Auditory Displays for In - Vehicle Technologies

sounds, such as tones and earcons, have symbolic relationships with their referent process, and symbolic relationships generally must be learned in the context of the interface. Abstract relationships may allow for more flexibility in the representation of different processes; any interface element can be represented via a symbolic association with no inherent relationship between the sound and its referent. The problem, however, is that abstract sounds are more difficult to learn and remember than sounds with some relationship to their referent (Bonebright & Nees, 2007; McKeown & Isherwood, 2007; Palladino & Walker, 2007; Perry et al., 2007; Smith et al., 2004), and some research has indicated that direct relationships are most easily learned (Keller & Stevens, 2004; Stephan, Smith, Martin, Parker, & McAnally, 2006). The caveat to this rule, however, is speech. With the exception of onomatopoeia, language has no ecological association with its referents. Yet the symbolic associations of language are so overlearned that speech and speechlike sounds (e.g., spearcons and spindex) tend to be easily learned and effectively used in interfaces (Bonebright & Nees, 2009; Jeon & Walker, in press; Palladino & Walker, 2007; Smith et al., 2004). AUDITory DIsplAys For IVTs: exTANT sysTeMs AND proToTypes AND poTeNTIAl exTeNsIoNs Audio has already seen widespread implementations in many in-vehicle systems, although extant applications of sound in IVTs have yet to leverage the richness of sound for information display. The sounds of most systems default to minimally informative tones, chimes, and beeps. A number of other design possibilities have been successfully prototyped for IVTs or tested in research scenarios that may readily generalize to invehicle auditory information display. We group our discussion of auditory displays for IVTs into four categories: (a) collision avoidance and hazard warning systems, (b) auditory displays that alert the operator to his or her state or the conditions of the vehicle itself (e.g., low fuel), (c) auditory displays for interacting with IVIS, and (d) auditory displays that aid navigation. collision Avoidance and Driving hazard Warning systems Warnings typically indicate a negative system state that requires immediate attention or action. For IVTs, a warning will likely indicate an unsafe driving scenario, such as a potential accident hazard, and auditory warnings may capture and orient attention more effectively than visual warnings (for a review, see Spence & Driver, 1997). Auditory warnings research has amassed a considerable literature that includes numerous sets of design guidelines (e.g., Ahlstrom, 2003; Campbell, Richard, Brown, & McCallum, 2007; Edworthy & Hellier, 2006b). Edworthy and Hellier (2006a) identified four characteristics of the ideal alarm as (a) easily localizable in space, (b) not susceptible to masking by other sounds, (c) not a source of interference with other communication (e.g., speech), and (d) easily learned and remembered. at HFES-Human Factors and Ergonomics Society on December 8, 2011 rev.sagepub.com Downloaded from Auditory Displays 71 In addition to meeting minimal acoustic parameters for localizability and detectability, auditory warnings need to be informative enough to provide the system operator with some indicator not only of the adverse event but ideally also of its nature and perhaps even of corrective actions to be taken. Although auditory warnings that match the operator’s current informational needs are preferable to less informative, generic warnings (Seagull, Xiao, Mackenzie, & Wickens, 2000), a warning needs to be brief enough to relay a message as efficiently as possible to allow time for the operator to act. As such, both nonspeech sounds, such as tones, earcons, and auditory icons, as well as brief speech messages have been used as warnings in vehicles. Researchers (Edworthy & Hellier, 2006b; Patterson, 1990) have arrived at the consensus that auditory warnings should be presented at least 15 dB higher than background noise, and Edworthy and Hellier (2006b) advise that warning signals should not exceed the background noise intensity by more than 25 dB. This offers some guidance for auditory warning design, as sound level meters capable of establishing a rough baseline level of noise in a given environment are relatively inexpensive and widely available. The interested reader is referred to Mulligan, McBride, and Goodman (1985) for detailed flow charts to aid in the design of nonspeech signals that are detectable in noise and Giguere, Laroche, Osman, and Zheng (2008) for a detailed methodology for optimizing the presentation of warnings in noisy environments. In some systems, when an adverse event is particularly important or takes precedence to other concurrent system information, it may be important to design urgency into the auditory signal. Haas and Edworthy (1996) showed that the perceived urgency of an auditory signal