Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Tennis skill assessment

Research Article
DOI: 10.1002/jst.112

Towards a wearable device for skill assessment and skill


acquisition of a tennis player during the first serve
Amin Ahmadi1,2, David Rowlands1 and Daniel Arthur James1,2,
1
Centre for Wireless Monitoring and Applications, Griffith University, Australia
2
Centre of Excellence for Applied Sport Science Research, Queensland Academy of Sport, Australia

In this article, the possibility of using wearable gyroscope sensors for skill
assessment and skill acquisition was investigated. Marker-based methods were Keywords:
. tennis
used initially to capture the fast rotational motions and simulate the outputs of . serve
gyroscope sensors. Utilizing the marker-based methods, the angular velocity of . assessment
the upper arm internal rotation, wrist flexion, and shoulder rotation were . skill
calculated for a range of athletes using the trajectory of Vicon markers with . inertial sensor
respect to the Plug-in Gait model during the first serve in tennis. Participants
from amateur to elite participated in this study. Thirty successful serves from
each participant were assessed. The results showed that the peak values of the
upper arm internal rotation, wrist flexion, and shoulder rotation just before
impact are indicative in classifying the participants’ skill level. It was shown
that all the three parameters, as well as the racquet head speed, increased as
the level of proficiency of the participants increased. A line (R2 5 0.89) was
fitted to the scatter data containing the upper arm internal rotation, wrist
flexion, and racquet head speed. The fit line is a function of upper arm rotation
and wrist flexion. The fit line can be used as a potential skill acquisition tool to
provide feedback on which variables (upper arm internal rotation, wrist
flexion, or shoulder rotation) need to be improved. The positions of three
gyroscope sensors to detect the same trends as those from the marker-based
methods were determined. Therefore, it is envisaged that gyroscope sensors
could be used for skill assessment and skill acquisition for a first tennis serve.
r 2010 John Wiley and Sons Asia Pte Ltd

1. INTRODUCTION assess the performance is based on the observation of an expert


person, such as a coach. However, there are two disadvantages
Evaluating the performance of athletes during competition or associated with this subjective method. First, since it is a sub-
even during training sessions has always been a hot topic jective method, different coaches could have slightly different
among coaches and sports scientists [1]. This is important, as ideas based upon their experience. Second, there are some fast
the correct evaluation feedback could result in enhancing the motions during an action that cannot be captured by human
performance of athletes. One common and traditional way to eyes. Therefore, the need for an objective method rather than
a subjective method was raised.
Videography was used by sports scientists to monitor and
*Griffith School of Engineering, Nathan Campus, Griffith University,
study the biomechanics of various actions, such as the tennis
170 Kessels Road, Nathan, QLD 4111, Australia. serve, to provide insight into physical activity levels associated
E-mail: d.james@griffith.edu.au with performance, as well as the skill-based technique involved

Sports Technol. 2009, 2, No. 3–4, 129–136 & 2010 John Wiley and Sons Asia Pte Ltd 129
Research Article A. Ahmadi, D. Rowlands and D. A. James

