Audible Beaconing with Accessible Pedestrian Signals

Abstract

Purpose

Although Accessible Pedestrian Signals (APS) are often assumed to provide wayfinding information, the type of APS that has been typically installed in the U.S has not had positive effects on finding crosswalks, locating pushbuttons, or providing directional guidance. This paper reports the results of research on crossings by blind pedestrians at complex signalized intersections, before and after the installation of APS with innovative audible beaconing features, designed to improve wayfinding.

Methods

Objective data on measures of street crossing performance by 56 participants was obtained at four intersections, two each in Charlotte, NC, and Portland, OR.

Results

In the first round of testing, APS with beaconing features resulted in only slightly improved wayfinding. Revisions to the audible beaconing features resulted in improved performance on four measures of wayfinding as compared to the pre-installation condition: beginning crossings within the crosswalk, ending crossings within the crosswalk, independence in finding the starting location, and independence in aligning to cross.

Implications for Practice

Use of APS that provide beaconing from the far-end of the crosswalk show promise of improving wayfinding at street crossings.

INTRODUCTION

Accessible or audible pedestrian signals (APS) have been installed around the world for many years. In some countries, such as Japan and the United States, the audible indication has been provided from an overhead speaker (pedhead-mounted) aimed across the street, intended to provide directional or wayfinding information to pedestrians during the crossing. In Australia, Sweden and other countries, a different type of APS has been used, which provides sounds from speakers at the pushbutton location. These pushbutton-integrated APS have not generally been expected to provide directional information during street crossings, although they include a locator tone that may help people find the crosswalk and home in on the opposite end of the crosswalk as they near it; the housing includes a tactile arrow which indicates the direction of the crossing actuated by the pushbutton.

Pedhead-mounted APS, as typically installed in the US, Japan, and Canada, have not been found to provide good directional guidance (Bentzen, Barlow & Franck, 2000; Carroll & Bentzen, 1999; Szeto, Valerio, & Novak, 1991; Uslan, Peck & Waddell, 1988; Wall, Ashmead, Bentzen & Barlow, 2004). Several researchers have evaluated modified pedhead-mounted APS, comparing simultaneous sounds provided from both ends of the crosswalk at the same time (typical installation), sounds alternating between ends of the crosswalk, and/or sound from only the far end of the crosswalk (Larouche, Leroux, Giguere, & Poirier, 2000, Poulsen, 1982; Stevens, 1993; Tauchi, Sawai, Takato, Yoshiura, & Takeuchi, 1998; Wall, et al. 2004). Far-end only signals resulted in more accurate crossings (Poulsen, 1982; Wall, et al., 2004) as measured by deviation from a straight line while crossing. Results on alternating signals were mixed (Larouche, et al., 2000; Stevens, 1993; Wall, et al., 2004). A pushbutton locator tone during the last half of the crossing improved crossing accuracy at simulated crosswalks (Wall, et al., 2004). These findings provided the basis for development of APS for this project.

This research is part of a multi-year project to examine the effectiveness of optimized APS for providing street crossing information to pedestrians who are blind. In a pre-installation phase, data were collected in three cities. Blind pedestrians crossed at two complex signalized intersections in each city without accessible pedestrian signals. The findings confirmed that pedestrians who are blind have considerable difficulty independently locating crosswalks, aligning to cross, determining the onset of the walk interval, ending their crossing within the crosswalk, and completing crossings before the onset of traffic perpendicular to their path of travel at complex, unfamiliar, signalized intersections without accessible pedestrian signals (Bentzen, Barlow, & Bond, 2004; Barlow, Bentzen, & Bond, 2005).

This paper presents research on wayfinding and orientation task performance prior to and following installation of pushbutton-integrated APS with audible beaconing features in two of these cities, Portland, OR, and Charlotte, NC. The third city is not included in this analysis because a different type of APS was installed there. Innovative audible beaconing features, providing tones for wayfinding from directional speakers, were successively refined and evaluated. Results related to crossing timing decisions before and after installation of APS were reported in a previous paper (Scott, Barlow, Bentzen, Bond, & Gubbe, 2008).

