Touch key design for target selection on a mobile phone

Abstract

Mobile phones with a touch screen replacing traditional keypads have been introduced to the market. Few studies, however, have been conducted on the touch interface design for a mobile phone. This study investigated the effects of touch key sizes and locations on the one-handed thumb input that is popular in mobile phone interactions. Three different touch key sizes (i.e. square shape with 4mm, 7mm, and 10mm wide) and twenty five locations were examined in an experiment. The results provided two groups of touch key locations (an appropriate and an inappropriate region) with respect to three usability measures including success rate, number of errors, and pressing convenience. In addition, a hits distributions based algorithm was applied to target selection tasks, which statistically improved the performance. The results of this study could be used to design touch keys so as to enhance the usability of mobile phones with a touch screen.

Full text

Touch Key Design for Target Selection on a Mobile Phone
Yong S. Park1, Sung H. Han2, Jaehyun Park3, Youngseok Cho4
Department of Industrial and Management Engineering
Pohang University of Science and Technology (POSTECH)
San 31 Hyoja, Pohang, South Korea 790-784
+82-54-279-28621,3,4, +82-54-279-22032
{drastle1, shan2, parkdog33, kilys4} @postech.ac.kr
ABSTRACT
Mobile phones with a touch screen replacing traditional keypads
have been introduced to the market. Few studies, however, have
been conducted on the touch interface design for a mobile phone.
This study investigated the effects of touch key sizes and
locations on the one-handed thumb input that is popular in mobile
phone interactions. Three different touch key sizes (i.e. square shape
with 4mm, 7mm, and 10mm wide) and twenty five locations were
examined in an experiment. The results provided two groups of touch
key locations (an appropriate and an inappropriate region) with respect to
three usability measures including success rate, number of errors, and
pressing convenience. In addition, a hits distributions based algorithm
was applied to target selection tasks, which statistically improved the
performance. The results of this study could be used to design touch keys
so as to enhance the usability of mobile phones with a touch screen.
Categories and Subject Descriptors
H.5.2 [Information Interfaces and Presentation]: User
Interfaces – Ergonomics, Input devices and strategies.
General Terms
Algorithms, Performance, Design, Experimentation, Human
Factors
Keywords
One-handed thumb input, mobile phones, touch screen, hits
distribution based algorithm, usability
1. INTRODUCTION
Touch screens are widely used for a variety of mobile devices
because they are highly intuitive and require little space to
implement [1]. Moreover, touch interfaces are easy to adjust the
design parameters, such as key size, spacing between keys and
location on the screen. Recently, mobile phones with a touch
screen replacing traditional keypads, e.g. AppleTM iPhone, are
coming into the spotlight.
Users tend to use only one hand when they use a mobile device
[4]. In other words, they hold a mobile phone with one hand and
interact with it using a thumb. In addition, they would use both
hands only when the user interface makes one hand interaction
impossible or difficult.
It is difficult to find studies investigating critical design factors
such as touch key size, touch key location, and touch recognition
area on a mobile device, although one handed interaction on a
mobile device is popular. A previous study investigated one-
handed thumb input on a PDA [6]. This study divided a PDA
screen into 3¯3 areas and examined usability of each region.
However, the results for the nine areas are not enough to be used
to design a mobile phone interface. Note that, mobile phones in
use often provide more than nine input elements simultaneously.
For example, an AppleTM iPhone can provide more than three
items in a row and in a column.
Input accuracy is critical to designing a mobile phone interface
since people are using mobile phones frequently for a variety of
purposes. Studies on improving input accuracy have been
performed for many interfaces, e.g. virtual keyboarding [3] and a
touch screen based keyboard [7].
This study aims to understand one-handed thumb input on a touch
screen and to enhance usability of a mobile phone with a touch
screen. To achieve the purposes, a human factors experiment is
conducted to investigate the effects of touch key sizes and
locations on usability of one-handed touch input. In addition, an
algorithm based on hits distribution is applied in order to support
user’s input and to improve touch input accuracy.
2. Experimental Methods
2.1 Subjects
A total of thirty right-handed subjects participated in a human
factors experiment. Their age ranged from 18 to 28 years old
(mean = 23.1, SD = 2.5). They had normal vision and no problem
to freely move their right thumb. Twenty of them had not used a
mobile device with a touch screen (e.g. personal digital
assistants), while the others had experienced for 1.2 years on the
average.
2.2 Experimental design
A within-subject design was used in the experiment, in which two
within-subject variables (touch key size and touch key location)
were included. The touch key size factor had three levels (square
shape with 4mm, 7mm, and 10mm wide). A pilot test revealed the
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
MobileHCI 2008, September 2–5, 2008, Amsterdam, the Netherlands.
Copyright © 2008 ACM 978-1-59593-952-4/08/09…$5.00.
SP
423