in the activity [1]. There are some disadvantages associated with gyroscopes was simulated using the marker-based methods. By
this method. One of the main disadvantages is that it is not using the Vicon standard Plug-in Gait marker placement (Vicon
possible to provide real-time feedback to the athletes, and in Motion Systems Ltd., Oxford, UK), four male tennis players
particular, tennis players during a training session, as tedious during the first tennis serve were assessed. One amateur, two
post-processing is required to extract and analyze the collected subelites, and one elite player were studied. In order to determine
data. This leads to use other technologies to monitor the the upper arm internal rotation, the MBVG method was used.
athletes during sporting activities. Inertial sensor technology as Other marker-based methods were used to determine the wrist
one of the growing technologies in the field of sports monitoring flexion and shoulder rotation. The upper arm internal rotation,
is becoming more popular, as it has some advantages over the wrist flexion, and shoulder rotation velocity were measured,
previous method. since they were reported as the main contributors for the tennis
Improvements in microelectronics and other microtech- serve after the maximum knee extension [11]. Also, the output
nologies have made it possible to take advantage of using from the developed methods were compared and found to be
miniaturized, light, inexpensive inertial sensors, including closely matched with those from the gyroscope sensors.
accelerometers and gyroscopes to capture and analyze the
movements of athletes during many sporting activities. For
instance, acceleration sensor technology has been used to 2. METHODS
analyze kinetic processes for golfers’ lateral swing [2] and in
swimming [3].Within many sporting applications, the sensors 2.1 Marker-Based Technology
are now used to measure and classify activity and effort
levels [4–5]. For instance, inertial sensors were employed to Four right-handed, male tennis players, including one
distinguish between amateur and subelite tennis players during amateur, two subelites, and one elite tennis player were studied
the first serve [6]. It is envisaged that these inertial sensors can in this experiment. The Vicon motion-capturing system using
be worn by athletes at all levels to monitor their performance the standard Plug-in Gait model was used, and optical markers
without hindering it. In spite of all the advantages, inertial were attached on the upper body of each participant. The place-
gyroscope sensors are not fast enough to capture fast-rate ment of markers, with respect to the standard Plug-in Gait
rotational motions. Therefore, a method to simulate the model, is shown in Figure 1a. Eight cameras were used to record
behavior of the gyroscope sensors is required. the data at 100 frames per second. The participants were to
Marker-based virtual gyroscopes (MBVG) were developed serve at a target region. If the serve was not inside the region, it
to overcome the drawback associated with gyroscopes to mea- would not count as a successful serve. This corresponded to the
sure the upper arm internal rotation [7]. The MBVG method area needed to serve the ball into the service box. Thirty success-
works with the help of optical monitoring motion capture sys- ful first serves were collected from each player for analysis. All
tems, such as Vicon (Vicon Motion Systems Ltd., Oxford, UK). the players used the same tennis racquet during the experiment.
The trajectory of three reflective markers, which were not in a Some marker-based algorithms were developed to calculate the
straight line in 3-D, were captured and passed to the MBVG upper arm internal rotation, wrist flexion, and shoulder rotation
algorithm to measure the rate of rotational motion. The main during the first tennis serve action.
purpose of the MBVG is to predict the maximum value, as well Upper arm internal rotation is one of the main con-
as the trends in rotational velocities when the inertial gyroscope tributors (54 per cent contribution) [11] to the forward speed of
is not able to measure the action due to high-speed motion. the racquet at impact during a first tennis serve [11]. Three
Also, the MBVG method can save a lot of time by eliminating markers on the right upper arm, including the right shoulder
the trial-and-error method to find out the best sensor position. (RSHO) marker, right elbow marker, and the right upper arm
Tennis is recognized as one of the most popular sports
around the world, as it is played at all levels, including socially
and professionally [8]. In order to win the match, or at least get
better results, athletes need to improve their ground strokes, as
well as their serves. According to Bahamonde [9], among
various strokes in tennis, the tennis serve is the most important
and critical stroke. It is also known that a fast serve can
dominant the game at the elite level [9–10]. Therefore, for a
tennis player to be more successful during the match, he/she
needs to master the serve action. In order to master the tennis
serve, it is important to recognize the main contributors to
produce the fast serve. According to Marshall and Elliott [11],
internal upper arm rotation, wrist flexion, and shoulder rota-
tion play critical roles in generatin the fast first serve from the
maximum knee flexion to hit the ball.
Figure 1. (a) Marker placement with respect to the Vicon Plug-In Gait
The aim of this article is show that gyroscope sensors as model, as well as the required vectors, to determine the wrist flexion
wearable devices can possibly be used for skill assessment and ƒ! ƒ! !
( V1 and V2 ) and shoulder rotation (U ). (b) Markers M1 and M2
acquisition during the tennis serve motion once the fast-enough attached at the sides of the head of the tennis racquet. Centre point C
gyroscopes are developed. In order to show this, the behavior of is also shown.

