EVALUATION OF BSS TECHNIQUES FOR MULTIPLE SPEAKER RECOGNITION

Abstract

The need for source (speech) enhancement is very important, because of the environment we are living in, which is composed of noise and other atmospheric disturbances, and this makes it almost impossible to record a source (speech) signal in pure form. In most of the mixed signals there is usually no information about each source such as its location and time distribution. In such situation the estimates of the original source signals is done based on the information of the received mixed signals, therefore the approach to be adopted in such cases to separate the signals must be one that does it blindly, thus the method Blind Source Separation is used in this work. In this dissertation work two BSS (Blind Source Separation) techniques namely Independent Component Analysis (ICA) and Time-Frequency Masking (T-F masking) have been explained and their performance is evaluated and compared with the parameter of SNR of separated signals. ICA is a method for finding underlying factors or components from multivariate (multidimensional) statistical data. It looks for the components which are both statistically independent and non-Gaussian. Time-frequency masking is another BSS technique which is able to separate source signals from single mixture and depends on the concept of W-disjoint orthogonality. In this work mixtures of simultaneous speeches are taken as source mixture. Though any independent and W-disjoint orthogonal source can be taken for experimental purpose.