Maximum Likelihood Detection from Multiple Bit Sources

Abstract

This paper deals with the problem of producing the best bit stream from a number of input bit streams with varying degrees of reliability. The best source selector and smart source selector are recast as detectors, and the maximum likelihood bit detector (MLBD) is derived from basic principles under the assumption that each bit value is accompanied by a quality measure proportional to its probability of error. We show that both the majority voter and the best source selector are special cases of the MLBD and define the conditions under which these special cases occur. We give a mathematical proof that the MLBD is the same as or better than the best source selector.