Auditory localization of multiple stationary electric vehicles.

Abstract

Current regulations require electric vehicles to be equipped with acoustic vehicle alerting systems (AVAS), radiating artificial warning sounds at low driving speeds. The requirements for these sounds are based on human subject studies, primarily estimating detection time for single vehicles. This paper presents a listening experiment assessing the accuracy and time of localization using a concealed array of 24 loudspeakers. Static single- and multiple-vehicle scenarios were compared using combustion engine noise, a two-tone AVAS, a multi-tone AVAS, and a narrowband noise AVAS. The results of 52 participants show a significant effect of the sound type on localization accuracy and time for all evaluated scenarios (p<0.001). Post-hoc tests revealed that the two-tone AVAS is localized significantly worse than the other signals, especially when simultaneously presenting two or three vehicles with the same type of sound. The multi-tone and noise AVAS are generally on par but localized worse than combustion noise for multi-vehicle scenarios. For multiple vehicles, the percentage of failed localizations drastically increased for all three AVAS signals, with the two-tone AVAS performing worst. These results indicate that signals typically performing well in a single-vehicle detection task are not necessarily easy to localize, especially not in multi-vehicle scenarios.