are 5 buttons enough- destination on touchscreen keyboards

1、Proceedings of the 4th International Conference on Automotive User Interfaces and Interactive Vehicular Applications (AutomotiveUI 12), October 1719, 2012, Portsmouth, NH, USAAre 5 Buttons Enough: Destination Input on Touchscreen KeyboardsDavid Wilfinger, Martin Murer and Manfred TscheligiChristian

2、Doppler Laboratory “Contextual Interfaces” ICT&S Center, University of Salzburg Austriafirstname.lastnamesbg.ac.atABSTRACTAlthough alternative means of text input in vehicles have already been developed, touchscreen keyboards still pose a standard solution for this input in modern vehicles. Un- til

3、alternative means of text input will be widely available in vehicles, it is important to further investigate potential improvements in button text input on touch screens. This paper presents a study in which we compared three means of text input for navigation destination entry on a touchscreen in t

4、he central console: two keyboards using a QWERTY and an ABC layout, respectively, and a novel 5Button in- put approach using only 5 buttons to type in city names. The study focused on performance, distraction, and user experience caused by the different keyboards. Results show that the QWERTY and AB

5、C keyboards performed better in terms of perceived workload than the 5Button approach. Performance and perceived usability were best with the QW- ERTY keyboard, even compared to the ABC keyboard. No significant differences between the systems were found in lane keeping behavior. The main downside of

6、 the 5Button approach was the increase of mental workload compared to the other keyboards.of safety. It is, therefore, necessary to investigate text input mechanisms in the vehicle for reducing the distractive effect of interactive technology that requires typing.An attention demanding text input us

7、e case in todays in- car systems is the input of navigation destinations due to the length of street and city names and the resulting com- plexity of the task. If the system uses a touchscreen as input modality, text input is usually conducted with an on-screen keyboard using either a QWERTY or an A

8、BC keyboard layout. Surprisingly, there is no data available identifying which of the two keyboard layouts is more distractive in the automotive context. Additionally, there is a potential for al- ternative keyboard layouts on central console touch screens for reducing driver distraction when typing

9、 text as secondary task. Therefore, this paper compares three means of text in- put on central console touchscreen for destination input: a keyboard with a QWERTY layout, a keyboard with an ABC layout, and a novel 5Button keyboard.2.RELATED RESEARCHBoth industry and research in the automotive field

10、have widely addressed different text input modalities, leading to a variety of solutions such as speech input, gesture recog- nition, or multifunctional rotary knobs. Research has come up with various approaches, for example, by investigating the potential of handwritten text input 5.Although it was

11、 found that touch screens may not be the least distractive way of destination input in an automotive user interface 9, touchscreen keyboards are widely used in infotainment systems of todays cars and in nomadic devices in the car, such as mobile navigation systems. Although dif- ferent text input me

12、chanisms for touch screens were already investigated in HCI research, a discussion is missing on how to improve text input in the vehicle by simply adjusting the keyboard layout on standard touch screens. We argue that even small adjustments have the potential to improve the drivers situation in ter

13、ms of reducing workload and distrac- tion since they can easily be integrated in todays touchscreen systems (e.g. on navigation devices).The effect of keyboard layouts on the user has already been investigated (e.g. 8, 3), but there is still research missing on the applicability of their results in

14、the automo- tive domain. We assume that the low amount of publications on touchscreen keyboard alternatives in the vehicle is caused by negative evaluation results (e.g. caused by a high learning curve 3) that are often obtained when proposing alterna- tive keyboard layouts. Although those results e

15、xist for other doma , the differences in the nature of the task betweenCategories and Subject DescriptorsH5.2. Information Interfaces and Presentation (e.g., HCI): User InterfacesKeywordsNavigation destination, text input, touchscreen1.INTRODUCTIONText input has always been a major concern in human

16、computer interaction (see 6 for an overview). In modern vehicles, functions were implemented that require the users to type in words or at least a certain amount of letters. Especially while driving, this can be a demanding activity, leading to distraction from the road followed by a reductionPermis

17、sion 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, to rep

18、ublish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.AutomotiveUI12, October 17-19, Portsmouth, NH, USA. Copyright (c) 2012 ACM 978-1-4503-1751-1/12/10 . $15.00277Proceedings of the 4th International Conference on Automotive User Interfaces and Inte

19、ractive Vehicular Applications (AutomotiveUI 12), October 1719, 2012, Portsmouth, NH, USAfree text input and the input of one term from a limited set of terms (e.g., one city out of all cities from one country) have to be taken into account when researching keyboard layouts. Additionally, the positi

20、oning of the touchscreen in the vehicle is different to text input on a desk or on a handheld device, preventing users from applying all ten fin- gers for typing. Results from other contexts might not be applicable in the automotive domain for that reason. We, therefore, see the need to investigate

