Signal sound positioning alters driving performance.
Many of the information systems in cars require visual attention, and a way to reduce both visual and cognitive workload could be to use sound. An experiment was designed in order to determine how driving and secondary task performance is affected by the use of information sound signals and their spatial positions. The experiment was performed in a driving simulator utilizing Lane Change Task as a driving scenario in combination with the Surrogate Reference Task as a secondary task. Two different signal sounds with different spatial positions informed the driver when a lane change should be made and when a new secondary task was presented. Driving performance was significantly improved when both signal sounds were presented in front of the driver. No significant effects on secondary task performance were found. It is recommended that signal sounds are placed in front of the driver, when possible, if the goal is to draw attention forward.
CITATION: Lundkvist, A., Nykanen, A., and Johnsson, R., "Signal Sound Positioning Alters Driving Performance," SAE Int. J. Trans. Safety 4(1):2016, doi:10.4271/2015-01-9152.
The number of systems in cars that need the driver's attention increases from year to year. These can include entertainment, communication or systems designed to help the driver, such as navigation or advanced driver assistance. All of these rely on information and interaction with the driver, which could in turn increase driver workload. Previous research has shown that cognitive processing of a secondary task has a negative impact on driving performance [1, 2, 3, 4, 5, 6, 7, 8]. Hence there is a risk that the driver does not have full attention on the road and can thereby miss vital elements in the surroundings. Klauer et al.  found from an analysis of driving data collected in traffic that drivers engaging in complex tasks can triple the crash risk as compared to an attentive driver. They also found that visual inattention was the main factor behind 78 % of the crashes that occurred in the study. Blanco et al.  discovered from three on-road studies that in-vehicle navigation systems which present multiple options of possible routes had a negative effect on driving performance compared to systems only presenting one route. The conclusion was that multiple options increase the required cognitive processing from the driver and that it negatively affects driving performance by reducing the situational awareness. The effect was found both when presenting information visually and audibly.
Shams and Kim  stated that the visual modality generally dominates for spatial tasks, for instance localization and object recognition, while the auditory modality dominates for temporal tasks such as timing between two events. However, it has been shown that auditory events help in locating spatial objects . Shams and Kim  found that attention in one modality spreads to other modalities, and auditory events could therefore increase visual attention even when the sound had no direct meaning for the visual task. It has been shown in other studies that auditory cues are effective for reducing reaction times in simulated driving environments [3, 4, 7, 8]. Ho and Spence  examined this by comparing spatial and non-spatial signal sounds and how it could direct the drivers' visual attention for situations when a car was rapidly approaching in front of the subject's vehicle (or behind, which could be seen through the rear view mirror). They concluded that "spatially predictive semantically meaningful auditory warning signals may provide a particularly effective means of capturing attention".
Most of the after-market navigation systems available today emit sounds from the position where the device is installed, which could be in conflict with the message's semantics. Lee  performed a study on the effect of congruency between sound source location and the message's semantics for in-vehicle navigation systems. The recall accuracy of navigation information was increased, and response time decreased, when the sound source location was congruent with the spatial information presented by sound, for instance when a verbal message for a left turn was played at the left ear of the driver.
Previous research gives good reasons to believe that driving performance and driver attention may be affected by sound and its location. It seems promising that using sound might reduce the workload and allow the driver to focus more on the road. Using spatially placed sounds for the tasks might also help in drawing attention to where it is needed using the attention capture capabilities of sound. The response time needed to interpret the signals might also be reduced, and could thus help the driver to operate the vehicle more safely. This study intends to evaluate the effects that the spatial positioning of sound signals has on driving and secondary task performance for simulated driving using Lane Change Task (LCT) [11, 12] and Surrogate Reference Task (SuRT) [12, 13] as a secondary task.
Objective: How are 1) driving performance (measured using LCT) and 2) secondary task performance (measured using SuRT) affected by spatial placement of two sound signals; a sound informing that a lane change shall be made and a sound informing that a new secondary task shall be solved.
