mvcisback / SSLVC

Sound Source Localization using Visual Cues
4 stars 1 forks source link

CS598ps_project

In this paper we present a creative approach to reconstruct 3D audio for multiple sources from a single channel input by detecting and tracking visual cues using supervised learning methods. We also discuss a similar approach for improving speaker’s classification from a video stream by employing both facial and speech likelihoods, or simply Multimodal Speaker Recognition on a video stream.

Videos assets are here: