Initial Design of Acoustic Vehicle Detector with Wind Noise Suppressor

Vehicle detection is one of the fundamental tasks in the ITS (intelligent transport system). We are developing an acoustic vehicle detector that relies on two microphones at a sidewalk [1]. The vehicle detector successfully detected vehicles as well as their traveling directions. However, the detector has difficulties in vehicle detection in a high wind condition due to wind noise. This paper presents a wind noise suppressor for the acoustic vehicle detector. Our simple idea is to remove frequency components corresponding to wind noise. Our acoustic vehicle detector relies on TDOA (time difference of arrival) of sound signals on two microphones to detect vehicles, which can be derived from a part of frequency components of vehicle sound signals. We experimentally analyze frequency components of wind noise and design a filter to reduce wind noise. Initial experimental evaluations revel that our vehicle detector with a wind noise suppressor successfully detected vehicles with an F-measure of 0.77 in a normal wind condition.