Ordering Finite-State Markov Channels by Mutual Information

Abstract: In previous work, an ordering result was given for the symbolwise probability of error using general Markov channels, under iterative decoding of LDPC codes. In this paper, the ordering result is extended to mutual information, under the assumption of an iid input distribution. For certain channels, in which the capacity-achieving input distribution is iid, this allows ordering of the channels by capacity. The complexity of analyzing general Markov channels is mitigated by this ordering, since it is possible to immediately determine that a wide class of channels, with different numbers of states, has a smaller mutual information than a given channel.