METHODS

Overview

In each city, each participant traveled four short routes, each requiring two or three crossings, for a total of nine crossings at the two intersections. In pre-installation testing in Portland, visual pedestrian signals were present for all crossings, with pushbuttons for five of the nine crossings. Post-installation, pushbutton-integrated APS were installed at seven crossings, with audible beaconing signals installed for 4 of those crossings. In Charlotte pre-installation testing, visual pedestrian signals were present for eight of the nine crossings, with pushbuttons for four of those crossings. Post-installation, pushbutton-integrated APS were installed at eight of the nine crossings, with audible beaconing installed for four of those crossings. Data were collected for several variables associated with each crossing subtask.

This paper reports only the wayfinding results associated with those crossings where audible beaconing features were installed. These were the crossings for which blind participants were observed to have the most difficulty with one or more of the wayfinding tasks measured during the pre-installation testing.

Materials

Intersections

Intersection geometry can be seen in Figure 1. The intersections were chosen in consultation with city staff, as examples of intersections with complex signal phasing or geometry. This paper focuses on crosswalks 1, 5 and 7 in Portland and crosswalks 6, 7 and 9 in Charlotte. On some crosswalks, participants crossed in two directions on the same crosswalk, depending on the intersection and route, and those are referred to in the results as different crossings. APS with beaconing features were added at these crossings without any change to intersection geometry or curb ramps, although new poles for the APS were added in some locations. If possible, the APS pushbutton was installed on the side of the crossing furthest from the center of the intersection; pole locations are shown on Figure 1. Beaconing speakers were installed on top of the visual pedestrian signals and aimed at the center of the crosswalk. Signal timing was the same both pre- and post-installation and met the requirements of the Manual on Uniform Traffic Control Devices (2003).

FIGURE 1.

Intersection diagrams, Portland (A, B) and Charlotte (C, D). This paper reports wayfinding results for crosswalks 1, 5, and 7 in Portland, and 6, 7 and 9 in Charlotte.

APS Devices

All of the APS were pushbutton-integrated, having pushbutton locator tones that repeated once per second and tactile arrows that vibrated during the audible walk indications and were oriented in the direction of the crossing. Volume of the pushbutton speaker was carefully adjusted so the locator tone and walk signal were normally audible only within 6–12 feet of the pushbutton. On this type of APS, volume adjusts continuously in response to ambient sound levels, and usually fluctuated between 30 and 80dB. In Portland, the audible walk signal was a rapid tick (approximately 10 times per second). In Charlotte, the walk indication was a speech message, followed by the rapid tick (e.g., “Kings, Walk sign is on to cross Kings, tick, tick, tick, tick.”). An additional speaker was attached to the pedestrian signal head to provide the audible beaconing (see Figure 2).

FIGURE 2.

APS installation: Left, pushbutton unit with integrated speaker; Right, beaconing speaker on top of pedestrian signal head.

In round 1, post-installation, the audible beaconing was provided simultaneously from pedhead-mounted speakers at both ends of the crosswalk, and was actuated by a pedestrian request (i.e., holding the pushbutton in for more than three seconds (Portland), or more than one second (Charlotte). In Portland, when audible beaconing was actuated, both the walk tone and the locator tone during the next pedestrian phase (during the walk and flashing don't walk signals) were elevated to a maximum volume (89 dB) from speakers aimed toward the center of the crosswalk. The louder sounds came only from the pedhead-mounted speakers; the pushbutton speakers were silent during the pedestrian phase if audible beaconing was actuated.

