Application of Multiple Hypothesis Testing for Beam Selection

摘要

The beam selection problem is one of the crucial problems for achieving high spectral efficiency in millimeter wave (mmW) systems. Most of the previous works use heuristic algorithms with fixed training length to solve this problem, without considering the optimal length of the training sequence. Thus the training length is often over-designed. In this paper, we show that beam selection by exhaustive search can be interpreted as an M-ary hypothesis test, where the optimal training length can be found using the formula of the selection probability. Using this relation, we design two algorithms based on composite hypothesis test theory to determine the optimal training length. Simulations show the applicability of this approach.