Sparse signal sensing with non-uniform undersampling and frequency excision

Abstract

We propose a novel compressive sensing algorithm for cognitive radio networks, based on non-uniform under-sampling. It is known that the spectrum of uniformly under-sampled signals exhibit frequency aliasing, whereby the frequency location is impossible. To alleviate aliasing, non-uniform sampling can be used. This, however, generates a high level of frequency leakage that prevents detection of weaker signals. To alleviate this problem, we introduce a novel iterative frequency excision technique that allows to detect tones or modulated signals below the original noise floor due to leakage. This method can be used in cognitive radio sensing engines, allowing to sense very wide bandwidths with a relatively low average sample rate. 20dB of leakage reduction can easily be achieved with this method.