130 www.sportstechjournal.com & 2010 John Wiley and Sons Asia Pte Ltd Sports Technol. 2009, 2, No. 3–4, 129–136
Tennis skill assessment

marker, were used to measure the angular rotation and thus to marker point LSHO. Vector U ~ is shown in Figure 1.
the angular velocity of the upper arm using the MBVG In order to calculate the shoulder rotation, some terms need to
method. The calculation was based upon the developed algo- be defined as follows: U ~s , the U
~ vector when an athlete is
rithm by Ahmadi et al. in 2009 [7]. The act of upper arm standing upright without any movement prior to the serve; P,
internal rotation is shown in Figure 2. horizontal plane (transverse plane) encompassing the U; ~ and
Wrist flexion is the next main contributor (31 per cent con- ~ is projected U
U, ~ onto the P plane through angle a.
tribution) [11] to the forward speed of the racquet at impact. Due to the normal movement of an athlete during the
Wrist flexion is the bending action of the wrist joint, as shown in tennis serve (trunk incline/decline), the U~ vector can make an
Figure 2. In order to determine the wrist flexion, three markers angle with the horizontal plane. Therefore, it is needed to
were used: one on the forearm (RFRA), one on the wrist project the U~ vectors first on a horizontal plane to obtain U ~P
(RWRB), and one on the hand (RFIN). Using the three mar- and then calculate the angle between the projected vectors U ~P
~ and vector V2
kers, vector V1 ~ were created, as shown in Figure ~
and the Us to determine the shoulder rotation angle. In other
~
1a. The V1 was defined as a vector from RWRB to RFRA, and words, shoulder rotation is angle b subtended between U ~P
~ was defined as a vector from RWRB to RFIN. The angle
the V2 and U~S on the P plane, as shown in Figure 3(b). The shoulder
between the two vectors was calculated and then differentiated rotation angular velocity can be then calculated by differ-
over time to obtain the wrist flexion angular velocity. entiating the calculated shoulder rotation angle over time.
Forward shoulder rotation (positive rotation about the In order to calculate the forward racquet head speed, two
medial axis) is another main contributor (10 per cent contri- markers, M1 and M2, were attached on the sides of the head of
bution) [11] to the forward speed of a tennis racquet at impact. the tennis racquet in a way that the median point of the two
In this article, instead of forward shoulder rotation, the term markers could define a point C as the centre of the head of the
‘shoulder rotation’ will be used. Shoulder rotation motion is racquet. The tennis racquet, the attached markers, and the
shown in Figure 3(a). calculated centre point C are shown in Figure 1(b). It should be
A marker on the RSHO and the left shoulder (LSHO) noted that throughout this article, the term ‘racquet head speed’
joints were used to define vector U ~ from marker point RSHO is used instead of ‘forward racquet head speed’. The horizontal
component of the centre point of the racquet head (forward
motion) was extracted to calculate the linear forward racquet
head speed. The linear velocity of the extracted centre point was
calculated by differentiating the position of point C over time.

2.2 Inertial Sensor Technology

Three inertial sensor-based devices [12] were used in this


study. Each sensor-based device contained one 1D ADXRS300
gyroscope sensor (Brisbane, Queensland, Australia) and was
sampled at 100 Hz [12].
The dimension of the sensor device is 52 mm long  34 mm
Figure 2. Upper arm internal rotation and wrist flexion (with 54 per wide  12 mm high, weighs approximately 22 g, and is small and
cent and 31 per cent contribution during the forward swing of the light enough to be mounted on different segments of an athlete.
serve, respectively) [11]. It is a microcontroller-based platform contacting a tri-axial ac-
celerometer to measure acceleration, a 1-D gyroscope to measure
angular velocity, on-board memory to record the sessions, radio
frequency (RF) link to control the unit from distance, LCD
screen to interact with the device, USB port to download the
collected sessions and charge the device, and five-way push
buttons to turn the device on and off and control data recording.
The technical details of the device are summarized in Table 1.

Table 1. Technical specification for the inertial sensor device [12].

