Sparse channel estimation for OFDM transmission based on representative subspace fitting

In channel estimation for OFDM transmission over sparse channels, the time-domain approach requires estimation of much fewer parameters than the frequency-domain approach, but one needs to estimate both the delay and the gain of each path. We consider the time-domain approach and propose an algorithm that finds a set of delays under a certain criterion on the approximation error. More specifically, we obtain one additional path delay at a time, with each newly added term selected to reduce by the most amount the difference between a subspace spanned by the estimated delays and that perceived at the pilot subcarriers. The proposed method can handle non-sample-spaced path delays and it allows arbitrary, irregularly spaced pilot allocations. Simulation results show some of its operational characteristics and demonstrate its superiority over simple frequency-domain interpolation methods.