Mean-based Tests And Aymptotic Performance For Digital Communications

Abstract

The method proposed is one that is based on the descrepancy of the means under different hypotheses. A hypotheses test is constructed in terms of conditional innovations sequences generated from the received sequence. Observing that the only statistics that change with hypotheses are the conditional means, the problem is treated as a test of the descrepancy of the means. An approximate sequential test is used and the test statistic is shown to be Frasersufficient. Asymptotic probability of error results are obtained using Cramer's theorem for maximum likelihood estimates.