Per-cluster-prediction based sparse channel estimation for multicarrier underwater acoustic communications

In this paper, we present a sparse channel estimator for zero-padded orthogonal frequency division multiplexing (ZP-OFDM) modulation in clustered sparse underwater acoustic (UWA) channels. The proposed channel estimator consists of two stages. In the first stage, channel prediction from the previous block to the current block is carried out to generate artificial measurements on OFDM subcarriers. One key innovation is to allow different clusters of channel paths to vary differently across the blocks, where the delay, amplitude and phase variations of the clusters are determined based on the channel estimate of the previous block and the measurements on pilot subcarriers of the current block. In the second stage, a sparse channel estimation is performed based on both the artificially generated measurements and the real measurements of the current block, accounting for their different reliabilities. Experimental results demonstrate that the proposed channel estimator considerably outperforms the conventional channel estimator without cluster-based adaptation.