Subband-split Co-Prime Spectral Analyzer with Convolution Window Prototype Filters

To remove the difficulty of recognizing cross-band sources and discriminating densely-distributed components in the existing undersampling based spectrum sensing methods, this paper proposes a subband-split coprime spectral analyzer (SUSCOSA) with convolution window prototype filters. Specifically, on basis of the subband-split concept, this SUSCOSA is synthesized by multiple sub coprime analyzers, whose prototype filters are analytically derived from convolution windows. In this way, SUSCOSA not only eliminates the spurious peak effect of the classical coprime analyzer, but also improves the spectral resolutions with times. Besides, this analytic filter configure mode makes the construction of SUSCOSA efficient and convenient. Moreover, SUSCOSA does not impose any constraint on source sparsity, thus it acquires superiority over the MWC based method that it not only can recognize sources spanning multiple frequency units but also can discriminate densely-located components. Numerical results show that, SUSCOSA can globally improve the coprime spectrum’s definition provided that sample consumption remains unchanged. Meanwhile, the resolvable source numbers (both densely-distributed components and cross-band sources are included) by SUSCOSA can exceed the upper bound of the MWC method, even when the former’s Nyquist sample consumption ratio is lower than half of the latter. These superiority presents SUSCOSA with vast potentials in varies of application fields relevant to wideband spectrum sensing.