Reduced-rank adaptive MMSE equalization for high-speed CDMA forward link with sparse multipath channels

This paper deals with adaptive reduced-rank, chip-level MMSE estimation for the forward link in CDMA based on the multi-stage nested Wiener filter (MSNWF). We first examine through simulations the convergence of 'block-adaptive' MSNWF based on training by pilot symbols; MSNWF operates in a low-rank subspace facilitating faster convergence than LMS and RLS. We then describe a "semi-blind" approach which exploits the underlying channel structure, specifically the fact that the channel impulse response is sparse, to adapt a much lower-dimensional equalizer. Our simulations show this low-rank MMSE equalizer converges very quickly to the asymptotic SINR.