Sparse signal processing has become a major component in modern signal processing theory, underpinning compressed sensing, linear regression, machine learning, big data processing, etc. It is based on the observation that most signals, under certain transform or dictionary, only contain a few significant components. Based on this sparsity, efficient ways for data acquisition, processing, and analysis can be developed.
The main aim of this experiment is to familar students with the concept of sparse signals and the easy-tounderstand techniques for sparse signal recovery.
In the first section, we highlight the limitations of least squares methods when applied to sparse signal processing, and suggest the paradigm shift from least squares approach to modern sparse recovery techniques. In the second section, three greedy algorithms are introduced to solve sparse recovery problem. Matlab implementations and numerical comparison of these three algorithms are required.