday, R., Baranowski,

J., Thompson, V., Jago, R., and Juliano Grifﬁth, M.

(2008). Serious Video Games for Health: How Be-

havioral Science Guided the Design of a Game on Di-

abetes and Obesity. Simulation & Gaming.

Wang, Y., Beydoun, M. A., Liang, L., Caballero, B., and

Kumanyika, S. K. (2008). Will all americans be-

come overweight or obese? estimating the progres-

sion and cost of the us obesity epidemic. Obesity,

16(10):2323–2330.

Whitehead, A., Crampton, N., Fox, K., and Johnston, H.

(2007). Sensor networks as video game input devices.

In Future Play ’07: Proceedings of the 2007 confer-

ence on Future Play, pages 38–45, New York, NY,

USA. ACM.

A WIRELESS BODY-WEARABLE SENSOR SYSTEM FOR DESIGNING PHYSICALLY INTERACTIVE VIDEO

GAMES