Multiband maximum likelihood signal detection based on compressive measurements

Cognitive radios impose challenges on the design of efficient signal detectors, including wide bandwidth sensing and large dynamic range support. The recently considered compressed sensing theory helps in relaxing the constraints on the design of the analog front-end. The maximum likelihood method introduced here is computationally simple since it does not require a signal reconstruction, unlike most methods introduced in the current literature. Moreover, the metric is optimum, works for any modulation scheme and is independent of the emitted signal knowledge and the number of occupied bands. The results are supported with Matlab simulations, a statistical study is performed and the probabilities of misdetection and false alarm are plotted for different scenarios, proving the efficiency of the estimator in a range of plausible SNRs and subsampling factors.