The Procedure

In initial experiments, vibration waveforms, durations and frequencies will be determined empirically for each device and held constant. The amplitude will be varied and the subjects will be asked to press a button when they detect a stimulus. The tests will be conducted for standing, walking and jogging. Amplitudes and activity levels will be randomized to compensate for learning effects and fatigue. As described above, experiments are being conducted to determine how well people can perceive various types of haptic stimuli while they are walking and jogging on a treadmill. Preliminary results with vibrotactile stimulation are consistent with our hypothesis that subjects are better at perceiving this type of feedback when they are stationary.

Figure A shows the fraction of stimuli that were detected while standing, walking and jogging by 8 normal subjects with a vibrotactile stimulator placed on their arm. The stimuli were 250 Hz sine waves bursts with 1 second duration. The control variable was the stimulus amplitude. Eight amplitudes were presented three times each in random order with random spacing. Figure A only show the lowest 5 levels as 100% were detected at the higher levels. Particularly at the lower stimulation levels, subjects were not able to detect the stimuli as regularly while they were moving and jogging. Figure B shows that the subjects’ response time is also longer when they are active, particularly at lower stimulus amplitudes.

With the stimulator placed on the leg, similar trends were found but with less consistency. We hypothesized that the reduction in the subjects’ ability to detect the vibrotactile stimulation was due primarily to the disturbance accelerations present while walking and jogging, which are greater on the leg than the arm. Also, these accelerations depend on where in the gait cycle the stimulus is presented. Because the timing of the stimulus relative to the gait cycle was not controlled in the experiment, we designed another experiment to investigate this issue. Shorter stimuli were presented at four times in the gait cycle; early stance (just after heel-strike), late stance (just before toe-off), early swing (just after toe-off) and late swing (just before heel strike). An accelerometer placed on the leg was used to detect markers indicative of the gait cycle phases.

As can be seen in Figure C, the fraction of stimuli detected was lowest in late swing and early stance, which are both near the event of heel strike where disturbance accelerations are largest. Figure D shows that the response time is longest when the stimulus is presented in early stance, just after heel strike. While stimuli presented in late swing had a low detection rate as shown in Figure C, the response time corresponding to those stimuli that were detected was relatively short. This indicates that the stimulus was likely detected before heel strike or not at all. While these results are preliminary, they seem to indicate that bursts of vibrotactile information should not be presented near the event of heel strike if possible.

Figure A shows the fraction of stimuli that were detected while standing, walking and jogging by 8 normal subjects with a vibrotactile stimulator placed on their arm.
Figure B shows that the subjects’ response time is also longer when they are active, particularly at lower stimulus amplitudes
Figure C shows that the fraction of stimuli detected was lowest in late swing and early stance, which are both near the event of heel strike where disturbance accelerations are largest
Figure D shows that the response time is longest when the stimulus is presented in early stance, just after heel strike