This experiment used Lane Change Task (LCT) [11, 12] which is a simple driving scenario and a secondary task based on the Surrogate Reference Task (SuRT) [12, 13]. A simple driving simulator was used for the driving scenario, consisting of LCT, a projector, a steering wheel and a secondary task screen (see Figure 1). The experiment was carried out in a sound recording studio at Lulea University of Technology. LCT and SuRT are laboratory tasks used to provide the subjects with a task similar to driving combined with a task similar to typical secondary tasks in a car, both allowing easy and controlled measurement of the subject's performance. The simplified setting limits the generalizability of the conclusions to be drawn, but as the objective of this study was to investigate how sound can be used to direct attention while driving the setup was considered suitable. Effects from changing positions of signal sounds were expected to be fairly small, and therefore an experimental setup giving sensitive and controlled measures was required. Even though this experiment does not prove that measureable effects will be reached under real driving conditions, it is likely that if positive effects can be shown using this laboratory setup there will also be positive effects of proper signal sound positioning for real car driving.
Lane Change Task (LCT)
LCT has been successfully used for repeatable and accurate retrieval of driving performance measures during simulated driving [12, 13, 14, 15], and is ISO standardized . LCT consists of a straight road with 3 lanes. The task in LCT is to change lane when information is displayed on signs by the side of the road (see Figure 2). A total of 18 lane changes occur on every track, consisting of 6 different lane change types (driving in the left lane; change one or two lanes to the right, driving in the right lane; change one or two lanes to the left, driving in the middle lane; change one lane to the left or to the right) that are repeated 3 times each. The order of the signs is randomized between tracks to avoid learning effects. The distances between signs varies between 140-188 m (mean distance 150 m), and the length of a track is approximately 3000 m .
The lane change information is displayed 40 m before passing the sign, and the subject should start the lane change immediately when the information has been recognized. Before the information becomes visible, the signs are by default blank. Since LCT is quite easy in its default state, the difficulty of the driving scenario was increased by removing the blank signs completely, so that the information popped up when the car was 40 m in front of the sign's hidden location. This makes it more difficult to predict when a lane change should be performed. The change of settings has been suggested to make LCT more realistic with regard to safety critical events . Furthermore, it gives a better measure of driver reaction time than the standardised measure. The standardised driving speed of 60 km/h was used in this study. Higher speeds would increase the driver workload and lower speeds decrease it. However, in pre-tests 60 km/h was shown to give a suitable workload. Choosing a suitable speed was found to be crucial for being able to measure effects from the positioning of the signal sounds on the tasks. When the workload was too low, the task was too simple and no effects were seen. When the workload was too high, the task was too difficult and the subjects were not able to manage it.
Measuring Driving Performance
[d.sub.s]=1/[summation over (y[member of][Y.sub.s])][square root of [([x.sub.y]-[r.sub.y]).sup.2]] 1
D=1/S[summation over (s)][d.sub.s] 2
where [d.sub.s] is the subject deviation in a section, [Y.sub.s], around a sign, [x.sub.y] is the subject lateral position at the longitudinal position y, [r.sub.y] is the lateral position of the reference path at the longitudinal position y, and [L.sub.s] is the length of the section. The total deviation is then given as the mean value of all sections, see Equation 2, where S is the total number of sections (number of lane changes). The analysis was only performed in proximity to the signs, starting 40 m bef ore the sign until 55 m after the sign (example shown as a shaded a rea in Figure 3) which constitutes a section [Y.sub.s]. All lane changes in the experiment were made within this range. By only doing the analy sis around the signs, random variations in how the subjects positioned themselves within a lane were reduced. This way, the driving performance measure became more affected by reaction time than lane keeping performance.