In Charlotte, there were two post-installation rounds of data collection with somewhat different beaconing features. In post-installation round 1 (Post-1), the walk indication and locator tone at elevated volume came simultaneously from pedhead-mounted speakers at both ends of the crosswalk, with the sounds also provided at normal (relatively quiet) volume from the pushbutton speakers. In Charlotte post-installation round 2 (Post-2), the software and wiring of APS was modified to add other features. When audible beaconing was called, volume at the pedestrian's starting location was not increased, but the locator tone was provided at maximum volume (110dB) during the subsequent flashing don't walk from the pedhead-mounted speaker on the end of the crosswalk opposite the location where the pushbutton was held (far-end). Far-end beaconing is technically more difficult, requiring additional wiring and controller modifications. Walk indications and locator tones were provided at normal volume (ambient sound responsive) from the pushbutton speakers at both ends of the crosswalk. Immediately after the extended button press during the flashing don't walk or don't walk signal, a pushbutton information message provided street names. This was followed by an “orientation tone”--seven repetitions of the locator tone at maximum volume from the pedhead-mounted speaker at the far end of the crosswalk.

Participants

Sixteen participants who reported their visual acuity as no light perception or light perception only participated in each test phase. During the post-installation testing sessions, half of the participants in each city were new to the study, while the other half had participated in an earlier testing session. This allowed practice effects to be evaluated as the cause of observed improvements and provided an adequate sample size from a limited population. There were a total of 56 different participants in these two cities. All participants were accustomed to crossing independently at signalized intersections using a long cane or dog guide and were unfamiliar with the intersections used in the study. Participant demographics were similar for the various testing phases, and similar across the two cities. Overall, 30 men and 26 women participated; ages ranged from 20 to 78, and mean ages for each testing session ranged between 44 and 48. In each group of sixteen, 11 to 14 participants used a long cane while others used dog guides.

Procedure

Participants were tested individually for approximately 1.5 hours, and traveled two routes at each intersection within their city. Order of intersections and routes was systematically varied. Guided approaches to all routes avoided the experimental crossings, thus avoiding learning immediately prior to each trial. Participants were accompanied at all times by an orientation and mobility (O&M) specialist who communicated instructions and was responsible for participant safety. Another researcher recorded observations and measured response times using a digital stopwatch.

At the beginning of all routes, participants were asked to assume that they needed to get to the other side of the intersection for an appointment. They were instructed to “cross the street in front of you, the perpendicular street, then cross the street beside you, the parallel street”, using their usual travel aid and techniques to accomplish the task. They could request assistance from the researcher with all or any part of the crossing task, except the use of the APS, if they would typically request assistance from another pedestrian. The researcher immediately provided requested assistance with any crossing subtask (locating the pushbutton, locating the crosswalk, aligning to cross, determining when to start crossing) or with the entire crossing.

While participants were locating the crosswalk and aligning, researchers only intervened when the starting location and alignment would result in participants crossing the wrong street, or at a clearly hazardous location or direction. Intervention occurred when starting crossing or while crossing the street only when participants were in, or stepping into, the path of moving vehicles. No information about the intersections was provided to the participants. Prior to pre-installation testing, participants were told that some crossings would have pushbuttons. Prior to post-installation trials, participants were told that APS had been installed at each intersection, but might not be installed for each crossing. Participants were thus unaware of which crossings would have pushbuttons or APS available. They were shown a demonstration model of the APS pushbutton and told that, where APS were installed, pushbuttons would have pushbutton locator tones, tactile arrows, and audible and vibrotactile walk indications, and the features were described. Participants were told that if they held pushbuttons in and heard a confirmation tone or message, audible beaconing would be provided and the beaconing feature was described. Finally, participants heard a recording of the locator tone, confirmation tone or message, and WALK indication, and were invited to ask questions. For the Post-2 installation, piloting indicated the need for on-site familiarization, so participants were familiarized with the pushbutton information message, the orientation tone, and the walk indications and beaconing locator tones at a crossing that was not on the experimental routes.

RESULTS

General

The results reported here are restricted to wayfinding measures (and the associated measures of independence) for only those crossings at which audible beaconing features were installed. Measures of wayfinding included starting within the crosswalk, starting from an aligned position, and ending within the crosswalk, as well as independence in each of these tasks. Independence was measured by requests for assistance or the need for intervention on each task (see Table 1). Additional results, including those related to timing measures, have been reported elsewhere (Scott, et al., 2008).

TABLE 1.

