Maximum likelihood performance over higher-order statistics for blind source separation in wireless systems

Blind source separation (BSS) has recently become an area of prime interest. Conventional adaptive source separation systems use a training sequence to estimate and separate sources with the help of predefined optimization criteria. In BSS, the key idea is to use the data statistics to get apriori knowledge and thus separate the sources blindly. Two important approaches to this regime are the maximum likelihood (ML) estimation and higher-order statistical (HOS) estimation. This paper presents the BSS problem in separating sources for a dual antenna communication system using the aforementioned algorithms. It has been shown that ML estimation outperforms HOS estimation for a wireless medium with noisy data transmission.