Digital TV Channel Prediction using Clustering Algorithms and Statistical Learning

Abstract

Due to the rise of new communication services, more portions of the electromagnetic spectrum must be relocated and their distribution optimized. With the digitization of the open TV service, it was observed that the distribution of channels in  he frequency band destined for this service generated an inefficient use of the radio spectrum. These unused frequency bands are the so-called void spaces. To establish efficient spectrum use, it is important to identify these spectrum gap  opportunities and use according to certain criteria. In this article, machine learning algorithms are proposed to identify new spectrum opportunities, through the signal levels received in the UHF frequency range of the Digital TV system. These  spectrum opportunities are generated from natural or artificial obstacles present in the propagation environment. Two measurement campaigns were carried out in a suburban area to obtain the level of the signal received in an area of approximately 240,000 square meters. From the received power values, machine learning algorithms were used to make prediction of the received signal levels. By using a reception threshold, it is possible to identify the shadow regions and possible  spectrum opportunities.