Résumé: To cross uncontrolled roadways, where no traffic-halting signal devices are present, pedestrians with visual impairments must rely on their other senses to detect oncoming vehicles and estimate the correct crossing interval in order to avoid potentially fatal collisions. To overcome the limitations of human auditory performance, which can be particularly impacted by weather or background noise, we develop an assisting tool called Acoussist, which relies on acoustic ranging to provide an additional layer of protection for pedestrian safety. The vision impaired can use the tool to double-confirm surrounding traffic conditions before they proceed through a non-signaled crosswalk. The Acoussist tool is composed of vehicle-mounted external speakers that emit acoustic chirps at a frequency range imperceptible by human ears, but detectable by smartphones operating the Acoussist app. This app would then communicate to the user when it is safe to cross the roadway. Several challenges exist when applying the acoustic ranging to traffic detection, including measuring multiple vehicles' instant velocities and directions with the presence many of them who emit homogeneous signals simultaneously. We address these challenges by leveraging insights from formal analysis on received signals' time-frequency (t-f) profiles. We implement a proof-of-concept of Acoussist using commercial off-the-shelf (COTS) portable speakers and smartphones. Extensive in-field experiments have been conducted to validate the effectiveness of Acoussist in improving mobility for people with visual impairments.

Langue: Anglais