Components Description

Processor Atmel ATMEGA 128


Sensors Kionix KXM52-1050,3axis 2G accelerometer
Figure 3. (a) Shoulder rotation in the transverse plane about the medial ADXRS300 gyroscope
axis (10 per cent contribution during the forward swing of the serve) and Radio Nordic NRF2401, 2.4 GHz radio with internal
the direction of rotation are shown.(b) Horizontal plane (transverse plane) patch antenna
!
(P), vector from right shoulder to left shoulder (U ), vector in a stationary Memory 128-MB flash memory
! ! !
position (U s ), and the projected vector (U P ), projected U onto the P Inputs/outputs LCD screen, USB port, and a push button
plane through angle a and the rotation angle b are shown.

Sports Technol. 2009, 2, No. 3–4, 129–136 & 2010 John Wiley and Sons Asia Pte Ltd www.sportstechjournal.com 131
Research Article A. Ahmadi, D. Rowlands and D. A. James

Figure 4. Placement of gyroscope sensors on the chest to measure


shoulder rotation (gyroscope A), the upper arm to measure upper arm
internal rotation (gyroscope B), and the hand to measure the wrist
flexion (gyroscope C).

The placement of the three gyroscope sensors to determine


the upper arm rotation, shoulder rotation, and wrist flexion is
shown in Figure 4. Gyroscope sensors were light enough to be
mounted on the body using double-sided tape. Gyroscope A,
which was mounted on the chest, determined the shoulder
rotation; gyroscope B, which was mounted on the upper arm,
determined the upper arm internal rotation; and gyroscope C,
which was mounted on the hand, determined the wrist flexion.
The correlation between the maximum peak of upper arm
internal rotation, wrist flexion, shoulder rotation, racquet head
speed, and skill level is presented in the Results and Discussion
sections. In addition, the output comparison between the
marker-based methods and gyroscope sensors is shown.
The following Results and Discussion sections are divided
into five topics: skill assessment, skill acquisition, removing rac-
quet head speed dependence, gyroscope sensor placement, and
simulated gyroscope. The focus of the skill assessment section is
to show how athletes can be assessed with respect to the peak
values of their main contributors during the first serve prior to
impact. The focus of the skill acquisition section is to show how
to apply the obtained results from the skill assessment section to
provide possible feedback to the athlete so that they are able to
compare their serves with an elite’s serves to try to improve their
swings. The focus of the removing racquet head speed depen-
dence is to show that skill assessment and skill acquisition can be
done in the absence of racquet head speed. The focus of the
gyroscope sensor placement section is to show that there is a close
relationship between the output of the sensors on the chest and
on the hand and those from the marker-based methods. Finally,
the focus of the simulated gyroscope section is to show that Figure 5. Peak of the angular velocity of (a) the upper arm internal
simulated gyroscope sensors are capable of measuring skill as- rotation, (b) the wrist flexion and (c) the shoulder rotation versus
sessment and skill acquisition for high-speed serves. racquet head speed. Participant 1 is an amateur, participants 2 and 3
were subelite, and participant 4 was an elite tennis player.
participant 1; J participant 2; + participant 3; } participant 4.

3. RESULTS (10 per cent contribution), with respect to the racquet head
speed during the first tennis serve for all the participants.
3.1 Skill Assessment Participant 1 was an amateur player, participants 2 and 3
were subelite players, and participant 4 was an elite player.
The angular velocity of the upper arm internal rotation,
wrist flexion, and shoulder rotation was calculated for the 31st
serves for each athlete. Figure 5 shows the relationship bet- 3.2 Skill Acquisition
ween the peak values of the main contributors containing the
upper arm internal rotation (54 per cent contribution), wrist In this section, a possible method for skill improvement is
flexion (31 per cent contribution), and shoulder rotation shown. Scatter plots for the upper arm rotation and the wrist

132 www.sportstechjournal.com & 2010 John Wiley and Sons Asia Pte Ltd Sports Technol. 2009, 2, No. 3–4, 129–136
Tennis skill assessment

Figure 6. Line of improvement as a function of upper arm internal


rotation, wrist flexion, and racquet head speed.

