Zero-forcing blind equalization based on channel subspace estimates for multiuser systems

The recovery of input signals in a frequency selective fading channel is a problem of great theoretical and practical interest. We present several new blind algorithms that utilize second order statistics for multichannel equalization. The algorithms are based on the subspace extraction of a preselected block column of the channel convolution matrix. For multiuser system, user signal separation can be achieved based on partial information of the composite channel response. The equalization algorithms do not rely on the signal and noise subspace separation and therefore tend to be more robust to channel order estimation errors.