Listening Experiments in Virtual Acoustic Environments: Framework and Application to Electric Vehicle Noise

Doktorsavhandling, 2026

Road traffic noise is one of the most widespread environmental health risks in Europe, with chronic exposure linked to elevated blood pressure, cardiovascular disease, sleep disorders, and reduced quality of life. To better understand these effects, controlled experiments measuring how humans respond to specific sounds are necessary. Such experiments require realistic laboratory reproductions of traffic noise, which is far from straightforward. This thesis addresses that challenge by developing a methodological framework for conducting listening experiments in virtual acoustic environments, consisting of highly realistic acoustic simulations under controlled conditions. Electric vehicles served as the central case study, connecting the work to one of the most significant ongoing transitions in urban transport.

At high speeds, electric vehicles produce noise levels comparable to those of conventional cars. At low driving speeds, however, electric vehicles can be nearly silent. For people with visual impairments and other vulnerable road users, this is a serious safety concern, as they have long relied on engine sound to detect, locate, and judge the speed of approaching vehicles. Regulations worldwide therefore require electric vehicles to emit artificial warning sounds at low speeds, yet it is challenging to define what makes such a sound effective. A sound that is easy to detect may be difficult for people to localize, and a sound that grabs attention outdoors might annoy nearby residents. Current implementations vary widely, ranging from engine-like rumbles to futuristic, tonal signals.

This thesis provides experimental evidence comparing different warning signal types across several aspects of how people perceive and respond to sound. In one set of experiments, participants localized stationary electric vehicles in scenarios with up to three vehicles simultaneously present, mimicking situations in car parks or at urban intersections. In another, participants listened to low-level electric vehicle noise transmitted through a closed window into a furnished living room. Across both settings, warning signals consisting of only two tones performed poorly: they were the hardest to localize, particularly when multiple vehicles were present, and elicited the highest annoyance ratings and greatest physiological stress responses even at very low indoor sound levels.

These findings carry direct implications for how electric vehicle warning sounds should be regulated and designed. More broadly, the framework developed here is not limited to electric vehicles and can be applied to a wide range of environmental noise problems, from railway, drone, and aircraft noise to the emerging question of how autonomous robots should sound as they navigate tomorrow's city streets.