Abstract
Advances in telecommunications have lead to modern technologies like conference calls, video calls, audio calls. This
widely communicate over hands- free devices, where this type of communication is affected by reverberations, mixing of sources,
acoustic echo etc. Goal of this paper is to estimate noise free original sources from mixture of multiple sources, and hence phenomenon
known as Blind Source Separation (BSS).This implementation of source separation and estimation is performed by Independent
component analysis (ICA). Aim of ICA is representing data linearly such that components must be statistically independent. De-noising
was performed using discreet wavelet transform (DWT) by decomposing and thresholding. Effective results were obtained when ICADWT
combinely worked than only ICA. Peak signal to noise ratio for ICA-DWT was more than only ICA.