increases as the loudness, pitch, and speed (i.e., rate) of the sound increased, with increases in loudness and pitch also resulting in faster responses. Other researchers (Guillaume, Pellieux, Chastres, & Drake, 2003), however, have shown that although the perceived urgency of an auditory warning is generally predictable from acoustic properties, the notable exceptions to this rule were instead explained by learned associations and cognitive representations of sound meaning. In Guillaume et al.’s (2003) study, for example, a stimulus that sounded like a bicycle bell should have resulted in a high urgency rating on the basis of predictions from acoustic models, but the sound was rated as having low perceived urgency. The researchers speculated that the cognitive representation of meaning for a bicycle bell overrode the urgency conveyed by the acoustic signal. Other research (Burt, Bartolome, Burdette, & Comstock, 1995) has indicated that the perceived urgency of a signal may change as the demands on the system operator change. A general rule regarding auditory warnings seems to be that increasing certain acoustic properties, such as frequency, intensity, and the temporal rate of a signal, will usually increase perceived urgency, and people subjectively feel that urgent auditory warnings are appropriate for driving situations that are associated with high urgency (Marshall, Lee, & Austria, 2007). Interestingly, however, the perceived urgency of a signal does not necessarily correlate with a faster objective response to the signal. A study (Sanderson, Wee, & Lacherez, 2006) tested auditory warnings that were designed according to a published international standard and found that participants perceived the standard’s “high-priority” alarms to indicate more urgency, yet they tended to respond faster to “medium-priority” sounds from the standard. at HFES-Human Factors and Ergonomics Society on December 8, 2011 rev.sagepub.com Downloaded from 72 Reviews of Human Factors and Ergonomics, Volume 7 collision Avoidance system (cAs) A CAS is an IVT that uses visual, auditory, or tactile warnings to inform the driver of potential impending accidents during which the vehicle is at threat of leaving the roadway or contacting other vehicles or objects. Varieties of CAS include adaptive cruise control, blind spot warnings, reverse warnings, lane departure warning systems, and rear-end collision avoidance systems. Researchers and designers have frequently used tones or noise bursts in studies of CAS, but other types of sounds may be more effective. Adaptive cruise control systems, for example, have begun to appear as a safety feature in many vehicles. Typically, the systems detect distances from a lead vehicle and automatically adjust the vehicle’s speed to maintain safe distances from lead vehicles when cruise control is engaged. Sound warnings in such systems indicate potentially dangerous states (e.g., collision threats) in which the driver needs to intervene with the system, and sounds may also indicate the engagement or disengagement of the system. Currently, the types of sounds used in adaptive cruise control systems seem to be simple, abstract beeps, chimes, and tones. Lin and colleagues (2009) recently tested the effectiveness of various types of simple auditory alerts for lane departures. Tone bursts and continuous tones of 500 Hz, 1750 Hz, and 3000 Hz were examined. Response times were significantly faster at 1750 Hz and 3000 Hz as compared with 500 Hz, and there were no differences for bursts versus continuous tones. Participants overwhelmingly felt that the bursts made better warnings, however, and most participants felt the 1750 Hz tone was the best choice for frequency. Lee, McGehee, Brown, and Reyes (2002) reported that a multimodal warning to indicate that a lead car was breaking significantly decreased collisions by 81% when the warning occurred close in time to the onset of braking. Later warnings were less effective but still showed improvement compared with no warnings, and the warnings improved performance for both distracted and undistracted drivers. The visual component of the warning was an icon above the instrument panel, and the auditory component consisted of pulsing bursts centered around 2500 Hz and presented 2 dB to 5 dB above ambient noise conditions, depending on the speed of the vehicle. Findings from another study (Suzuki & Jansson, 2003) suggested that both monaural and spatially predictive stereo auditory warnings (beeps) were less effective (i.e., resulted in slower reaction times) for correcting lane departures than haptic feedback delivered via the steering wheel when participants were naive to the meaning of the warnings. When participants were instructed on the meaning of the warnings, performance was equivalent for both auditory and haptic warnings. Interestingly, the researchers found that stereo warnings to predict the side of the lane departure did not facilitate corrective steering away from the departure. Instead, the