Comparison of Average Wayfinding Results Between Pre-and Post-APS Installation for Four Crossings in Each City^†

	Portland		Charlotte
	Pre-Install	Post-Install	Pre-Install	Post-Install Round 1
Started within crosswalk	79.2	100.0***	72.4	78.8
Found starting location independently	93.8	100.0	82.8	96.2
Started correctly aligned	68.8	81.0	51.0	48.5
Aligned independently	94.3	100.0	68.2	75.6
Ended within crosswalk	51.6	75.6	23.2	48.7
Crossed independently	77.6	82.1	78.0	82.1

^†All numbers reported are percentages reflecting the average percentage of four crossings each individual performed in the stated manner.

The number of participants included in each analysis fluctuated some as a result of using only trials in which beaconing was actuated and participant task completion was independent. The total sample size (N) for the various analyses ranged between 27 and 30.

^***p<0.001

Only data that was collected during independent travel was used in analysis, so appropriate pre-and post-installation comparisons can be made. This results in a small to moderate amount of missing data for nearly all variables in the data set (e.g., where participants ended their crossings was not recorded if they required an intervention during the course of the crossing; alignment was not recorded if assistance was requested and provided in aligning to cross). To compute inferential statistics and evaluate changes in performance, percentages were calculated as follows. For each variable, started within the crosswalk, for example, a percentage was calculated for each participant based on how many of the crossings were begun from within the crosswalk. For the post-installation data, the average was calculated using only the trials in which the beaconing feature was activated by use of an extended button press. Because percentages were calculated across multiple trials by each participant, despite rates of use around 50%, most participants were still included in each analysis as they may have had two or three usable trials and one which was not usable due to not using the beaconing or an intervention. An average of all participants' percentages was then calculated for each city by condition. In Portland, the beaconing was activated on approximately 50% of the crossings. In the first round of post-installation testing in Charlotte, the use rate was 48%, while the rate of use was 98% during the second round of post-installation testing.

Study 1: Pre-Installation vs. Post-Installation-1 (simultaneous tones from both ends of the crosswalk)

Independent t-tests between new and returning participants were performed for each variable on post-installation data. These tests compared the average post-test percentage scores of participants who completed both pre- and post-test with those who completed only the post-test. Returning participants did not differ from new participants, thus there was no evidence of learning resulting from participation in pre-installation testing. Therefore, the data from the two groups were combined, and inferential statistics were computed with between-subjects analyses (N=32).

Wayfinding Measures

In Portland, there was significant improvement in the rate of starting within the crosswalk [t(28)=4.093, p<0.001], and a trend toward increased ending within the crosswalk [t(27)=1.824, p=0.079] Other wayfinding measures in Portland, and all wayfinding measures in Charlotte (Post-1), revealed no significant change in pedestrian behavior following APS installation. Thus, APS installation resulted in only slightly improved wayfinding, and no negative effect on performance.

Audible Beaconing – Round 1

On only half of the crossings, participants who used the beaconing-enabled pushbuttons chose to activate the beaconing. The simultaneous beaconing signal, even with the beaconing locator tone during the flashing don't walk, did not significantly assist participants in finding the correct destination, although there was an increase in mean percentages. In Portland, participants on average finished 76% of their beaconing-actuated crossings within the crosswalk, while the average was 70% of crossings where the pushbutton was used but beaconing was not activated. In Charlotte, 48% of crossings when beaconing was called ended within the crosswalk, as compared to 39% without the beaconing.

Study 2: The Effect of Additional Beaconing Features on Measures of Wayfinding in Charlotte

Additional beaconing features were implemented between the two post-installation tests in Charlotte in an attempt to improve wayfinding. Half of the participants in post installation testing had participated in one or both of the previous testing sessions, however, statistical analysis again revealed no evidence of practice effects or learning. Thus the groups were combined for all analyses (N = 32).

