A recent work investigated whether random measurements of sparse signals provide an efficient method of representing sparse signals. How the random measurements are encoded into bits are the basis for the performance of the source coding. Even very weak forms of universality are precluded using familiar forms of quantization. The recent work also showed that recovery of the sparsity pattern is asymptotically impossible and that the mean-squared error peformance is far from optimal.
ASJC Scopus subject areas
- Signal Processing
- Electrical and Electronic Engineering
- Applied Mathematics