3D Auditory Displays for Driver Assistance Systems

Sammanfattning: A constantly increasing number of advanced driver assistance systems are implemented in cars.The driver is required to interpret the data reported by these systems while driving, and thereforeit is important to know how the information signals provided by such systems affect the driver.Driving relies to a large extent on vision, and most current advanced driver assistance systemsuse visual information, requiring visual attention from the driver. This conflicts with the drivingtask and leads to an increase in inattention and in cognitive and visual workload. Off-loadingvisually demanding tasks to other sensory modalities should be beneficial in terms of the driver’sability to divide attention resources. The aim of this thesis is to gain knowledge on how the driveris affected by information and interaction with driver assistance systems utilising 3D auditorydisplays. The goal is to provide knowledge of how to create robust interfaces that assist the driverwhile maintaining the necessary safety aspects while driving.The first paper investigated the importance of sound source placement within the car compartment.Since sound can be a very effective means of capturing attention, it could cause dangeroussituations if used in the wrong way. If sound is used to draw attention, one must be sure to notdraw the driver’s attention away from something more critical. The study showed that soundcan improve attention to the road when the sound source location is in front of the driver.Previous research has shown that a combination of sensory modalities increase the robustnessof perception. To avoid additional load on the visual modality, the second paper investigatedthe usefulness of vibrotactile information, in combination with auditory and/or visual signals.There was no significant increase in performance when adding vibration to a sound signal, butvibration and sound on their own led to similar performance. It can be argued, however, thatcombinations are still useful since they reduce the risk of an important signal being missed orneglected. In this investigation, all signals containing a visual component provided the fastestresponse time.The third paper investigated the usefulness of a 3D auditory display applied on a parking assistancesystem. A user study was conducted to evaluate drivers’ workload and performancewhen using different system implementations. It was found that a 3D auditory display was wellreceived as long as the sound only came from one direction at a time. Providing too much informationto the driver by using simultaneous sound sources at different locations was perceived asconfusing and annoying. In some cases, the effort required to perform a successful parking wassignificantly reduced when using a 3D auditory display system compared to a traditional parkingassistance system using only two sound sources. There were also indications that 3D soundcan reduce drivers’ mental workload and frustration with parking assistance systems.Placing sound sources in the car where they are most suitable can be difficult, especially if soundsources outside the car are required. The fourth paper investigated if it would be possible to use acrosstalk cancellation technique in combination with binaural recordings to reproduce 3D audiousing only two loudspeakers in the car compartment. It was found that when the loudspeakersare placed close to the ears of the driver, crosstalk cancellation can be effective inside the car compartment by avoiding strong reflections. Closely placed loudspeakers reduce filter creationcomplexity and improve sound quality.The fifth paper tackles the problems of within-the-head perception and front-back confusion,commonly found in binaural recordings and synthesis. It has been shown that reverberant binauralsounds are perceived more externalised than anechoic sounds, and tend to suffer less fromfront-back confusion. However, information sound signals should rarely be reverberant. Theidea was to create an image-source model with focus on early reflections and compare it withtraditional artificial head recordings. The image-source model performed better than the artificialhead recordings in terms of front-back confusion, and at the same time considerably limitedthe inherent room acoustics in the sound.In conclusion, meaningful and informative sound signals can raise driving performance if theyare placed in front of the driver. This encourages the driver to maintain focus on the road, byusing the attention capture capabilities of sound. A 3D auditory display can be beneficial forsystems such as parking assistance, since it helps lowering driver effort, mental workload andfrustration. Also, 3D auditory displays increase the drivers’ sense of safety and comfort, by augmentingthe drivers’ surrounding and thereby increasing their situational awareness. If a 3Dauditory display should be implemented in the car, a transaural system with crosstalk cancellationshowed promise for virtual sound source reproduction. However, binaural synthesis hasinherent problems that need to be addressed, specifically front-back confusion. It is hard to recommendusing crosstalk cancellation with binaural synthesis at the current state for locationcritical signals due to localisation confusion. However, using more than one modality for informationsignals makes the signals difficult to miss and could also help with the front-back confusionproblem, but further research is needed. If a transaural system with crosstalk cancellationis chosen as reproduction format, make sure to carefully select the loudspeaker placementsin order to avoid as many problems with filter generation as possible, since this will cause errorsin crosstalk suppression and affect the reproduction sound quality. Directive loudspeakersmounted in the neck rest, or ceiling, are good candidates for implementation.