The APS used during Post-2 did result in significantly improved performance on four measures of wayfinding as compared to the pre-installation condition (see Table 2). Participants began and ended a greater percentage of crossings within the crosswalk [t(30)=2.819, p<0.01; t(27)=4.321, p<0.001; respectively]. There was also increased independence with respect to both finding the starting location and aligning to cross [t(30)=2.236, p<0.05; t(30)=4.698, p<0.001; respectively]. Despite the increased independence aligning to cross, rates of accurate alignment remained low (65%). Successfully ending within the crosswalk was thus often due to course corrections during crossing, after having started misaligned.

TABLE 2.

Charlotte - Comparison of Average Results Between Pre-Installation and Post-2, and Between Post-1 and Post-2^†

	Pre-Install	Post-2	Post-1	Post-2
Started within crosswalk	72.4	91.7**	78.8	91.7
Independently found starting position	82.8	96.9*	96.2	96.9
Started correctly aligned	51.0	64.6	48.5	64.6
Independently aligned	68.2	98.4***	75.6	98.4*
Ended within crosswalk	23.2	76.7***	48.7	76.7
Independently crossed	78.0	73.4	82.1	73.4

^†All numbers reported are percentages reflecting the average percentage of the four crossings across all participants that each individual performed in the stated manner.

The number of participants included in each analysis fluctuated some as a result of using only trials in which beaconing was actuated and participant task completion was independent. The total sample size (N) for the various analyses thus ranged between 27 and 32.

^*p < 0.05

^**p < 0.01

^***p < 0.001

Compared to Post-1 performance, the Post-2 feature changes installed in Charlotte led to one additional improvement in participant wayfinding performance (see Table 2). There was a significant increase in independence aligning to the crosswalk [t(27)=2.391, p<0.05], although accurate alignment remained poor. Note that the intersection where beaconing was installed in Charlotte was a skewed intersection with conflicting vehicular cues; stop lines and idling perpendicular traffic were not parallel to the crosswalk. Following the introduction of new beaconing features for Post-2, ending within the crosswalk increased considerably, from 49% to 77%, a trend toward significant improvement [t(26)=1.890, p=0.070]. While there was no significant change in independently finding the starting position or starting within the crosswalk, the independent task performance was already quite high during Post-1 and remained high during Post-2.

In Post-2, the beaconing feature was actuated on nearly every crossing (62 of 64), while it had been actuated on only half of the crossings in Post-1. This change in participant behavior may have been attributable to either, or both, the hands-on familiarization with device features in Post-2 versus just hearing a recording of device features in Post-1, or that participants simply perceived the additional features in Post-2 as helpful and were highly motivated to actuate them.

DISCUSSION AND CONCLUSIONS

Across the two cities, and at all locations where APS were installed, data showed numerous improvements in safety and independence, and no negative impacts of APS-installation. The addition of APS resulted in a nearly two-second overall reduction in starting delay across the two cities (Scott, et al. 2008)

Post-installation in Portland, APS resulted in significant improvement in crossings started within the crosswalk, which was probably related to the pushbutton locator tone. One crosswalk (#1) was offset from the corner and participants were more likely to find the pushbutton, and the correct starting location, post-installation. Researcher observations and participant comments indicated that when the audible beaconing was called, it was quite difficult to hear the walk indication at the starting location because the sounds only came from the overhead speakers on both ends of the crosswalk and were aimed at the center of the crossing. In addition, the audible beaconing, provided by the louder locator tone throughout the flashing don't walk, did not seem to improve “ending within the crosswalk”, as researchers had expected.

Results of the Portland testing were used to refine the technology and its operation for the first round of post-tests in Charlotte. In Charlotte Post-1, audible beaconing was modified to provide audible walk indications from both pushbutton speakers and overhead speakers, rather than from just overhead speakers as in Portland. In addition, the length of button press required to call the beaconing signal was shortened, because some individuals were observed to attempt to call the audible beaconing, but not hold the button for three full seconds. Research had meanwhile determined that a one-second button press is adequate to prevent unintended calling of the beaconing by the general public (Noyce & Bentzen, 2005). Despite these changes, results were similar to those in Portland, without any real improvement in “ending within the crosswalk”; less than 50% of crossings were completed within the crosswalk.

