Adaptive Channel Estimation for Underwater Acoustic MIMO OFDM Systems

Long acoustic multipath limits the applicability of MIMO OFDM channel estimation methods that require matrix inversion of size M_TL, where M_T is the number of transmit elements and L is the multipath spread measured in 1/B, the bandwidth inverse. To overcome this problem, sparse nature of the channel is exploited in an algorithm based on a compact signal representation that leads to two forms of adaptive implementation: one that requires matrix inversion but of a reduced size, and another that completely eliminates it. Channel estimation is coupled with phase tracking and prediction to enable decision-directed operation in the presence of non-uniform Doppler distortion, which in turn provides improved performance and reduced overhead. Performance is demonstrated on real data, recorded during a recent experiment in the coastal North Atlantic.