Joint signal detection and channel estimation in multi-scale multi-lag underwater acoustic environments

We consider the problem of jointly detecting a known signal and estimating the channel in the frequency range used by underwater acoustic communication systems. A multi-scale multi-lag propagation environment is considered and no prior knowledge of the channel order is assumed. The proposed detection/estimation method is based on the framework of multifamily generalized likelihood ratio tests applied to a signal lying in a union of subspaces. The result is a ”tuning-free” orthogonal matching pursuit algorithm with a stopping criterion that does not require the knowledge of the number of channel taps or the noise variance. The performance is illustrated with replay simulations using real shallow-water channels measured in the Mediterranean Sea. Numerical results show that the proposed method outperforms competing algorithms in terms of both detection probability and channel estimation error. In addition, channel estimation does not exhibit a performance floor as observed with fixed-order-based approaches.