Sparse Sampling Strategies and Parameter Identifiability
Statistical Signal Processing
Antenna Array Processing with Application in Radar Signal Processing.
Sparse Representation, Reconstruction and Learning
A central theme in my doctoral research has been that of understanding the role of sparse sampling strategies in pushing the limits of identifiability in parameter estimation. In particular, we ask the question: Can Sparse Sampling Strategies be employed to possibly increase the number of parameters that can be identified ? Also, do such strategies depend on the underlying statistical distribution of the signals ? In the course of my doctoral work, I have found interesting answers to these questions, leading to development of novel sampling strategies and algorithms that are fundamentally capable of estimating parameters in a largely underdetermined observation model that often arise in array processing. Another key contribution of my research has been to develop a new paradigm for support recovery of sparse signals which explicitly utilises the correlation present in multiple measurement vector models. Our approach has shown that it is fundamentally possible to increase the level of recoverable sparsity using the proposed correlation aware framework over existing techniques that do not use such information.
