Abstract

This work discusses and implements the application of speaker recognition forthe detection of collaborations in YouTube videos. CATANA, an existingframework for detection and analysis of YouTube collaborations, is utilizingface recognition for the detection of collaborators, which naturally performspoor on video-content without appearing faces. This work proposes an extensionof CATANA using active speaker detection and speaker recognition to improve thedetection accuracy.

 

Quick Read (beta)