participants looked ahead at the road before taking corrective actions with steering, despite the fact that the location of the tone indicated the direction of the departure. The finding by Suzuki and Jannson (2003) that training on the meaning of an auditory warning improves performance suggests that abstract tones and beeps may not make the most intuitive signals for collision warnings in IVTs. To this effect, a number of studies have shown that auditory icons likely offer a better option for warning sounds in vehicles. A study (Belz, Robinson, & Casali, 1999), showed that auditory icons (the sound at HFES-Human Factors and Ergonomics Society on December 8, 2011 rev.sagepub.com Downloaded from Auditory Displays 73 of tires skidding for an imminent front or rear collision and the sound of a horn honking for an imminent side collision) produced significantly faster breaking response times than did tones, a visual warning condition, and a control condition without warnings. Furthermore, the auditory icon display alone resulted in performance that was as fast as several multimodal conditions. In addition, auditory icons showed a considerable advantage compared with tones for identifiability of the meaning of the auditory signal, and participants generally preferred a multimodal display. A similar experiment (Graham, 1999) found that the same auditory icons (a car horn and skidding tires) resulted in faster braking response times than did a tone or a speech warning (“ahead”), but the auditory icons also elicited more false-positive braking in inappropriate situations. The car horn in the study was perceived by users to be an appropriate indicator for a potential collision, and the tone was perceived to be the least appropriate and least liked warning. Auditory icons represent a simple change away from the typical tone warnings of traditional interfaces, yet this small change might result in considerable safety advantages. Furthermore, in a recent study (McKeown, Isherwood, & Conway, 2010), a stronger ecological relationship between an auditory icon warning and an impending potential rear-end collision event improved reaction times, which led the authors to suggest that auditory icons act as “occasion setters” that prime reactions. spatial Audio for collision Warnings For many potential sources of collisions, knowledge of the location of the threat may offer relevant information and potentially suggest corrective action to the driver. Sounds emanating from a spatial location can capture visual attention and actually facilitate visual perception (McDonald, Teder-Salejarvi, & Hillyard, 2000); the implications of this crossmodal facilitation may be very important for collision avoidance in vehicles. In one study, nonspeech sounds that spatially cued drivers to the location of a simulated event requiring intervention (braking or accelerating) produced large improvements in response times (Ho & Spence, 2005). Ho and Spence (2009) further showed that people oriented faster (i.e., had a faster head movement response time) to auditory warnings presented close behind them (40 cm behind their heads) as compared with a waistline vibrotactile warning, a peripheral visual warning, or a condition in which the auditory warnings were farther away (80 cm) and in front of them—a location that roughly corresponds to the location of in-dash radio loudspeakers in many vehicles. In another study, however, no difference was found in driver response times between a condition that used a generic master alarm (abstract tone patterns) to warn of a potential danger as compared with a series of multiple distinct abstract sounds that indicated more specific information about the location and type of danger (Cummings et al., 2007), so clearly, this is an area where further research would elucidate more conclusive evidence and design heuristics. Spatialized audio presentation, however, offers another feasible approach to alerting the driver to possible collision hazards (or the cockpit pilot to possible targets, etc.). For sounds that are maximally localizable, Edworthy and Hellier (2000) recommended the use of sounds with multiple harmonics and low fundamental frequencies (also see Wightman & Kistler, 1983). at HFES-Human Factors and Ergonomics Society on December 8, 2011 rev.sagepub.com Downloaded from 74 Reviews of Human Factors and Ergonomics, Volume 7 AUDITory AlerTs For VehIcle AND DrIVer coNDITIoNs In much of the literature, auditory alerts are synonymous with warnings. We define alerts here as brief messages that assume a lower priority than warnings. These signals may not necessarily indicate a negative system state and also may not require immediate action. Auditory icons, earcons, musical sounds, pure tones, speech messages, and spearcons are all candidate types of sounds for alerts and reminders. Alerts and reminders may convey information to the system operator about ongoing processes, prospective actions to be taken at a later point in time, or optional courses of action.