Vibration Perception in Mobile Contexts Idin Karuei* Zoltan Foley-Fisher† Sebastian Koch‡ Russ MacKenzie§ Mohamed El-Zohairy¶ Karon E. MacLean║ Department of Computer Science, University of British Columbia, Vancouver, Canada significant and ignoring it will harm the effectiveness of tactile ABSTRACT user interfaces. Human sensitivity to vibration declines in mobile contexts. Designers of wearable haptic systems must account for the effects 3 METHODS of movement and distraction so that tactile display information is 16 volunteers (8 male) took part. The participants perceived consistently. We compared the sensitivity of seven (counterbalanced for gender and condition) sat in a chair in one body sites in simulated mobile contexts, and found that the thigh condition, and walked on a treadmill in the other. A tall chair is least and the wrists the most sensitive of the sites tested. maintained a consistent view of the screen between conditions. We attached thirteen vibrotactile displays to seven body sites KEYWORDS:Vibration, Sensitivity, Mobile Contexts, Movement, corresponding to wearable sites based on past studies and Distraction, Wearable Haptics. potential practicality: chest (left and right, directly below the collar bone), spine, outer thighs, stomach (left and right, halfway INDEX TERMS:H.5.2 [Information Systems]: User Interfaces – between navel and hip bone), feet (on the top surface of the foot), Haptic IO. wrists, and upper arms. During half of the trials in each condition, participants 1 INTRODUCTION performed a workload task targeting vision, memory, and Many body sites have been considered for wearable tactile attention as shown in Figure 1. A screen four meters wide and displays, and vibratory information has improved performance of three meters high displayed twenty-five blocks bouncing slowly pilots and drivers. It is well known that some body sites are less around a three-dimensional room; one block was highlighted and sensitive than other areas, e.g. back versus wrist – a function of participants counted the times the highlighted block hit any of the skin type and sensor density and composition. When a body part walls. The task was chosen for its controllable continuous is in motion, it becomes less sensitive to stimuli [1] and vibration workload typical of normal pedestrian activity, with distraction patterns may be misinterpreted or undetected. For wearable haptic adjusted so that participants would not fall off the treadmill. systems, often used in ambulatory situations this is especially Participants reported their collision count at the end of each troublesome. Merely increasing vibration intensity is workload condition. unsatisfactory due to concerns of power and comfort. In this paper we tackle unpredictable vibration sensitivity by seeking body sites that are less susceptible to changes in sensitivity, comparing diverse sites (a few studied elsewhere) in a single study. 2 RELATED WORK Wearable tactile systems have been the focus of many papers in the last decade due to its variety of applications. Ertan et al. introduced a wearable navigation system for guidance of blind users in unfamiliar indoors areas [2]. They used a vibrotactile display consisting of a 4-by-4 array of micromotors embedded in the back of a vest to communicate a stop signal or the four cardinal directions to the user. Bosman et al. developed a wearable haptic guidance system that could be attached to both wrists of a pedestrian to guide him inside unknown buildings [3]. Tsukada and Yasumura developed a belt with eight vibrotactile haptic displays to guide a pedestrian towards destinations, predefined locations, or valuables left behind [4]. Subjects could feel vibrations when stopped but often failed to recognize Figure 1: Experiment setup. Subject responds to vibrations while vibrations when walking; they could stop for a moment to walking on the treadmill and doing the visual task. recognize the direction of the vibration. This suggests that the effect of movement on detection of tactile stimuli which has been During all conditions, participants pressed the right-hand button studied in the field of neural psychology [5][6] is in fact on a modified computer mouse when they detected vibration from any tactor. Vibrations were presented in randomized sites, intensities in randomized order, and the interval between 500 ms- * e-mail: idin@cs.ubc.ca duration tactor vibrations was randomized between four and six † e-mail: zoltan@ece.ubc.ca seconds. Reactions later than 3500 ms were discarded. ‡ e-mail: skoch@cs.ubc.ca § e-mail: rmacken1@cs.ubc.ca 4 RESULTS ¶ e-mail: zohairy@cs.ubc.ca As the dependent variable (detection rate) is dichotomous, we ║ e-mail: