A hybrid MMSE approach to channel shortening for underwater acoustic OFDM

Underwater acoustic (UWA) channels pose a significant challenge to multicarrier communication systems. Doppler effects necessitate short block lengths, so that the channel can be approximated as time-invariant, while large delay spreads require long guard intervals to prevent inter-symbol interference. Channel shortening can be used to alleviate this double bind by compressing the received channel and reducing the effective delay spread. Two simple shortening filters are the Time Reversal (TR) and Minimum Mean Squared Error (MMSE) filters. The MMSE filter is a well-studied equalizer, but leads to very long filter lengths. The TR filter is very simple to compute and results in a short filter length, but does not achieve sufficient shortening in sparse UWA channels with a single receive phone. We propose a hybrid approach between these two algorithms, based on the fact that the MMSE filter converges to the TR filter for increasing noise level. By artificially increasing the assumed noise level, we can tune this Hybrid MMSE filter to provide a tradeoff between the performance of the MMSE filter and the simplicity of the TR filter. We perform a series of numerical simulations to show that, for a single-input single-output UWA multicarrier system, the Hybrid MMSE filter can be tuned to provide a shortening filter of fixed length that outperforms both the regular MMSE and TR shortening filters.