21、both established and experimental keyboard layouts for improving text input in the vehicle.input. The ABC keyboard as also quite common among mo- bile navigation solutions, using a more logical arrangement of keys than the QWERTY layout. Especially for novices, the ABC keyboard can be advantageous b

22、y easing the learn- ing of the layout. Both ABC and QWERTY keyboards followed the same approach, displaying a list of cities the moment that less than 10 cities were remaining. The two established keyboards were included to compare them and to serve as a benchmark for the 5Button keyboard approach.4

23、.RESEARCH GOALSIn the presented work, our goal was to find out, which key- board layout (ABC, QWERTY, 5Button) was suited better for destination entry on touch screens in the central console. In detail, we wanted to investigate which keyboard layout1. had the better input performance.2. created the

24、better user experience in terms of perceived usability and workload.3. had less negative effects on the driving performance through distraction.In this work, we focused on destination entry as use case for investigating the keyboard layouts since it poses the most complex text input task in current

25、vehicles.3.KEYBOARD LAYOUTSFor the purpose of exploring improvement potentials in touchscreen destination entry and to compare it to estab- lished approaches we developed a 5Button keyboard together with our partner organization. The main idea behind the ap- proach is that an increase of button size

26、 on a keyboard will support the driver since an input is possible with less accu- racy in hand-eye coordination and with less visual searching. Following the basic principles of Fitts law 4, the combina- tion of larger targets combined with a comparable or smaller distance allows a faster input of l

27、etters.We, therefore, propose a system using only five buttons for typing in a navigation destination. The main idea behind the proposed keyboard layout is the possibility to increase the button size while maintaining about the same level of button presses needed to type in a city name. For that pur

28、pose, the complete set of possible inputs (i.e. cities) is divided into five groups according to their alphabetical or- der. Let us assume the user wants to drive to the German city of Frankfurt. The names of German communities are divided based on their first letter, forming the five groups (Group

29、1: A-E, Group 2: F-J, .). Users are now required to select the group that their city is located in. In our ex-5.USER STUDYFor the purpose of investigating the research goals, we set up an end user study in our laboratory. A simple driving simulator mockup was used for the study, which included a sea

30、t, pedals, a steering wheel, and a 32” screen to display the driving simulator. For the driving task we used the lane change test driving simulator 7. A resistive touchscreen (8”, 800 x 600 PX resolution) was mounted in a central consolelike position. We used a withubject design to reduce er-ample,

31、Group 2 contathe letter “F” for “Frankfurt”. In arors associated with individual differences. Therefore, eachsecond step, the “F-J” group is again divided into five sub- groups, now only containing the cities with the first lettersparticipant used all three keyboard layouts. Participants be- gan wit

32、h a baseline track, then they drove three tracks with alternating keyboards before finishing with another base- line track to take different driving behaviors and expecta- tions towards driving performance into account. The order of systems was counterbalanced to reduce learning effects. Participant

33、s were recruited through university mailing lists and received e 20 for their efforts.The trial with each keyboard followed the same order and was divided into two parts: The system was used by the participants with and without a primary task (driving and non-driving) to focus on different aspects o

34、f the system. Be- fore starting the tasks, participants could practice using each system by typing in one city, if something was unclear the present researcher answered questions.In a first step, participants typed five cities without a driv- ing task, so users could completely focus on the input mo

35、dal- ity. We used this to measure the performance of the system independent from workload induced by a driving task.In a second step, participants typed in city names while driving at a constant speed of 60 kilometers per hour. The track was limited to a duration of three minutes, according to the l

36、ane change test method in 7. While driving on the track, the present researcher read out city names to the par- ticipants, one city at a time. We decided to read out the city names to avoid visual distraction caused by giving the taskF-J. Figure 1 shows the touchscreen content after making the first

37、 selection. Again, the user selects the group that holds the city which is typed in. These steps are repeated until only ten cities are left. These ten remaining cities are then shown in a list to choose from.Figure 1: 5Button text input layout after selecting the group F-J to type in “Frankfurt”.Ad

38、ditionally, we included a standard QWERTY and an ABC keyboard layout in the prototype. The QWERTY keyboard posed the state-of-the-art in keyboard based text278Proceedings of the 4th International Conference on Automotive User Interfaces and Interactive Vehicular Applications (AutomotiveUI 12), Octob

39、er 1719, 2012, Portsmouth, NH, USAto participants. When the input was successful, the next city name was read out until the three minutes track time was over. If a mistake was made, the input was repeated. Participants were asked to type in as many cities as possi- ble, the amount of successfully ty

40、ped in cities was recorded and used as control variable for the mean lane deviation. Since participants were told to find a good balance between driving and typing in cities, we expected some participants to dedicate a different amount of resources to the road and to the system than others. We wante

