A practical sparse channel estimation for current OFDM standards

Wireless channels especially for OFDM transmissions can be precisely approximated by a time varying filter with sparse taps (in the time domain). Sparsity of the channel is a criterion which can highly improve the channel estimation task in mobile applications. In sparse signal processing, many efficient algorithms have been developed for finding the sparsest solution to linear equations (basis pursuit, matching pursuit) in the presence of noise. In current OFDM standards, a number of the ending subcarriers at both positive and negative frequencies are left unoccupied (for ease of analog filtering at the receiver) which results in an ill-conditioned frequency to time transformation matrix. This means that the initial estimate for the impulse response of the channel (in time) easily varies as the noise vector changes. Thus in this case we cannot use most of the proposed algorithms in sparse signal processing. In this paper, we propose iteration with adaptive thresholding and MMSE methods to overcome this difficulty. Simulation results indicate that the proposed method is almost perfect for stationary channels and only minor performance degradation is observed with increase of Doppler frequency.