Speaker: Sandipana Dowerah
Data and place: November 4, 2021, at 10:30 – Videoconference
Abstract: Speaker verification has recently been deployed in several real-world applications including secured access to bank services via telephone or internet. However, verification based solely on voice remains a modality with limited reliability under real conditions including several acoustic perturbations. Speech signals can be corrupted by environmental noise as well as room reverberation which severely affects the speaker verification performance. This talk is a presentation of the impact of multichannel speech enhancement pre-processing as a front-end to Speaker Verification in a noisy and reverberant environment. We analyze the impact in realistic scenarios involving additive noise and reverberation on the speaker verification performance. Main efforts have been focused on the front-end multichannel processing and the acoustic conditions during the enrollment phase.
