Performance comparison of reconstruction algorithms in discrete blind multi-coset sampling

This paper investigates the performance of different reconstruction algorithms in discrete blind multi-coset sampling. Multi-coset scheme is a promising compressed sensing architecture that can replace traditional Nyquist-rate sampling in the applications with multi-band frequency sparse signals. The performance of the existing compressed sensing reconstruction algorithms have not been investigated yet for the discrete multi-coset sampling. We compare the following algorithms - orthogonal matching pursuit, multiple signal classification, subspace-augmented multiple signal classification, focal under-determined system solver and basis pursuit denoising. The comparison is performed via numerical simulations for different sampling conditions. According to the simulations, focal under-determined system solver outperforms all other algorithms for signals with low signal-to-noise ratio. In other cases, the multiple signal classification algorithm is more beneficial.