Te Sun Han, Waseda Univeristy, Japan

Title: Multi-user Hypothesis Testing with Coding Error

Abstract:
Distributed multi-sensor systems are considered, instead of a single sensor system, which are, for example, to widely and effectively detect an unknown object. In these systems each sensor i is assigned rate R_i, (i=1, 2,...,k) to send the signal i (received by sensor i) to a common information processing center C, which is requested to make an optimal decision about " absence or presence " of an unknown object. More specifically, suppose that we have two sensors A, B (at remote sites) which receives signal X, Y, respectively (correlated: null hypothesis H), if there is no object; on the other hand, if there is an object, which receive signals X, Y, respectively (correlated: alternative hypothesis H). The problem treated here is that under rate constraint R_i and given type I error probability Pr(H|H)≤ ε, one tries to make type II error probability Pr(H|H) as small as possible. The novel aspect in this talk is to incorporate "coding error probability" (in addition to hypothesis error probabilities), which was out of scope in the literature. We discuss a possibility to further improve type II error probability by taking account of coding aspects.