Making an Impression: Force-Controlled Pen Input for Handheld Devices

CHI 2005 | Late Breaking Results: Posters

April 2-7 | Portland, Oregon, USA

Making an Impression: Force-Controlled Pen Input for Handheld Devices Sachi Mizobuchi1, Shinya Terasaki1, Turo Keski-Jaskari2, Jari Nousiainen2, Matti Ryynanen2, Miika Silfverberg2 Nokia Research Center {sachi.mizobuchi, shinya.terasaki, turo.keski-jaskari, jari.nousiainen, matti.ryynanen, miika.silfverberg}@nokia.com 1 2 Arco Tower 17F, 1-8-1 Shimomeguro, Meguro, FIN-00045 Nokia Group Finland Tokyo 153-0064 Japan RELATED WORK

ABSTRACT

Research on human ability of force (weight) discrimination has a long history in the field of human perception, beginning with early psychophysical studies by Weber [2]. He found that the smallest just noticeable difference (JND) in stimulus energy is a constant fraction of the intensity of the stimulus. This relationship is known as Weber’s law (later extended over a range of sensation to become the logarithmic relationship between physical intensity and psychological sensation that is predicted by Fechner’s law).

The properties of force-based input on a handheld device were examined. Twenty-one participants used force input to set 10 different target levels representing consecutive force ranges (0 to 4N) with visual feedback (digits or bar graphs) or no feedback. Both accuracy and speed were greater with analog feedback (bar graph). Statistical comparisons of adjacent targets/digits indicated that subjects differentiated roughly seven input levels within the set of ten force ranges actually used. Time taken to input the target force increased significantly with the size of the target force, suggesting that smaller force ranges should be considered in future implementations of force input. The results are discussed in terms of the design of appropriate feedback for force input.

There has been relatively little research on human ability to control force. One exception is the work of Srinivasan & Chen [3] who examined force tracking against a rigid object, using the index finger. In their experiment, participants were asked to control the force applied to a force sensor using their index finger pad, under a number of different conditions (including an anesthetized fingertip to examine the effect of removing tactile feedback). From their results, they suggested that haptic interfaces need to have a force resolution of at least 0.01N in order to make full use of human haptic capabilities.

Author Keywords

Force sensitive touch screen, handheld device, pen user interface ACM Classification Keywords

H5.2 User Interface: Input devices and strategies I.3.6 Methodology and Techniques: Interaction techniques

Recent research by Ramos et al. [4] focused on force control with a pen user interface. They investigated human ability to perform discrete target acquisition tasks by varying stylus force. They used four different selection techniques and found that their “Quick Release” selection technique was preferable overall. They suggested that the force space could be divided into 6 levels. However, since their findings were based on force values provided by a specific device (Wacom Intuos tablet), without showing correspondence to quantified levels of physical force, it is difficult to generalize their findings to other devices. Furthermore, their experiment was done with a desktoptype tablet, and thus the properties of force input on a handheld device have yet to be explored.

INTRODUCTION

Force control provides users with a natural form of manual input to control analog information. Current commercial products that use force input include game controllers (e.g. the Dual Shock2 for the Sony PlayStation2) and drawing tablets (e.g., the Wacom Intuos). It is also suggested that using force sensing for authentication [1]. The control of force should also be useful with pen-based handheld devices. It is a potential space saving technique that can be used for selecting menus, text/number input, changing windows, zooming/panning, etc. This research aims to clarify requirements for force control- based interaction design.

Raisamo [5] compared 3 different force-based manipulations with direct manipulation and time-based manipulation on an information kiosk with a force-sensitive screen. In his study, one of the force-based manipulations (a non-linear technique that directly mapped the force level

Copyright is held by the author/owner(s). CHI 2005, April 2–7, 2005, Portland, Oregon, USA. ACM 1-59593-002-7/05/0004.

1661

CHI 2005 | Late Breaking Results: Posters

April 2-7 | Portland, Oregon, USA

(analog feedback). In the Number condition, target level was shown on the left of the screen. As participants pressed in the red circle, the number in the right changed based on the pressed force level (stronger force produced a higher number). When the pressed force exceeded the range of level 10 (more than 4.1N), the number on the right changed to “11”. The task for the participants was to set the number on the right (using force input) to the target number shown on the left. When the number corresponding to the force input remained at the same level for one second, the test program gave visual (the circle changed to white) and audio feedback, and saved the corresponding force value. After a 2 second pause, the next target level/number was shown. The target number for each trial was selected randomly from the digits 1 through 10.

to the size of the selection area) was ranked as being as good as the direct manipulation technique, even though it was slower than that technique. Thus, force-based interaction techniques can provide alternative interaction styles in future products, if designed appropriately. EXPERIMENT Objectives

An experiment was conducted to study properties of force input on a handheld device relevant to design of input methods. The following questions were addressed; 1) How accurately/quickly can people control the force they exert with a pen on a force sensitive screen of a handheld device? 2) How does visual feedback affect the performance, and what happens when the type of visual feedback changes? 3) How many different force ranges can be used in interacting with a handheld device, and how should they be distributed across the continuum of possible forces?