flexion as dependant variables, and the racquet head speed as


an independent variable, are shown in Figure 6.
Upper arm internal rotation and wrist flexion were chosen
as they have more contribution effects and importance to the
maximum racquet head speed after the maximum knee flexion
in the first tennis serve. As shown in Figure 6, there is well-
separated clustering for different skill levels. According to
the shape of the scatter data, a straight line can be fitted to the Figure 7. Use of a line of improvement in the absence of racquet head
data. The least-squared fit technique was applied to create speed by (a) using upper arm and wrist data and (b) upper arm, wrist,
the fit line (R2 5 0.89). The equation of the fit line is: and shoulder data.

x  18:19 y  5:45 z  20:94 Pearson’s correlation coefficient (r) and significant difference
¼ ¼ ¼t ð1Þ
30:90 19:04 9:65 test results (P) were used to quantify the relationship between
the slow motion serve and the normal speed serve action. The
correlation between the slow motion serve and the normal speed
3.3 Removing Racquet Head Speed Dependence serve was found to have similar trends (r 5 0.8680, Po0.0001).
A previous study [7] has shown that a gyroscope can
follow the trends of a slow motion serve. Figure 8(a,b) shows
It has already been shown in Figure 5 that there is a linear
that a gyroscope can follow the trends of the wrist flexion and
relationship between each main contributor (upper arm in-
the shoulder rotation for a slow motion serve. This shows that
ternal rotation, wrist flexion, and shoulder rotation) to the
gyroscopes are capable of following the trends of a serve.
racquet head speed. This means that the values of the con-
It was shown that gyroscopes can capture the components
tributors were increasing as the racquet head speed increased.
of the movements for a tennis serve. Further in the text,
Therefore, it is possible to remove the racquet head speed and
simulated gyroscopes are developed and used for classifying
define the line of improvement in 2-D by only using the upper
athletes during a high-speed first serve in tennis.
arm data and wrist data instead of the 3-D case as shown in
Figure 7(a). It is also possible to remove the racquet head
speed and define the line of improvement in 3-D by using the
upper arm, wrist, and shoulder data as shown in Figure 7(b). 4. DISCUSSION

4.1 Skill Assessment


3.4 Gyroscope Sensor Placement
In Figure 5(a–c), clear bands can be seen between the main
The aim of this section is to show that gyroscope sensors contributors and the racquet head speed. A relationship between
could be used as wearable devices to determine the peak of the each main contributor and the racquet head speed can be seen for
upper arm internal rotation, wrist flexion, and shoulder rotation each banding. The band shows that the racquet head speed is
during the forward motion of the tennis serve and thus, they can increased for increasing skill level as expected from the literature
be used as a potential skill assessment and acquisition tool. Due [11]. It can be seen that the upper arm internal rotation, wrist
to the limitation of gyroscope sensors to detect the fast rate of flexion, and shoulder rotation are increasing for increasing skill
rotational motions, slow motion serves rather than a normal level. For instance, participant 4 (elite player) has higher peak
power first serve were performed. The biomechanic movement values than participant 1 (amateur player), and the peak values
of the slow motion serve was observed to be similar to that of from participants 2 and 3 are higher than those of participant 1
the normal speed serve, except that ball was hit with less power. and lower than those of participant 4.

Sports Technol. 2009, 2, No. 3–4, 129–136 & 2010 John Wiley and Sons Asia Pte Ltd www.sportstechjournal.com 133
Research Article A. Ahmadi, D. Rowlands and D. A. James

Figure 9. Suggested method to use the line of improvement is shown.


P1 is a new collected data point and P2 is the mapped version of P1 in
!
a way that both P1 and P2 have the same racquet head speed. d is
the distance vector between P1 and P2.

