TV White Space Spectrum Analysis Based on Machine Learning

Exploration of TV white space (TVWS) is a promising solution to mitigate the spectrum shortage and provide opportunities for new applications. In this paper, we present a detailed analysis of spectrum utilisation over TVWS at different locations in London. Both short-term and long-term outdoor measurement campaigns are conducted over large scales to better understand the spectrum features and variations across multiple locations and time periods. Different from most fixed-location-only measurements, we also drive along the main streets of London with a portable moving node to measure the on-route spectrum density along with the corresponding geographical information, which allows us to study the features and variations of spectrum use through a continuous space. To better analyse the dynamic spectrum utilisation, a machine learning based analysis algorithm is developed over the real-world measurements. This approach allows us to characterise the similarity and variability in spectrum usage within and among different channels, locations, and time instances, which is critical for the secondary system deployment to efficiently exploit the white space.