A Novel Method for Blind Segregation of Speech and Image Data Using Independent Vector Analysis.

Abstract

Independent Vector Analysis (IVA) is a handy algorithm that is used to separate a Convolutive mixture into its constituent signals that are most common in acoustic surrounding. While postulating that sources are independent and mixing is linear. IVA strongly excludes the permutation issue in the frequency domain by the adoption of spherical dependency model to the entire frequency bins. Which was one of the dilemma with Independent Component Analysis (ICA) even though it was handled by several engineering solutions. Till now IVA algorithm segregates the mixture into its components that are generated by homogenous sources, while in this paper heterogeneous sources are introduced. Quality of splitting is determined utilizing signal-to-noise ratio (SNR)