Blind Diversity Reception and Interference Cancellation using ICA

In this paper, we consider blind diversity reception and interference rejection in multi-antenna communications context, in terms of maximizing the output signal-to-interference-and-noise ratio (SINR). More specifically, we demonstrate that independent component analysis (ICA), although originally designed for noise-free linear models, is able to provide essentially the best possible output SINR among all linear transformations of received data in noisy linear models. In particular, our experiments indicate that one of the most widely applied ICA algorithms, equivariant adaptive source identification (EASI) algorithm, is, in practice, identical with SINR maximizing generalized eigenfilter in terms of SENR, even though it does not use explicit knowledge of the channel states and noise statistics. We also show that, in a special case of interference-free (that is, noise only) system, the EASI algorithm attains the greatest diversity gain blindly, i.e., performs as a blind maximal ratio combiner (MRC).