Abstract: Blind separation of sources consists of recovering a set of signals of which only-instantaneous linear mixing is observed. This paper presents a novel blind source separation method when noises are complex isotropic SaS process. SaS processes can describe many-signals and noises with impulsive nature, but its second order and high order statistics are infinite, so, a subspace approach is used to process the observed data, then joint approximate diagonalization of eigen-matrices is used to estimate the mixing matrix and source signals. Computer simulation shows the high performance of the proposed method.