Self-adaptive maximum-likelihood sequence estimation
Bernd‐Peter Paris · 2002
The problem of estimating the most likely state sequence of a discrete-time finite-state Markov process with unknown parameters observed in independent noise arises in many important problems in digital communications, including self-adaptive equalization and adaptive multi-user detection. A maximum likelihood criterion over both the input sequence and the parameters is introduced for estimating the state sequence without using an embedded training sequence. Asymptotically, this estimator is close to the maximum-likelihood sequence estimator with completely known parameters. To facilitate the search for the most likely state sequence, we introduce computationally simple algorithms which are guaranteed to converge. Performance of the self-adaptive maximum-likelihood sequence estimator for the blind equalization problem is illustrated through numerical examples.>