M-ary Digital Communication Systems

Thus far we have considered only binary decision problems, that is, models where the decision space consists of two elements. In such problems the signal space either consisted of two elements or was partitioned into two disjoint subspaces. Let us now develop the more general situation in which the signal set consists of M possible signals or M classes of signals with a priori probabilities π;i = 1, …, M. Such a model arises in certain communication systems and in pattern-recognition problems. For example, in communications, instead of recognizing or detecting a binary sequence bit by bit, as we have done previously, we might delay the decision until a sequence of length n has been observed, at which point the receiver attempts to decide which signal has been transmitted from a class of M = 2 n possibilities, that is the class of possible sequences of binary bits of length n. This is known as block encoding.