ICA with Particle Filtering for Blind Channel Estimation in High Data-Rate MIMO Systems

A novel approach to joint blind channel estimation and data recovery is presented for high date rate multiple-input multiple-output (MIMO) systems operating over flat Rayleigh channels. The technique is based on independent component analysis (ICA) with particle filtering to track the time-varying channel. Given one value from the channel matrix coefficients\´ second-order statistics, non-stationary independent component analysis with a generalized exponential density function is used to separate each source signal. The performance is evaluated by simulation and is compared with optimal coherent detection as benchmark. Improved performance is demonstrated over the "conventional" blind approach with Kalman-based estimation, and over two known pilot-aided systems. Finally, the effect of time-selectivity of the channel on error performance is also assessed by simulation.