Title: Structured Non Uniform Sampling for Spectrum Estimation
Speaker: Piya Pal (UMD)
This talk explores the role of sparsity, structure and statistical information in developing efficient sub-Nyquist sampling strategies for spectrum estimation. The statistical information contained in wideband wide sense stationary random signals alone can be exploited to sample such signals at rates significantly lower than the Nyquist rate. When additional information regarding the low dimensional structure of the signal is available (in the form of sparsity and/or low rank), it is possible to further reduce the sampling rate by incorporating statistical priors alongside the structural information. As an illustrative case study, we will consider the problem of spatial line spectrum estimation and demonstrate that sparsity and correlation can be jointly exploited to identify more lines than the number of spatial sensors. In this context, we will also present recent and ongoing research on classical versus sparsity promoting approaches, the effect of basis mismatch, and analyze some recent ‘gridless’ algorithms for line spectrum estimation.