The secondary task software was based on the Surrogate Reference Task (SuRT) [12, 13]. The secondary task was to locate a circle with a slightly larger radius on a screen displaying a number of distractor circles. For the distribution of circles, the screen was sectioned into a matrix and 66 randomly selected sections out of 130 contained a circle. Each circle's position was randomized within a section. This distribution scheme was utilized to avoid drastic changes in difficulty level and uneven distributions. See Figure 4 for a screenshot of the secondary task. The subjects answered on which side of the centreline of the screen the larger circle was located by the use of 2 input buttons, positioned to the right of the steering wheel. The idle time between tasks was randomized between 2-5 seconds so that the appearance of a new task could not be predicted. A preliminary study was conducted to determine the time interval. It was subjectively determined over several trials that 2-5 seconds gave a good balance between driving and working on the secondary task. The various parameters for the secondary task can be found in Table 1. The secondary task performance was measured as response time from that a new task was displayed to user input. Error rate and number of answers were recorded.
Custom software was written for real-time analysis of output data from LCT to allow sounds to be played simultaneously when new sign information was displayed. The LCT software remained unchanged; however the engine sound was removed during the experiment to avoid masking of the examined signal sounds. The driving task sound was a composition of 3 repetitions of a click sound to give the impression of a turn signal. The second click was pitch shifted down 200 Hz. The time signal can be seen in Figure 5. The secondary task sound was a Microsoft Windows XP sound (ding.wav), commonly used in Windows XP to indicate that there is information on the screen. The secondary task triggered the sound when a new task was presented, of which the time signal can be observed in Figure 6.
The sounds were selected with spatial properties in mind so that they should be easy to locate. Sanders and McCormick  suggested that signal sounds where the dominant frequency components are located between 1500-3000 Hz should be avoided if good localization is required. This is mainly because the crossover region [18, 19] is located in this area, and thus causes high localization blur, making it difficult to hear exactly where the sounds are originating from . Blauert  defined the crossover region as a region where neither the interaural time difference nor interaural intensity difference contain enough information for localization of a sound source. Outside this area, the localization blur can go down to as little as 1[degrees]. The driving task sound had its energy peak at approximately 1600 Hz, which is just inside the crossover region (see Figure 7). Catchpole, McKeown and Withington  suggested the use of broadband noise for good localization. Since the driving task sound is mainly transient, the energy distribution can be compared to broadband noise, which has energy outside the crossover region as well.
It has been suggested by Edworthy, Loxley and Dennis  that harmonic overtones can be used to enhance spatial localization. However, they also said that fundamental tones higher than 1000 Hz should be avoided, since it can make the sound aversive. The secondary task sound has its fundamental frequency in the 800 Hz band and a secondary component at about 3150 Hz (see Figure 8).
Sanders and McCormick  recommended that the sound signals should have duration of at least 500 ms. If the duration is shorter, this could be compensated for by an increase in intensity. The duration of the secondary task sound was around 650 ms (see Figure 6). Since the duration of the driving task sound was shorter than 500 ms (see Figure 5), the equivalent sound pressure level was increased as suggested by Sanders and McCormick. The A-weighted equivalent sound pressure level for the duration of the sound signal used during the experiment was 72 dB(A) for the driving task sound and 70 dB(A) for the secondary task sound. The appropriate sound pressure levels were determined by subjective evaluation. The driving task and secondary task sounds were distinctly different from each other to avoid confusion in cases where the sounds were played at the same time.
The driving task was projected onto a screen about 4 m in front of the subject. To reduce the background noise level during the experiment, the projector was placed inside a sound insulated enclosure. The secondary task screen was located approximately 1 m away from the subject, 45[degrees] to the right. Response buttons for the secondary task were located to the right of the steering wheel (see Figure 9). Loudspeakers were placed at 2 positions:
1. In front of the driver at 1 m distance.
2. Under the secondary task display, 45[degrees] to the right, at 1 m distance.
The loudspeaker positions can be seen in Figure 9. To account for spatially congruent, and incongruent, visual and auditory information, 4 test cases with different combinations of loudspeaker positions were selected (see Table 2). Positions behind the driver were not considered, since this is of less importance in real car user interfaces. Positions to the left of the driver were not considered either, as secondary tasks are typically placed to the right of the driver in a left hand drive car. Switching the sound locations completely could possibly affect driving and secondary task performance negatively, but this was not considered as it would be of no use in a real life scenario.
