MULTIPLE HYPOTHESIS TESTING FOR COMPRESSIVE WIDEBAND SENSING

The classical Compressive Sensing (CS) techniques for wideband spectral sensing consist of a two-stage estimation detection approach. A novel approach is proposed to solve the detection problem directly from observations with incomplete frequency information, for e.g., signals acquired using sub-Nyquist rate sampling. The wideband occupancy detection problem is formulated as a multiple hypothesis testing problem under a non-Bayesian framework, and a Neyman-Pearson-like criterion is proposed. The detector based on exhaustive search performs better than the conventional CS based techniques. However, it is impractical for large block sizes. Hence, we propose a sub-optimal greedy algorithm whose complexity and performance can be traded-off by construction.