There was also no improvement in starting within the crosswalk in Charlotte. Pushbuttons were almost 10 feet from the edge of the street on 3 of the 4 crossings, and as participants moved to the curb line after pushing the button, it seemed they used stopped traffic and curb alignment to decide where to begin crossing, which put them outside the crosswalk area. While the pushbutton locator tone may have helped participants find the pushbutton in Post-1, it did not lead to improved `starting within the crosswalk', as was seen in Portland.

Generally, Post-1 showed that providing louder walk indications and pushbutton locator tones simultaneously from both ends of the crosswalk did not improve wayfinding. This was corroborated by concurrent research which evaluated a combination pedhead-mounted and pushbutton-integrated APS unit, and an APS with the option of increased volume of the locator tone from the pushbutton-integrated speaker, among others (Harkey, et al., 2007).

In Charlotte Post-2, beaconing features were further modified in response to previous results and participant suggestions. Audible beaconing was provided only from the far-end speaker, and an orientation tone was added. These modifications resulted in faster crossing initiations (Pre-installation, 8.2 second average starting delay; Post-1, 3.7 seconds; Post-2, 2.3 seconds), and a considerable increase in the percentage of crossings ending within the crosswalk, from 23% to 77%. While the orientation tone may have led to somewhat improved understanding of ending location, and produced increased alignment independence, rates of accurate alignment remained relatively poor (64%).

Pushbutton information messages (street names) from the pushbutton speaker were followed by 7 repetitions of the orientation tone from the far end of the crosswalk, during which participants could confirm the location of the opposite end of the crosswalk. Researchers observed that participants would sometimes align toward the orientation tone, then move up to the edge of the street. The improvement in starting within the crosswalk might be related to participants using the orientation tone as a clue to the correct direction to walk from the pushbutton toward the edge of the street.

Data on their alignment was recorded just before participants started to cross. Participants often realigned while waiting to cross, listening to perpendicular traffic, which seemed to result in poor alignment at the point when data was recorded. The effect of the beaconing locator tone from the far end of the crosswalk in Charlotte was obvious as participants initiated their crossings. Participants were often observed to correct misalignment upon hearing the loud locator tone from the other end of the crosswalk as they began their crossing. Subjective responses to the Post-2 audible beaconing were exciting. Several participants stated that they wanted the feature installed everywhere, because it would help them cross at complex or skewed intersections.

Because the beaconing was provided by the far-end pushbutton locator tone sound, rather than louder walk indications, the potential for hearing walk indications from the wrong crosswalk was minimized or eliminated. Audible beaconing has not previously been recommended at locations with channelized right turn lanes, split phasing, and other similar situations because of the researchers' concerns that there might sometimes be ambiguity regarding which crosswalk was being signaled if pedestrians who are visually impaired heard a loud walk signal, but it wasn't for the crossing they intended to make. More evaluation is needed at those types of locations, but researchers were encouraged by the lack of ambiguity of the beaconing in Post-2 testing.

These results are quite promising, although they are related to crossings at only one intersection, but an intersection with very difficult long skewed crossings, as can be seen in Figure 1, with confusing traffic cues for blind travelers. At this location, the walk signal was only 4 seconds long, so participants heard the far-end locator tone very soon after they began to cross. A longer walk indication would delay the onset of audible beaconing and might make it less successful, as pedestrians would have more time to have veered from the crosswalk area, and might not be able to hear the audible beaconing. Alternatively, pedestrians might wait at the curb until they heard the loud locator tone, thus delaying their crossing.

Implications for practice

The results of this research indicate that, while APS provide information about the status of the pedestrian signal, APS generally do not provide good wayfinding information, particularly when sound is presented simultaneously from both ends of crosswalk. O&M specialists and individuals who are blind need to understand that there is little or no improvement gained by installing loud signals, unless the signals provide effective audible beaconing. They also need to understand the issues involved in audible beaconing in order to effectively advocate for beaconing where needed.

It may be possible to provide effective directional information from far-end speakers, but more research is needed in various situations to confirm results found in this study. More evaluation is planned at smaller intersections, with different walk indication lengths, and where the intersection is more confined by nearby structures.