three touch key sizes represented small, medium and large touch
key sizes, respectively.
The touch key location factor had 25 different levels. Each
location was one of center points of 25 rectangular areas that had
the same width and height (that is, one-fifth of a touch screen
width and one-fifth of a touch screen height, respectively). Each
area had its own ID (See Figure 3). In the experiment, the center
point of a square touch key was positioned one of 25 touch key
locations. That is, although touch keys had different sizes, the
center points of them were located on the same position when
they are located in the same touch key location. Figure 1
illustrated an example of experimental conditions, which had a
touch key size of 10 mm.
26 Pixel
294 Pixel
240 Pixel
Targ et k ey
Size: 10mm long
(46 Pixels)
Location: area ID of 9
Figure 1. An example of experimental tasks
2.3 Dependent measures
Two types of dependent measures (pressing performance and a
subjective satisfaction score) were collected in the study. The
pressing performance measures included success rate, number of
errors, and pressing deviation. The success rate was calculated
based on the number of tasks correctly pressed at the first press.
The pressing deviation was calculated by the difference between a
center of a target and a centroid of a pressed area.
In order to obtain the subjective satisfaction score, each subject
was asked to rate pressing convenience for each experimental
condition using a nine point rating scale. The pressing
convenience meant how easily the subjects could press a target.
2.4 Apparatus
A commercial PDA with a touch screen size of 240¯320 pixels
(HPTM iPAQ rz1717) was used to implement an experimental
prototype because there were a few mobile phones that equipped
touch screens which could be easily manipulated in the
experiment. It also had smaller body size than other mobile
devices with touch screens (e.g. PMPs and other PDAs), which
could provide a device size similar to real mobile phones
2.5 Experimental tasks
Each experimental task consisted of two states, a stand-by state
and an input state. In the stand-by state, the experimental
prototype was waiting for any user press on a touch screen. If a
subject pressed any location on a touch screen in the stand-by
state, the state changed to an input state after 0.3 second delay.
That is, a blue touch key, a target key, was presented. In case that
a subject pressed the blue key correctly, a ‘beep’ sound was
provided. Then a stand-by state for the next task started. If a
subject failed to press the target correctly, a pressed location for
every press was recorded and no response was provided by an
incorrect press. Subjects who failed to press the target correctly
were asked to press the targets until they succeeded.
2.6 Experimental procedure
Each subject was given written instructions on the experimental
objectives and procedures at the beginning of the experiment.
Then he/she was asked to hold a PDA with his/her right hand and
to practice pressing targets by the right thumb. In case that thumb
movements by the right hand were interfered by the PDA body,
the subject was allowed to put his/her left hand underneath the
PDA in order to support easy and free thumb movements like real
mobile phone use. Examples for the two methods to hold a PDA
were presented in Figure 2.
Figure 2. Two methods to hold a PDA. Left and Right figures
showed one-hand and two-hands holding, respectively.
The main experiment consisted of three blocks. Each block
consisted of 25 experimental conditions (i.e. 25 different touch
key locations with the same touch key size). For each
experimental condition, the two-state experimental task was
repeated 10 times. That is, each subject carried out a total of 250
tasks in each block. After completing each block, the subjects
rated pressing convenience for 25 touch key locations. The
presentation order of three blocks was determined by the Latin-
square balancing technique.
3. Results
The ANOVA on ranks, one of non-parametric statistical
techniques, was applied to two error related measures, the success
rate and the number of errors, and the pressing convenience
because of non-parametric characteristics of the data [2]. The
results showed all measures were affected by the touch key size
and the touch key location. Table 1 showed the ANOVA results
including F-statistic and p-value.
424
SP

