A prediction method using instantaneous mixing plus auto regressive approach in frequency domain for separating speech signals by short time fourier transform

C Anna Palagan; K Parimala Geetha

Abstract

A prediction method using instantaneous mixing plus auto regressive approach in frequency domain for separating speech signals by short time fourier transform

The revealed works of separation of speech signals, the most disadvantages is that the incidences of distortion speech at intervals the signal that affects separated signal with loud musical noise. The thought for speech separation in normal Blind Source Separation (BSS) ways in which is solely one sound supply in an exceedingly single area. The projected methodology uses as a network that has the parameters of the Instantaneous Mixing Auto Regressive model (IMAR) for the separation matrices over the entire frequency vary. A trial has been created to estimate the simplest values of the Instantaneous Mixing Auto Regressive model (IMAR) model parameters using two matrices W and G by suggests that of the maximum-likelihood estimation methodology. Supported the values of those parameters, the supply spectral half vectors square measure calculable. The whole set of Texas Instruments Massachusetts Institute of Technology (TIMIT) corpus is utilized for speech materials in evolution results. The Signal to Interference quantitative relation (SIR) improves by a median of 5dB unit of measurement over a frequency domain BSS approach.

Author(s): C Anna Palagan, K Parimala Geetha
Abstract | Full-Text | PDF

Share this