It indicates that the higher one is on the line, the closer to the
professional serve. This line is not the line of ‘best technique’, but
indicates a traversal path that can be followed to improve from
amateur to elite. It should be noted that this line is generated
from the available population of athletes and would benefit from
a greater number of players and serves. However, the line is still
an indicator of skill acquisition and skill improvement.
All athletes have different needs, so the different require-
ments from athletes in different levels dictate the way the line is
used. It is up to the coaches and sport scientists to interpret
the results and provide the relevant feedback to a player. An
example method of using the line of improvement for the 3-D
Figure 8. Comparison between the gyroscope sensor output (mea- case is as follows. Data point P1 consists of the racquet head
sured) and the marker-based developed methods (calculated) for (a) speed, wrist flexion, and upper arm internal rotation collected
shoulder rotation angular velocity (b) and wrist flexion angular velocity from a player during the first serve in tennis. The collected data
during the slow motion tennis serve. measured; calculated.
point (P1) can then be mapped onto the line of improvement to
give point P2, as shown in Figure 9. P2 is obtained in such a way
Also, distinct clustering can be seen between the different that both points (P1 and P2) have the same racquet head speed.
skill levels. In Figure 5, it can be seen that the lower cluster The components of the distance vector d~ between P1 and P2
belongs to an amateur player, the middle cluster relates to the can identify the amount of upper arm rotation and wrist flexion
subelite players, and the top cluster corresponds to the elite improvement needed to approach the line of improvement.
player. This is also expected due to the fact that elite players Once the athlete has approached the line, it is possible to climb
generate more racquet head speed, which means that upper up the line to obtain a higher skill level. The ability to traverse
arm internal rotation, wrist flexion, and shoulder rotation are the line may be limited based upon the physiology of the player,
also increased, since they contribute approximately 85 per cent which may prevent him/her traversing further.
to the racquet head speed at impact. Therefore, According to As will be discussed in the next sections, the ultimate aim is
Figure 5a–c, athletes can be assessed and classified with respect to use the gyroscope sensors and measure the players and pro-
to the peak of the upper arm internal rotation, wrist flexion, vide real-time feedback on the field during a training session.
and shoulder rotation respectively prior to impact. However, using the gyroscope sensors, the racquet head speed
cannot be easily determined. Therefore, it is needed to show that
skill assessment, as well as the skill acquisition, are still feasible in
4.2 Skill Acquisition the absence of the racquet head speed. In the following section,
it is shown that it is possible to remove the dependence of the
It can be seen in Figure 6 that higher values on the line racquet head speed and still see banding and separate clustering.
correspond to more skilful athletes. For instance, low racquet
head speed, upper arm, and wrist values belong to amateur
player right at the bottom of the line. Those values are growing 4.3 Removing Racquet Head Speed Dependence
for subelite players and the highest values correspond to the elite
players. The line shows a progression from amateur to subelite to Figure 7(a) shows the relationship between the upper arm
elite, so it is possible to think of this line as a line of improvement. and wrist data and the skill level. According to the shape of the

134 www.sportstechjournal.com & 2010 John Wiley and Sons Asia Pte Ltd Sports Technol. 2009, 2, No. 3–4, 129–136
Tennis skill assessment

data, a straight line was fitted to the data (direction vector


V~d1 ¼ h10; 16:5i, R2 5 0.85). It shows that the line of im-
provement is valid in the absence of the racquet head speed.
Similar to the 3-D case, the line of improvement (direction
vector V~d2 ¼ h37:9; 9:05; 11:44i, R2 5 0.87) can be defined in 4-
D by including the shoulder rotation as another dependant
variable and can be reduced to the 3-D case by removing the
racquet head speed as shown in Figure 7(b). Figure 7(b) shows
that the line of improvement can be defined by using the
data of the three main contributors (upper arm rotation,
wrist flexion, and shoulder rotation) in the absence of the
racquet head speed. The linear relationship between the
three main contributors and the skill level is clearly shown
in Figure 7(b). Therefore, skill assessment and acquisition
can be done without any knowledge of the racquet head
speed values.

