Model-Based Signal Subspace Channel Tracking for Correlated Underwater Acoustic Communication Channels

Abstract

Many underwater acoustic channels exhibit correlated (semideterministic) multipath arrivals. Such channels are often time-varying with extensive delay spread and yet have a limited number of degrees of freedom due to cross-path (cross-tap) correlation. Traditional least square algorithms used for channel estimation do not exploit this correlation structure. In this paper, a model-based channel tracking algorithm is proposed for correlated underwater acoustic communication channels. To exploit the cross-tap correlation, the channel impulse response (CIR) is projected into a lower dimensional signal subspace consisting of a set of uncorrelated channel components. Assuming constant bases, the channel variations are then represented via an autoregressive (AR) model of these uncorrelated components. This leads to a dynamic model of reduced dimension which can be effectively processed using a Kalman filter, yielding improved channel tracking performance. By tracking only a small number of degrees of freedom in the signal subspace, significant savings in computations can be achieved compared with the conventional recursive least square (RLS) algorithm. Performance is demonstrated with at-sea data.