In total, 30 subjects participated in the study, where 5 were female. The mean age was 22 years ([sigma] = 2.4 years). All subjects had a driver's licence at the time of participation, and the mean possession time was 3 years ([sigma] = 2.6 years). No subject reported any visual impairment that could not be compensated by eye glasses and all subjects had self-reported normal hearing. It has been shown that self-reporting is sufficiently sensitive for determination of hearing loss [23, 24]. The subject population only contained young drivers, and this limits the possibilities to draw conclusions on how older people are affected by placements of signal sounds.
The driving test consisted of driving while solving the secondary task. The subjects trained for the secondary task on its own, followed by driving a track without the secondary task. A third training round was performed where the subjects drove and carried out the secondary task simultaneously. A randomly selected test case including sound was used during the combined test round. Training is important since it has been shown that LCT can suffer from learning effects . Hence, it is important that the subjects are well familiarised with the simulator before starting the experiment. Otherwise, the subjects will be significantly worse at the beginning of the experiment than at the end, masking the effects of the differences between the cases to be compared. Each subject was tested for each case once (a total of 4 rounds). The order of the test cases (including baseline) was randomised for each subject.
A repeated-measures ANOVA was performed on the driving performance. The test showed that the different sound positions had an effect on driving performance, F(3, 87) = 3.270, p = .025, [[eta].sub.2] = .101. Mauchly's test indicated that the assumption of sphericity was not violated ([chi square](5) = 10.003, p > .05). As a post-hoc test, Bonferroni corrected pairwise comparisons were made (Table 3), showing that playing both signal sounds in front of the driver gave significantly better driving performance than the baseline without sound (p < .05).
A repeated-measures ANOVA showed no significant differences on secondary task response time, F(3, 87) = .771, p > .05, [[eta].sub.2] = .026. Mauchly's test indicated that the assumption of sphericity was not violated ([chi square](5) = 1.469, p > .05). The mean response times for the different cases can be found in Table 5. For the error rate, a repeated-measures ANOVA showed no differences between the sound locations, F(3, 87) = 1.880, p > .05, [[eta].sub.2] = .061. The error rate was maintained at approximately 1 %.
This study has shown that LCT can be used as a simple driving simulator to acquire measurable effects of sound positioning. However, in this study, the difficulty level of the simulator was adjusted from the default values in LCT by replacing the stationary signs with popup signs. The analysis of the driving data was only performed in close proximity of the signs (40 m before to 55 m after), to avoid large influence from lane-keeping performance and emphasize reaction time. Sound sources that are placed in well selected positions can have a positive effect on driving performance. By playing all information sounds during the task in front of the driver, attention was directed to the front which gave rise to a significant improvement in driving compared to not using any sound. However, no improvement or reduction in performance was found for the secondary task. A reason may be that the driving task is still too easy, so that the subjects could keep focus on the secondary task equally well between test cases. Furthermore, the secondary task was positioned so that a new event could be easily detected in the peripheral vision. Since the secondary task was not only to detect that a new task became visible, but also to solve the task, any peripheral reaction time benefits of adding sound may be masked by the time it took to solve the task.
In real cases, attention to the road is probably stronger. Still, in case of distraction in other areas or in case of too high workload, it might be beneficial to be able to direct a driver's attention.
The reason there was only an effect for the case with both sounds in front of the driver compared to no sound may be because of the bias towards driving. It may translate to a larger effect or quicker response in real driving if the driver is focused elsewhere than on the road. Furthermore, if the sound source should be located anywhere else than on the road, it might keep the driver focused on the wrong area in case of important driving related information.
