The UCSB Personal Guidance System

A research team at University in California Santa Barbara has in a period of several years developed a personal guidance system that utilize spatial audio in route navigation for visually impaired users [?????]. The prototype they developed has many similarities with the Sound Guide prototype developed during our study. We were not aware of this research during development and user testing, but since the prototypes share so much virtual sound display could provide more directly perceptual information of the environment.

25http://www.tracab.com/

26A larger set of directions was attempted, but performance issues limited the number to eight

The long term goal of the research group is to ”contribute to the development of a portable, self-contained system that will allow visually impaired individuals to travel through familiar and unfamiliar environments without the assistance of guides.” [?, p 196].

The prototype developed was not intended to replace mobility aids such as white canes and guide dogs, but rather help in the more general navigation and orientation challenges. The prototype was developed with a user-centered approach where iterative development was guided by test results with visually impaired users. The system that was used in ? will here be presented.

Concept

The system was designed to guide participants through a route that consisted of several waypoints positioned at corners and intersections.

The prototype was used to guide the visually impaired participant in the direction of the next waypoint. When the position of a waypoint was reached, the next waypoint in line was presented. A waypoint was considered reached when the user was within 2.1 meters of its position.

The orientation of the waypoint with respect to the listener was presen­

ted through sound. Two spatial displays were used in the experiment. One was a haptic pointer interface (HPI) where the user was expected to scan the environment with the use of a hand-held pointing device. If the HPI was pointed within 10^◦ of the next waypoint, a fast beeping sound was heard to confirm the direction. The user could then continue in that dir­

ection with the assurance that he/she was heading in the right direction if the beeping sound was heard. The other spatial display used spatial au­

dio filtering to convey direction. In the virtual sound condition a rapid beeping sound was spatial filtered to appear from the direction of the next waypoint.

Both displays was turned on and of with the help of a single button. When the button was pressed the sounds conveying directional information could be heard, when the button was released no directional information was given. The distance to the next waypoint was in addition read aloud of a synthetic voice every eight second.

System

The system was run on a notebook computer carried in a backpack. The prototype used a Trimble 12-channel differential GPS receiver that provided an accuracy of position that was below 1 meter during the tests [?]. This accuracy is way better than the accuracy of GPS-receivers in more portable devices, but the researcher justified their choice of a ’bulky system’ with the hope that smaller systems with the same accuracy eventually will become commonplace. The whole system weighted 2.5 kilograms.

A small magnetic sensor to detect orientation (compass sensor) was positioned at the hand-held pointer to accommodate scanning when the HPI spatial display was used. The 2006 article does not specify where

31 2.4. PROJECTS UTILIZING SPATIAL AUDIO

they positioned the compass sensor when using the virtual sound display.

Earlier experiments positioned this either mounted in front of the users waist, or on top of the headset [?]. If the sensor is positioned on the headset all head rotation will lead to an update of the virtual sound environment.

When mounted on the body a change of body orientation is required to update the virtual sound environment.

Study

The experiment had several goals. Earlier test had only been conducted on open places and not in urban areas that better reflect visually impaired pedestrian everyday routes. One route was therefore set to a urban area with intersections, sidewalks and streets. Another route was set in a more open area in a park with curved paths. The routes was intended to be traversed without aid from the researchers, where the participants navigated with the UCSB Personal Guidance System and a traditional white cane to help in avoiding obstacles.

The researchers was interested in seeing how often people would choose to use the spatial display and when they chose to do so. Would a more structural environment lead to less use of the system, than in a less structural environment as a park?

The third goal was to compare the two different spatial displays, that was designed based on earlier tests of a larger variety of displays.

Eight subjects that were legally blind was tested and a total of 30 trials was completed. The two routes was conducted with both the HPI and the virtual sound display. The time to complete each route, the amount of time with directional guidance, errors made and the distance travelled were recorded.

Results

Both spatial displays proved very successful. All participants in the study manage to follow the routes and reach the target in all trials without the need for outside assistance [?]. To determine how accurate they were in following the route, actual travel distance was compared to an ideal travel distance. The mean extra distance was 18%, but as low as 10% in one third of the trials. This indicate that the participants manage to follow directional information with high accuracy. One participant managed to walk past a waypoint, but discovered it and turned around to regain the path. Another participant got of course and got trapped in a bike rack. Apart from these incidences the UCSB Personal Guidance System proved to be of great use.

Street Park

Virtual Sound HPI Virtual Sound HPI

406.7 438.5 344.3 427.5

Table 2.1: Mean times in seconds of all conditions (SD: 95.5-138.4) from ?

The mean time of route completion for all conditions is summarized in table 2.1. The use of the virtual sound displayed proved to lead to significantly shorter route completion than did the HPI display (t(7) = 1.80, p = .058).

The total time the spatial displays was accessed show that the parti­

cipants requested more directional guidance when in ”semi-bound” envir­

onments like the park, compared to the more ”bounded” environment of the street (t(₇) = 4.83, p < .001). The amount of ”travel skill” seem to affect how much directional guidance the participant request. Those participants who had the shortest completion time on each route also requested less directional guidance (t(₆) = 0.73, p < .05).

The participants reported great satisfaction with the system. They said they wanted this types of spatial displays in commercial GPS-aids and that they would have used the directional guidance features when needed.

They reported the spatial displays to feel accurate in giving directional information, that they felt it to be safe and easy to use. The virtual sound display was reported to more easy to use, but slighter less safe. The prototype gave distance information every eight second. When asked how often they wanted to hear such information the responses given were between 5 and 10 sec. The participants wanted the ability to change this frequency on demand.