Table 1. Summary of the ANOVA on ranks results
Touch key size Touch key location
Success rate F(2,58)=437.1,
p<0.01 F(24,696)=6.6,
p<0.01
Number of errors F(2,58)=501.0,
p<0.01 F(24,696)=6.5,
p<0.01
Pressing
Convenience F(2,58)=57.1,
p<0.01 F(24,696)=75.34
p<0.001
As post-hoc analyses, the Student Newman-Keuls (SNK) tests
were conducted on significant main effects (i.e. the touch key size
and the touch key location) at the 0.05 significance level. The
results revealed, as expected, the number of errors decreased as
the touch key size increased. In addition, it was provided that the
larger the touch key size, the higher the success rate and the
pressing convenience.
Two groups of touch key locations, appropriate and inappropriate
regions, were identified by the SNK tests. An appropriate region
provided good usability in terms of each dependent measure,
while an inappropriate region provided poor usability. With
respect to all three measures, there was significant difference
between the two groups at the 0.05 significance level. Figure 3
illustrates two groups, in which the darkest areas and white areas
represent appropriate and inappropriate regions, respectively.
24
19
14
9
4
22
17
12
7
2
1086
252321
201816
151311
531
24
19
14
9
4
22
17
12
7
2
1086
252321
201816
151311
531
24
19
14
9
4
22
17
12
7
2
1086
252321
201816
151311
531
Success Rate Number of Errors Pressing Convenience
Figure 3. Appropriate and inappropriate regions for three
measures. The darkest area and white areas represent
appropriate and inappropriate regions, respectively.
Each number is an ID of each area.
The ANOVA on the deviation data showed the touch key size
(F(2,58)=5.95, p=0.004) and the touch key location
(F(24,696)=7.61, p<0.001) significantly affected on x-axis
deviation at the 0.05 significance level. Similarly, y-axis
deviation was significantly influenced by the touch key size
(F(2,58)=5.95, p=0.005) and the touch key location
(F(24,696)=7.61, p<0.001). Table 2 shows touch key locations
with the largest deviation.
Table 2. Touch key locations with the largest deviation
x-axis y-axis
Positive direction 1, 11, 21 1, 4, 5
Negative direction 15, 20, 25 21, 22, 24
4. Discussions
Small touch keys have poor performance in terms of the success
rate and the number of errors according to the results. Touch keys
with the size of 4mm provided the lowest performances in
boundary region (i.e. areas with IDs of 4, 20, 21, 22, 24, and 25).
For example, the success rate and the number of errors in the area
with an ID of 4 were 54.3% and 4.8, respectively. Also, touch key
location with an area ID of 22 provided 61.0% and 5.6.
The poor performance could be explained by the amount of
feedback information during pressing a target and the input
recognition algorithm of the touch screen. Feedback information
is necessary during a rapid human movement to check if a
movement reaches a target [5]. The users can get only visual
information from a touch screen because it is difficult to
implement tactile feedback on a touch screen. Worse yet, when
using a mobile device by one hand, it is difficult to get even
visual information if targets are visually interfered by the hand
and fingers. A traditional keyboard is one of input devices that
have quite high performance. It provides tactile feedback so that
users can easily recognize which keys they are pressing and
whether the keys are successfully pressed or not.
Touch keys are activated only if the centroid of a pressed area
falls in the recognition area of a target key, which requires users
to take locations of the centroid during pressing tasks. In case of
pressing a target by a thumb, however, it is difficult to calculate
and take the centroid of a pressed area in mind. Worse yet, due to
the anatomy of the hand, users need considerable thumb flexion
and extension to press a target in some areas (e.g. areas with IDs
of 20, 22, and 25) of the boundary region, which can make
pressing accuracy quite low.
Adjusting a location of a touch recognition area is one of the
solutions to improve the poor pressing performance. The solution
seems to acceptable since it is simple and easy to manipulate
design parameters by a software manner. In addition, from the
collected deviation data, it is easy to obtain hits distributions from
which a movement range of a recognition area could be obtained.
The solution was applied to two touch key designs (4 mm touch
keys located in the areas with IDs of 4 and 22), as a case study.
Figure 4 illustrates hits distributions of the two designs. It showed
pressing pattern of each touch key design. For example, the Y-
axis deviation of the touch key location of 22 showed that
subjects tended to press positions upper than center of targets
because most hits had negative Y-axis deviation. Movement
ranges of a recognition area in both directions were determined
using the hits distributions. Then, success rates were re-calculated
within the movement ranges. The Friedman test revealed
significant differences between the original success rates and the
re-calculated maximum success rates for both touch key designs
(for both, p=0.01) at the 0.05 significance level. Specifically, the
touch key with an area ID of 4 had the maximum success rate of
65.7% (21% increase compared to the original success rate of
54.3%) when a recognition region moved 5 pixels in the x
direction and -2 pixels in the y direction. The success rate of the
other design increased by 9.8% (from 61.0% to 67.0%) in case
that a recognition region moved -2 pixels and -3 pixels in the x
and the y directions, respectively. Note that, the success rates
SP
425

