Training-based Adaptive Channel Tracking for Correlated Underwater Acoustic Channels

Multipath arrivals in many underwater acoustic channels are cross-correlated due to recent study. By exploiting the cross-correlation of mulipath arrivals, an efficient type of channel tracking for underwater acoustic channel is proposed by decomposing the cross-correlation into the channel principal components with low rank and the corresponding channel subspace. To track the channel, the channel principal components are modeled as an autoregressive (AR) process, and a Kalman filter tracks the channel components based on this AR model. The channel subspace is also tracked by recursive algorithms. However, these multi-procedure algorithms leave many undecided parameters which can affect the performance substantially. And the mismatch of the priori model is inevitable for underwater acoustic channels. In this paper, we present a training-based adaptive channel tracking algorithm. With the help of a short prior sequence of data, the parameters for the trackers are obtained through training. And an adaptive Kalman filter is used to correct the mismatch of the model which is also training-based. Performance of the proposed algorithms is demonstrated with real sea data. For the real sea data analyzed, the channel tracking accuracy is improved both in calm sea and rough sea.