This study has shown that playing all information sounds in front of the driver oriented attention to the driving task. For this case, the driving performance was improved. The reaction time and error rate for the secondary task were unaffected by the use of sound, and therefore the position of the sound did not matter. It was shown that sound has the capacity of drawing attention to where it comes from, even if the results were not so strong. In the experiment positive effects on attention to driving could be reached. No effects on the attention to the secondary task could be seen. Despite the vague results, it would be recommended to place signal sounds where attention is wanted. The effect of placement is small, but under some circumstances it may help. This may lead to less and/or shorter off-road glances and contribute to safer driving.
[1.] Klauer S. G., Dingus T. A., Neale V. L., Sudweeks J. D. et al., "The Impact of Driver Inattention on Near-Crash/Crash Risk: An Analysis Using the 100-Car Naturalistic Driving Study Data," Tech. Report, National Highway Traffic Safety Administration, Springfield, Virginia, April 2006.
[2.] Blanco M., Biever W. J., Gallagher J. P., and Dingus T. A., "The Impact of Secondary Task Cognitive Processing Demand on Driving Performance," Accident Analysis and Prevention 38:895-906, 2006.
[3.] Ho C. and Spence C., "Assessing the Effectiveness of Various Auditory Cues in Capturing a Driver's Visual Attention," Journal of Experimental Psychology 11(3):157-174, 2005.
[4.] Ho C., Tan H. Z., and Spence C., "The Differential Effect of Vibrotactile and Auditory Cues on Visual Spatial Attention," Ergonomics 49(7):724-738, 2006.
[5.] Lee Y.-L., "The Effect of Congruency Between Sound-Source Location and Verbal Message Semantics of In-Vehicle Navigation Systems," Safety Science 48:708-716, 2010.
[6.] Cao Y., Mahr A., Castonovo S., Theune M., et al., "Local Danger Warnings for Drivers: The Effect of Modality and Level of Assistance on Driver Reaction," in 14th ACM International Conference on Intelligent User Interfaces, 239-248, 2010.
[7.] Sodnik J., Tomazic S., Dicke C., and Billinghurst M., "Spatial Auditory Interface for an Embedded Communication Device in a Car," in International Conference on Advances in Computer-Human Interaction 1:69-76, 2008.
[8.] Sodnik J., Dicke C., Tomazic S., and Billinghurst M., "A User Study of Auditory Versus Visual Interfaces for Use While Driving," International Journal of Human-Computer Studies 66:318-332, 2008.
[9.] Shams L. and Kim R., "Crossmodal Influences on Visual Perception," Physics of Life Reviews 7(3):269-284, 2010.
[10.] Ngo M. and Spence C., "Crossmodal Facilitation of Masked Visual Target Identification," Attention, Perception & Psychophysics 72(7):1938-1947, 2010.
[11.] ISO 26022:2010, "Road Vehicles - Ergonomic Aspects of Transport Information and Control Systems - Simulated Lane Change Test to Assess In-vehicle Secondary Task Demand," 2010.
[12.] Mattes S., "The Lane-Change-Task as a Tool for Driver Distraction Evaluation," Quality of Work and Products in Enterprises of the Future, 57-60, 2003.
[13.] Petzoldt T., Bar N., Ihle C., and Krems J. F., "Learning Effects in the Lane Change Task (LCT) - Evidence From Two Experimental Studies," Transportation Research 14:1-12, 2011.
[14.] Bruyas M.-P., Brusque C., Tattegrain H., Auriault A., et al., "Consistency and Sensitivity of Lane Change Test According to Driving Simulator Characteristics," Intelligent Transport Systems 2(4):306-314, 2008.
[15.] Lundberg K. and Stralin A., "Forardistraktion - En studie av metoden Lane Change Test," Master's thesis, Lulea University of Technology, 2005. Swedish.
[16.] Petzoldt T., Graichen L., and Krems J. F., "Modifying the Lane Change Task - How Does Increased Unpredictability of Lane Changes Affect Performance?," in 3rd International Conference on Driver Distraction and Inattention, Gothenburg, Sweden, 2013.