41、d to allow an interpre- tation, whether single participants only cared about driving or ignored their driving performance to increase their text input speed.In the experiment, participants were requested to type in German city names, representing a standard task when choosing a new navigation destin

42、ation. We included cities, that were comparable in terms of input effort, i.e. clicks needed to type in a city, between the ABC and QWERTY layouts and the 5Button keyboard. For that purpose, the number of clicks needed for each city was computed for each input mechanism. The QWERTY layout, for examp

43、le, al- lowed the first 10% of cities to be typed in with a minimum of 4 presses, while the 5Button layout required 5 clicks at least. Cities were selected, which were in the same percentile (10, 30, 50, 70, 90) in all three input mechanism (compared to the clicks required for all other cities) to e

44、nsure a fair comparison over the conditions.5.1MeasuresFor investigating the effect of the input modalities on user experience, we focused on two factors of UX, namely per- ceived workload measured with the NASA Raw Task Load Index (RTLX) 2 and perceived usability measured with the System Usability

45、Scale (SUS) 1 questionnaire. In order to asses the input performance of the modalities, we recorded the time to complete typing five cities without parallel driv- ing. The SUS questionnaire was handed to the participants after typing cities without parallel driving task, while the RTLX questionnaire

46、 was filled out after the dual task con- dition. In order to measure the distraction caused by the text input modality and its effect on driving performance, we used a Lane Change Test (LCT) 7 to collect data on the mean lane deviation of participants during text input. The mean lane deviation serve

47、d as a measure for impaired lat- eral control and therefore distraction from the driving task caused by the secondary task.Table 1: Time needed to type in five cities without driving task (sec) and perceived usability (SUS; 0- 100).long to type five cities. A post hoc test with Bonferroni correction

48、 showed that all differences were highly significant (p0.001).This finding was confirmed when we computed the results of the SUS questionnaires that were filled out by the par- ticipants after typing in the five city names (Table 1). A repeated measures ANOVA revealed a significant effect of the inp

49、ut modality on the perceived usability (F(2, 58) = 85,245, p 0.001). A post hoc test with Bonferroni cor- rection showed statistically significant differences between the SUS values of the QWERTY keyboard and the 5Button keyboard (p 0.001). The difference in the SUS scores be- tween the ABC keyboard

50、 and the 5Button keyboard were lower but aga tatistically significant (p0.001). Also the SUS scores of the ABC and QWERTY keyboards revealed a statistically significant difference (p0.05). We can, there- fore, conclude that the QWERTY keyboard was preferred by the participants in terms of perceived

51、usability, followed by the ABC keyboard. The 5Button keyboard was rated fairly negative, compared to the other input mechanisms.In a second step, we investigated the distraction and work-load caused by the keyboards in the dual task condition, i.e. typing in cities while driving. We computed the mea

52、n lane deviation over all driving conditions (2 baseline + 3 text input). A repeated measures ANOVA with a Greenhouse- Geisser correction revealed a statistically significant effect of the experimental condition on the mean lane deviation (F(2.277, 66.033) = 22.947, p 0.001). When conductingpost hoc

53、 tests using the Bonferroni correction, we found sta- tistically significant differences between each baseline condi- tion and the dual task conditions (p0.01), but not between the different text input modalities. We can, therefore, con- clude that the driving performance in terms of mean lane devia

54、tion was negatively affected by all navigation desti- nation input tasks, but that no statistically significant dif- ferences were found between the different input modalities. This finding shows that the distraction from the road was equal over all three input modalities. This could also be explain

55、ed by users dedicating a constant amount of their re- sources over all dual task conditions to driving. Differences in the qualities of the systems would then become visible in the amount of cities, which were successfully typed into the system over the period of three minutes.Therefore, we computed

56、 the amount of cities, that were typed in during the three minute period of driving the lane change task. A repeated measures ANOVA revealed a sta- tistically significant effect of the keyboard layout on the amount of cities completely typed in three minutes (F(2, 58) = 411.120 p 0.001). On average,

57、 most cities could be typed in with the QWERTY layout, followed by the ABC layout and the 5Button keyboard. A post hoc test revealed that these differences were highly significant (p0.01).6.RESULTSIn total, 30 participants took part in the study in total (21 female, 9 male). On average, the particip

58、ants were 32 years old (SD: 10) and possessed a drivers license for 13.03 years (SD:10). Participants reported that they drove on average 9647 (SD: 11819) kilometers per year, ranging from 100 to 50000 kilometers.In a first step we computed the time needed to type in five cities, see Table 1. To com

59、pare the mean time to completion over all three keyboards, a repeated measures ANOVA with a Greenhouse-Geisser correction was computed and revealed a significant effect of the input modality on the time needed to complete typing five cities (F(1.021, 58) = 83.874 , p0.001). Typing five cities with the QWERTY keyboard took less tim