Training Symbol Design for 2-D Least-Squares Channel Estimation

Underwater acoustic channel is always a 2-D dispersive channel. For 2-D least-squares channel estimation, the design of training symbols is an open problem. In this paper, (i) we derive the optimal relation between the dimensions of the training matrix and the ffective memory lengths of the equivalent discrete-time channel model, and (ii) we propose a simple algorithm for searching suitable training symbols.