Blind source separation is a popular technique which is used in the fields of signal processing,audio, video and image processing. BSS is used to separate the mixed signals with only knowing the mixed signals and knowing very little about original signal characteristics. The
separated signals should be very good approximations of the source signals. In particular, the blind source separation algorithm tries to estimate the Mixing Matrix. In my thesis, I have studied the blind source separation of signals based on its second order statistics.
The problem of blind source separation is studied considering the following cases: when the signal is modelled as non-stationary, cyclo-stationary and quasi-stationary. A closed form solution to the blind source separation of speech signals considering speech to be a
quasi-stationary source is studied and implemented.