A comparison between multi-sensor and CP-based spectrum sensing for TV white spaces

The design of an efficient Cognitive Radio [1], [2] system often requires the development of a spectrum sensing unit [3], [4] making the wireless system or network aware of the available transmission opportunities. In order to provide an indication about channel availability, a binary hypothesis testing experiment must be periodically performed [5]. Several methods have been proposed for the computation of such statistics: a comprehensive description can be found in [6] and references therein. The Digital TeleVision (DTV) white spaces offer an interesting context for the evaluation of novel spectrum sensing techniques. Starting from an analysis of the statistical properties of real DVB-T signals generated using a software-defined transmitter, this work aims to compare several well-known spectrum sensing algorithms: single-sensor Cyclic Prefix-based sensing (CP-based), and multi-antenna methods based on eigenvalue decomposition of the sample covariance matrix. Our research aims at establishing relationship among different algorithms by evaluating trade-offs of detection performance, algorithm complexity, number of sensors employed and sensing time. The first results obtained are shown here in terms of Receiver Operating Curve (ROC) and detection probability PD under constant false alarm probability PFA. The algorithms input data consists of real DVB-T signals generated using a software-defined DVB-T transmitter