maclean@cs.ubc.ca performed a logistic regression on five factors: Intensity, Task, 28 Movement, Body Site (within-subject) and Gender (between- subjects). Gender, Task, Movement, and Body Site (with spine as a reference point) are categorical variables. The omnibus test of the model coefficients is significant (p < 0.001). The regression results are listed in Table 1, where main effects Gender, Intensity, Movement and Body Sites are seen to be statistically significant. Table 1: Results of logistic regression B S.E. Wald df Sig. Exp(B) Gender(1) .215 .064 11.158 1 .001 1.240 Intensity 1.692 .036 2195.889 1 .000 5.429 Task(1) .054 .064 .700 1 .403 1.055 Movement(1) 1.778 .071 625.023 1 .000 5.919 BodySite 649.684 6 .000 BodySite(1) -1.186 .121 96.079 1 .000 .305 BodySite(2) .878 .125 49.060 1 .000 2.407 BodySite(3) -.972 .121 64.962 1 .000 .378 BodySite(4) -2.086 .125 279.697 1 .000 .124 BodySite(5) .096 .121 .622 1 .430 1.100 BodySite(6) -.102 .121 .715 1 .398 .903 Figure 3: Sensitivities for five intensities across the four conditions. Constant -2.651 .117 511.428 1 .000 .071 All body sites are negatively affected by movement, but some The six Body Site levels in the table are Foot, Wrist, Stomach, sites more than others, as illustrated in Figure 3. Thighs, and to a Thigh, Chest, and Arm; Spine is the reference level. Foot, Wrist, lesser extent feet and stomach are particularly strongly affected. Stomach, and Thigh are significantly different than Spine; Figure These are also areas of motion: the feet can feel heel strikes on the 2 further shows that the Wrist is more sensitive than the Spine, treadmill surface, while the stomach undergoes twisting motions and the Foot, Stomach, and Thigh are less sensitive. as the arms swing. 5 CONCLUSION The results of our experiment confirm the effect of body motion on detection of vibrations. We discovered that movement in a typical mobile context (i.e. walking) affects detection of vibrations on the thighs more than other body sites. Also, reaction times to vibrations are significantly reduced during walking. However, it appears that visual distraction in a mobile context may not have a significant effect on detection of vibration on any body site. In general, the thigh is not suited for applications that require discriminating among vibration patterns in everyday wearable haptics. This may be of interest to cell phone users who typically receive vibration notifications on the site most Figure 2: Detection rates at different body sites. susceptible to movement effects. On the other hand, the data suggest that the chest, upper arm, and wrist are sufficiently For Gender, males show a slightly higher detection rate of sensitive to lower energy vibrations while the body is in motion. 65.3% compared to 63.0% for females. The Movement factor showed an important result: participants detected 73.9% of stimuli REFERENCES when sitting, but only 54.4% while walking. As suggested by the [1] C. E. Chapman, M. C. Bushnell, D. Miron, G. H. Duncan, and J. regression results, Task did not show significant differences; the Lund. Sensory perception during movement in man. In Experimental detection rate with and without the visual distraction task was Brain Research, vol. 68, pages 516-524, 1987. 64.4% and 63.8%, respectively. Predictably, there were strong [2] S. Ertan, C. Lee, A. Willets, H. Tan, and A. Pentland. A wearable results for Intensity: at intensity 4 (the strongest) almost all stimuli haptic navigation guidance system. In Digest of the Second were detected, while at intensity 0 (weakest) only 16.7% were International Symposium on Wearable Computers, pages 164-165, detected. 1998. Interaction effects are difficult to analyze using regression, so [3] S. Bosman, et al. GentleGuide: An exploration of haptic output for we will present these results graphically rather than with tests of indoors pedestrian guidance. In Lecture Notes in Computer Science, significance. There is a strong interaction between movement and pp. 358-362, 2003. intensity, as shown in Figure 3. At the highest intensity there is no [4] K. Tsukada and M. Yasumura. Activebelt: Belt-type wearable tactile difference between movement conditions, while at lower display for directional navigation. In Lecture Notes in Computer intensities the detection rate is much lower while walking. Science, vol. 3205, pp. 384-399, 2004. [5] R. W. Angel and R. C. Malenka. Velocity-dependent suppression of cutaneous sensitivity during movement. In Experimental neurology, vol. 77, pp. 266-274, 1982. [6] L. Post, I. Zompa, and C. Chapman. Perception of vibrotactile stimuli during motor activity in human subjects. In Experimental Brain Research, vol. 100, pp. 107-120, 1994. 29