Abstract
Source separation has numerous applications, namely, biomedical signal processing, exploration seismology, speech and image enhancement, localization, etc., which includes the estimation of the mixing model. The technique of source separation aims to estimate both the original sources and the mixing model using only the observations. We have analysed the source separation problem with the assumption that sources are statistically independent. On the basis of prior statistical information of the source signals and joint probability distributions function (PDF) between the observed signals, the information of the mixing model has been extracted and examined. It has been observed that the shape of joint PDF is highly useful in characterizing the mixing coefficients.