Maximum Likelihood Sequence Estimation for Unknown, Dispersive, and Time Variant Communication Channels

摘要

The Maximum Likelihood Sequence Estimation (MLSE) algorithm is one of a number of techniques developed for processing signals with intersymbol interference. The MLSE algorithm is optimum in the sense that it minimizes the probability of error. This report describes the intuitive and mathematical basis for the MLSE algorithm. The derivation assumes the channel is known and time invariant. In Navy communications (underwater acoustic and hf radio channel) channel characteristics are unknown to the receiver. This study's approach to the problem is to attach to the transmitted message a preamble consisting of a known bit sequence. The receiver uses it to estimate channel response. The MLSE algorithm is applied to the received data using the estimated channel response in lieu of the actual. Performance is then compared under ideal (known channel characteristics) and realistic (estimated channel characteristics) conditions, as a test on the effectiveness of the channel estimation procedure. This approach can be extended to channels exhibiting time variations. Updates to the matched filter and metrics are performed so that the channel distortion is more precisely represented by the estimated encoding. This is done periodically by injecting a short, known sequence into the message stream. The adaptive MLSE algorithm was tested on signals recorded during an hf field test.