increased significantly, although the recognition areas moved less
than 6 pixels, about 1.4 mm long.
0
2
4
6
8
10
-15 5 2
5
X- axis Deviation
0
2
4
6
8
10
-19 -2 15
Y- axis Deviation
X- axis Deviat ion Y- axis Deviat ion
X- axis Deviat ion Y- axis Deviat ion
Touch Key Size: 4mm, Touch Key loc ation: 4
Touch Key Siz e: 4mm, Touc h Key loc ation : 2 2
0
2
4
6
8
10
-23 - 4 15
0
2
4
6
8
10
-27 -7 1
3
Hits (%)Hits (%)
Hits (%) Hits (%)
Figure 4. Hits distributions of two touch key designs
The subjective satisfaction score (the pressing convenience)
seems to be higher in the center region (i.e. areas with IDs of 7, 8,
9, 12, 13, 14, 17, and 18) than other regions (See Figure 3). This
result is consistent with [6]. This study, however, provides more
detailed results because the previous study divided a touch screen
into nine regions.
Input elements that require a highly accurate control, e.g. a shutter
button for a camera function, could be located in the leftmost
areas with IDs of 6, 11, and 16 because the areas had higher
success rate and lower number of errors than other regions (See
Figure 3). Also, center areas with IDs of 7, 8, 9, 12, 13, 17 and 18
are recommended for general input elements because they provide
high pressing convenience and their performance in terms of the
success rate and the number of errors is not poor.
5. Conclusion
This study was conducted to investigate effects of touch key sizes
and locations on mobile phone usability in terms of the success
rate, the number of errors, and the pressing convenience. Also, a
hits distribution based algorithm was used to improve one-handed
thumb touch input on a mobile phone.
The results provided the larger touch key size, the higher
performance and subjective satisfaction. In addition, two types of
touch key locations (an appropriate and an inappropriate region) were
statistically identified with respect to the success rate, the number of
errors, and the pressing convenience. Finally, an algorithm to adjust a
location of a recognition area was applied to two touch key designs that
provided poor pressing performance. The algorithm statistically
increased the success rates. The results of this study could be used to
design touch keys so as to enhance the usability of mobile phones with a
touch screen.
Pressing serial targets are one of common tasks that frequently
happen in mobile phone use, while this study focused on pressing
a single target. Further studies are required regarding serial target
selections on a mobile touch screen. Different hits distribution
based algorithms to help pressing tasks (e.g. on-line adaptation,
adjusting size of a recognition area, etc.), also, are worth to be
examined in further studies.
6. ACKNOWLEDGMENTS
This work was supported by the Korea Science and Engineering
Foundation (KOSEF) grant funded by the Korea government
(MOEST) (No. R01-2006-000-11142-0)
7. REFERENCES
[1] Colle, H. A. and Hiszem, K. J. 2004. Standing at a kiosk:
Effects of key size and spacing on touch screen numeric
keypad performance and user preference. Ergonomics. 47, 13,
1406-1423.
[2] Hesel, D. R. and Hirsch, R. M. 2002. Statistical methods in
water resources. U.S. Department of the Interior. DOI=
http://pubs.usgs.gov/twri/twri4a3/
[3] Himberg, J., Häkkilä, J., Kangas, P., and Mäntyjärvi, N.
2003. On-line personalization of a touch screen based
keyboard. In Proceedings of the Intelligent User Interfaces
Conference (Miami, Florida, January 12 - 15, 2003). IUI’03.
ACM Press, New York, NY, 77-84.
[4] Karlson, A. K., Bederson, B. B., and Contreras-Vidal, J. L.
2006. Understanding single-handed mobile device
interaction. Technical Report. University of Maryland at
College Park.
[5] Meyer, D. E., Abrams, R. A., and Wright, C. E. 1990. Speed-
accuracy tradeoffs in aimed movements: toward a theory of
rapid voluntary action. M. Jeannerod, Ed. Lawrence Erlbaum
Associates, Mahwah, New Jersey, 173-226.
[6] Parhi, P., Karlson, A. K. and Benderson, B. B. 2006. Target
size study for one-handed thumb use on small touchscreen
devices. In Proceedings of the Mobile HCI Conference
(Helsinki, Finland, September 12 - 15, 2006). MobileHCI’06.
ACM Press, New York, NY, 203-210.
[7] Zhai, S., Sue, A., and Accot, J. 2002. Movement model, Hits
distribution and learning in virtual keyboarding. In
proceedings of CHI (Minneapolis, Minnesota, April 20 - 25,
2002). CHI’02. ACM Press, New York, NY, 17-24.
426
SP