Front-end processing for joint maximum likelihood channel and sequence estimation

The problem of performing joint maximum-likelihood estimation of a digital sequence and an unknown dispersive channel impulse response is considered starting from a continuous-time model. Previous investigations of this problem have not treated the front-end (FE) processing in detail. It is shown that a fractionally-spaced whitened matched filter, matched to the known data pulse, provides a set of sufficient statistics when a tapped-delay-line channel model is assumed. A signal model at the output of this FE processor, as well as representative sub-optimal FE processors, is described. Recursive computation of the approximate likelihood functional is presented and a general receiver structure, which may be interpreted as a generalization of the previously introduced technique of per-survivor processing, is suggested. Simulations conducted in the tracking mode illustrate that FE processing techniques suggested in the existing literature may significantly degrade performance.