Universal Quickest Sensing of Spectrum Change in Millimeter Wave Communications: A Data Driven Approach

As millimeter wave is becoming the fundamental signaling technology of the physical layer standard in the next generation cellular network, it also brings about many questions and challenges. Not all the existing theories and methods for traditional wireless communications can apply directly to millimeter wave network because of the adoption of directional antenna, blockage effect, and the unprecedentedly large bandwidth. Among them, spectrum sensing is one of the open challenges, for the purpose of dynamic spectrum access in the millimeter wave band. In this paper, we propose a data driven sensing technique based on the mean recurrence time of random process to efficiently detect the change in the primary user (PU) activities, which can tolerate small fluctuations in the distribution. The proposed spectrum sensing works well without a priori knowledge of the PUs, and does not take the assumption of independent and identically distributed observations on the PUs. It can also serve as a general framework for change detection in other areas.