Tracking Channel Variations in a Time-varying Doubly-Spread Underwater Acoustic Channel

Channel impulse responses (CIRs) are estimated for a doubly-spread acoustic channel in this paper by (1) estimating explicitly the path-dependent Doppler frequency shifts assuming a time-invariant CIR for individual blocks of data, and (2) estimating the time-varying CIR symbol by symbol including the phase change due to the Doppler. For the latter, a subspace based tracker is used to track the time variation of the channel basis vectors with their amplitudes tracked using a Kalman filter aided with a state-space model. For each method, the performance is measured by the normalized signal prediction error, the error between the data and the predicted signal based on the estimated CIRs, assuming the transmitted signal is known. Analysis of at sea data shows a 7 dB improvement in signal prediction error by tracking CIR using the model-based approach compared with the conventional methods based on explicit Doppler estimation.