Some properties of the log-likelihood difference of mismatched maximum likelihood sequential estimation

Due to the consideration of additive white Gaussian noise, the performance evaluation of most communication receivers applying the maximum likelihood principle (MLSE) is based on the Euclidean distance, denoted by d²0. The present authors consider another parameter, the log-likelihood difference for describing the performance of a MLSE receiver under uncoded modulation. This parameter proves to be simpler and exhibits some nice properties under the mismatched receiver situation. Some bounds on the tolerable mismatch are given. The performance of a blind channel estimator is illustrated