Analysis of Deep Neural Network for Speaker Identification Task with Video Input

Abstract

: Speech processing area is wide area in computer science and telecommunication field. Speech recognition, speaker recognition and speech synthesis are prevalent tasks in speech processing area. This research work focuses on speaker identification part of speaker recognition. Deep learning shown significant performance in speech re3cogntiion area. This fact has motivated to analyze deep neural network .for another tasks like speaker recognition. Deep autoencoder has been proved good for feature representation purpose, this work has experimented deep autoencoder for speaker classification activity of speaker identification. Significant accuracy has been achieved for speaker classification. Speaker identity obtained used for extracting speeches of interested speaker from mixed speaker videos.