4.4 Simulated Gyroscope

The aim of this section is to show that simulated gyroscope


sensors are capable of measuring skill assessment and skill
acquisition for high-speed serves. In the previous sections, the
marker-based methods were used to calculate the upper arm
internal rotation, wrist flexion, and shoulder rotation only,
and athletes were assessed based upon the peak values of the
angular velocities of those three elements. However, gyroscope Figure 10. (a) Skill assessment using the simulated gyroscope.
(b) Line of improvement created using the simulated gyroscope.
sensors show more complex rotations due to linkage of seg-
participant 1; J participant 2; + participant 3; } participant 4.
ments of the body. Therefore, simulated gyroscopes based
upon marker positions were developed to simulate the output
behavior of the gyroscope sensors. In this section, the effect of To determine if skill assessment can be seen using simu-
segment linkage on the upper arm internal rotation, wrist lated gyroscopes, the racquet head speed versus simulated
flexion, and shoulder rotation during the forward motion of gyroscopes were plotted, as seen in Figure 10. In Figure 10(a),
the tennis serve is discussed. well-separated clusters are apparent for different skill levels.
A previous study [7] indicated that the output of a simu- In Figure 10(b), different clusters for different skill levels are
lated gyroscope sensor on the upper arm is influenced by the shown. It can be seen that a line of improvement (direction
upper arm internal rotation, as well as the shoulder rotation. vector V~d3 ¼ h21:75; 5:5; 20:94i, R2 5 0.87) can be fitted to the
Therefore, simulated gyroscope for the upper arm will contain clusters as a potential tool for skill assessment.
the summation of the calculated upper arm internal rotation Since the simulated gyroscope sensors can be used to
and the calculated wrist flexion only. In Figure 10(a), the peak model a high-speed serve, it can be suggested that the sensor
values for the shoulder rotation plus upper arm internal technology as a wearable technology can be used to assess the
rotation versus racquet head speed is shown during a high- performance of athletes and to provide the required feedback
speed tennis serve. A clear banding can also be seen, as well as to the athletes on the field during a training session. However,
different clustering for different skill levels. This indicates that due to the technology limitations, the currently-available gy-
a gyroscope sensor on the upper arm can allow skill level to be roscope sensors are not yet capable of measuring the fast ro-
distinguished. tational motion. As a result, as soon as technology is advanced
In order to determine the wrist flexion, a simulated gyro- enough to develop high-range gyroscopes, it is feasible to
scope needs to be placed on the hand, as shown in Figure 4. As employ them as a training device on the tennis court.
can be seen in Figure 8(b), the measured sensor output and the
calculated wrist flexion angular velocity only are very close.
Thus, a simulated gyroscope for wrist flexion will consist of the 5. CONCLUSION
calculated wrist flexion only.
In order to determine the shoulder rotation, a simulated In this study, athletes were assessed according to the mea-
gyroscope needs to be placed on the chest, as shown in Figure 4. surements of the main contributors between the point of the
It was found that the gyroscope sensor on the chest is not maximum knee flexion and the point of impact during the first
greatly influenced by any segment linkages of the body during serve in tennis. The trajectory of marker positions on the upper
the service action, as can be seen in Figure 8(a). Therefore, the body with respect to the Vicon Plug-in Gait model was used to
calculated shoulder rotation can be used for the simulated develop the marker-based methods to calculate the angular ve-
gyroscope as it can follow the trends. locity of each main contributor to generate the serve. The peak

Sports Technol. 2009, 2, No. 3–4, 129–136 & 2010 John Wiley and Sons Asia Pte Ltd www.sportstechjournal.com 135
Research Article A. Ahmadi, D. Rowlands and D. A. James