[17.] Sanders M. S. and McCormick E. J., "Human Factors in Engineering and Design," McGraw-Hill Science/Engineering/Math, 7th ed., 1993.
[18.] Blauert J., "The Psychophysics of Human Hearing," Cambridge: MIT Press, rev. ed., 1996.
[19.] Feddersen W. E., Sandel T. T., Teas D. C., and Jeffress L. A., "Localization of High-Frequency Tones," Journal of the Acoustical Society of America 29:988-997, 1957.
[20.] Moore B. C. J., "An Introduction to the Psychology of Hearing," London: Academic Press, 5th ed., 2003.
[21.] Catchpole K. R., McKeown J. D., and Withington D. J., "Localizable Auditory Warning Pulses," Ergonomics 47:748-771, 2004.
[22.] Edworthy J., Loxley S., and Dennis I., "Improving Auditory Warning Design: Relationship Between Warning Sound Parameters and Perceived Urgency," Human Factors 33:205-231, 1991.
[23.] Sindhusake D., Mitchell P., Smith W., Golding M., et al., "Validation of self-reported hearing loss. The Blue Mountains Hearing Study", International Journal of Epidemiology 30:1371-1378, 2001.
[24.] Gomez M. I., Hwang S., Sobotova L., Stark A. D., et al., "A Comparison of Self-Reported Hearing Loss and Audiometry in a Cohort of New York Farmers", Journal of Speech, Language, and Hearing Research 44:1201-1208, 2001.
This project was made in co-operation between Lulea University of Technology and Volvo Car Corporation, and was financed by the FFI - Vehicle and Traffic Safety programme.
Andre Lundkvist and Arne Nykanen
Lulea University of Technology
Table 1. Parameters used for the secondary task. Parameter Value Number of Distractors 65 Distractor Radius 14 px Target Radius 20 px Circles' Thickness 2 px New Task Delay Time 2-5 s Screen Resolution 1280 x 1024 px Screen Size 19 in Table 2. Experiment test cases containing both sound source locations congruent and incongruent with the visual tasks. Case Driving Task Secondary Task Baseline No sound No sound 1 45[degrees] 45[degrees] 2 In front 45[degrees] 3 In front In front Table 3. Pairwise comparisons of driving performance. Units for mean difference, standard error and confidence intervals are meter. Perf. Perf. Mean Std. Lower Upper (I) (J) Diff. (I-J) Error Sig. (1) Bound Bound Basel. 1 .030 .034 1.000 -.067 .127 2 .052 .028 .417 -.026 .131 3 .085 (*) .027 .024 .008 .162 1 Basel. -.030 .034 1.000 -.127 .067 2 .022 .026 1.000 -.053 .097 3 .055 .031 .507 -.032 .142 2 Basel. -.052 .028 .417 -.131 .026 1 -.022 .026 1.000 -.097 .053 3 .033 .020 .636 -.023 .088 3 Basel. -.085 .027 .024 -.162 -.008 1 -.055 .031 .507 -.142 .032 2 -.033 .020 .636 -.088 .023 Perf. Effect (I) Size ([r.sup.2]) Basel. .160 .330 .501 1 .160 .155 .315 2 .330 .155 .296 3 .501 .315 .296 (1) Adjustment for multiple comparisons: Bonferroni. (*) The mean difference is significant at the .05 level. Table 5. Mean secondary task response time for the different test cases. Case Mean (ms) STD (ms) Baseline 1090 184 1 1087 190 2 1076 192 3 1062 174
|Printer friendly Cite/link Email Feedback|
|Author:||Lundkvist, Andre; Nykanen, Arne; Johnsson, Roger|
|Publication:||SAE International Journal of Transportation Safety|
|Date:||Apr 1, 2016|
|Previous Article:||Assessment of compressive thoracolumbar injury potential and influence of seat cushions on vertical impact loading of a seated occupant.|
|Next Article:||Response times for visual, auditory and vibrotactile directional cues in driver assistance systems.|