(英) |
Applications of machine learning in wireless communications are introduced in this presentation. In millimeter-wave (mmWave) communication, throughput estimation using camera imagery has been studied. Human blockage in mmWave communication sharply degrades link quality, and the blockage could be predicted from imagery since camera imagery contains geometry of transmitter, receiver, and pedestrian, and shape and mobility of pedestrian, which characterize human blockage in mmWave communications. Experimental results demonstrate that the scheme estimates throughput from imagery. In a spectrum sharing system, we proposed a scheme to estimate region where harmful interference could occur and update primary exclusive region (PER) in order not to cause harmful interference. Support vector machine (SVM) learns interference region and decides appropriate boundary for PER. The simulation results demonstrate that the area of PER with the proposed scheme is smaller than that of the fixed circular PER setting with the same interference probability. |