values of the upper arm internal rotation, wrist flexion, and REFERENCES
shoulder rotation angular velocities were calculated for athletes
with different skill levels and plotted against racquet head speed. 1. Sprigings R, Marshall B, Elliott B, Jennings L. A three-dimensional
kinematic method for determining the effectiveness of arm segment
Clear banding and well-separated clustering were shown for rotations in producing racquet-head speed. Journal of Biomechanics 1994;
different skill levels. Due to the fact that skill acquisition was an 27: 245–254.
important aspect of this study, the line of improvement was 2. Ohgi Y, Baba T. Measurement of deceleration of golfer’s sway and unlock
developed. The line was the best fit through the clusters obtained timing in driver swing motion. In Subic A, Ujihashi S (Eds), The Impact of
from the available population and indicated the path of im- Technology of Sport. Australasian Sports Technology Alliance: Japan,
2005; 349–354.
provement from amateur players to elite player. The distance
3. Davey N, Anderson M, James DA. Validation trial of an accelerometer-
vector between any new collected data and the mapped data on based sensor platform for swimming. Sports Technology 2008; 1(4): 202–207.
the line contains required vector components information on
4. Harding JW, Mackintosh GC, Martin DT, Hahn AG, James DA.
how to fix the deficiency. It was also shown that due to the linear Automatic scoring for elite half-pipe snowboard competition: important
relationship between the racquet head speed and all the three sporting development or techno distraction? Sports Technology 2008; 1(6):
main contributors (upper arm internal rotation, wrist flexion, 277–290.

and shoulder rotation), skill assessment/acquisition can be ap- 5. Wixted A, Thiel D, Hahn A, Gore C, Pyne D, James DA. Measurement of
parent without the use of the racquet head speed. energy expenditure in elite athletes using MEMS–based inertial sensors.
IEEE Sensors Journal 2007; 7(4): 481–488.
It was also shown that there is a reasonably close match
6. Ahmadi A, Rowlands D, James DA. Investigating the translational and
between the calculated results and the measured results using rotational motion of the swing using accelerometers for athlete skill assessment.
the inertial gyroscope sensors during the slow motion serve in Proceedings of IEEE Sensor Conference; 22– 25 October 2006, Daegu, Korea.
tennis. It was found that all the required angular velocities for 7. Ahmadi A, Rowlands D, James DA. Measuring upper arm rotation in
the skill assessment and skill acquisition can be obtained using tennis serves using videography, inertial and a novel marker based
three gyroscope sensors mounted on the upper arm, the chest, technologies. Journal of Sports Engineering, submitted April 2009.

and the hand. Therefore, it will be possible to measure the 8. Ahmadi A, Rowlands D, James DA. Technology to monitor and
athletes on the field and provide them with time feedback when enhance the performance of a tennis player. In Pope N, Kuhn K, Forster J
(Eds), Digital Sport for Performance Enhancement and Competitive Evolution:
the fast gyroscopes are developed. Intelligent Gaming Technology 2009; pp 101–121. USA: IGI Global.
In this study, three gyroscopes were suggested to measure 9. Bahamonde R. Changes in angular momentum during the tennis serve.
performance, and thus classify the athletes according to the Journal of Sports Sciences 2000; 18: 579–592.
peak of the main contributors during the first tennis serve. In 10. Gordon BJ, Dapena J. Contributions of the joint rotations to racquet speed
future, further studies will be required to minimize the number in the tennis serve. Journal of Sports Sciences 2004; 24(1): 31–49.
of sensors and ideally use only one sensor to capture the whole 11. Marshall RN, Elliot BC. Long-axis rotation: The missing link in
swing. proximal-to- distal segmental sequencing. Journal of Sports Sciences
Overall, this article suggests a method to examine the use 2000; 18: 247–254.

of gyroscope sensors as a wearable device to assess the per- 12. Davey N, Wixted A, Ohgi Y, James DA. A low cost self contained plat-
formance of tennis players during the first serve. This helps form for human motion analysis. Proceeding of the 3rd Asia Pacific
Conference on Sports Technology; 23–26 September 2007, Singapore.
athletes with different skill levels to be monitored and assessed
in the real environment (tennis court) instead of laboratories,
and obtains real-time or close to real-time feedback on the field Received 11 July 2009
during training sessions. Also, since the gyroscope sensor Revised 9 October 2009
technology is cheap compared to the other technologies, it Accepted 12 October 2009
makes it possible that a wide range of tennis players could
benefit from using the sensor technology. Published online 7 February 2010

136 www.sportstechjournal.com & 2010 John Wiley and Sons Asia Pte Ltd Sports Technol. 2009, 2, No. 3–4, 129–136

You might also like