In the Bar condition, a bar divided into 11 ranges was displayed on the screen and a target level was shown in red text under the bar. As participants pressed inside the red circle, the white bar extended from left to right based on the force level pressed (the bar moved to the right as participants pressed more strongly). When the pressed force exceeded the maximum (level 10) target range, the edge of the white bar moved to the right of the number 10 on the scale. The task for the participants in this condition was to set the right edge of the bar to the target area/number and to keep it there for one second.

Participants

Twenty-one (14 males, 7 females) employees and trainees from the same company, ranging in age from19 to 40 (mean 32.2, SD=5.6) participated in the experiment. Eighteen of them were right-handed and three of them were left-handed. Apparatus

The testing device had a 640 x 320 pixel LCD covered with a resistive touch panel. Four micro force sensors (HFD-500, manufactured by Hokuriku Electric Industry in Japan) were located under the edges of display, each capable of sensing a force of up to 5N. Force on the point being pressed on with the pen/stylus was calculated based on the values of the four sensors. The process of calculating the force based on the sensor data was calibrated prior to the experiment, so that it corresponded to the actual physical force in Newtons. The device was covered with a plastic case (W121 x H69 x D27.5mm) and connected to a laptop PC (IBM Thinkpad X24) with RS-232C and DVI connectors.

Figure 1. Visual feedback conditions. (left) Number condition. (right) bar condition. Procedure and Design

Participants did the experiment individually. They sat in a quiet room holding the testing device in their non-dominant hand, with a pen in their dominant hand. The test session had 2 parts; a) pressing with visual feedback conditions and b) pressing with a no-feedback condition. First, participants performed the target acquisition tasks under the two visual feedback conditions. Half of them started with the Number condition, and the other half started with the Bar condition. In each condition, they had as many trials (and at least 10) of practice as they needed, until they felt that they were completely familiar with the task. Participants were asked to complete two blocks in each of the two visual feedback conditions, with 50 trials per block. During the blocks, and after each condition, participants were asked to provide the following feedback: 1) If there were particular target level(s) they felt it was harder to set than others; 2) their preference for type of visual feedback and the reason; 3) their judgment of how fatigued they felt. After completing two visual feedback conditions, the participants were asked to acquire the target under the Number condition without being shown feedback. A fully within-subjects factorial

The testing program was run on the laptop PC and the interface viewed by participants in the experiment was displayed in landscape mode on the handheld testing device using a multi-display function. The force data input by participants during each trial was logged on the laptop PC. The input device was the stylus for an iPaq h 3630. That stylus had been found to be most preferred among various available designs of styli for PDAs in our previous testing. It was 109.5mm in length and weighed 5.94g. Tasks

Ten force ranges were used as target levels in a series of target acquisition tasks. Each target level spanned a range of 0.41N (thus level 1 was 0.41N or less, level 2 was from more than 0.41to 0.82N, and so on, with level 10 containing forces greater than 3.67N up to 4.08N. We used two different visual feedback conditions: a Number condition (discrete feedback), and a Bar condition

1662

CHI 2005 | Late Breaking Results: Posters

April 2-7 | Portland, Oregon, USA

was disturbing during the task.” or ”Bar feedback made me feel that I have to set the edge of the bar exactly in the middle of the range”. There was no significant difference in performance between the feedback conditions for this smaller group of six participants (Figure 5 (b) F < 1 for both speed and accuracy). No obvious relationship was discernible between personal characteristics (age, gender, nationality, experience of pen-based device) and which of the two feedback preference groups the participants belonged to.

design with repeated measures was used. The independent variables were visual feedback types (Bar/Number/NoFeedback) and Target levels (1-10). The dependent variables were target acquisition time and error rate. RESULTS Accuracy and speed by conditions

Figure 2 shows the mean error rate of each target level by conditions. Error rate in the no-feedback condition was higher, and especially so for the higher target levels. When feedback was provided, the Bar condition was more accurate than the Number condition (F[1,20]=11.67, p