Low complexity 2D projection-based channel estimators for MC-CDMA

We compare here low complexity, two dimensional channel estimators for MC-CDMA, which use projection to perform filtering and/or interpolation. The principle is to project a first channel estimate obtained from pilot symbols onto a subspace where lies most of the channel energy, thus performing rank reduction. For advanced schemes we study here, the projection subspace can be robustly derived from singular value decomposition (SVD). Simple schemes, namely sample and hold and linear interpolation can be viewed also as projection schemes, with easy projection subspaces. This generalization makes SVD based algorithms easier to understand. The advanced schemes have a very low complexity, their design is robust, and extensive simulations show that their performances are good: they are 0.5 dB close of perfect channel estimation for QPSK, and at most 1 dB close for 16QAM.