Prediction Based Method for Faster Compressive Sensing Reconstruction Using OMP

Orthogonal Matching Pursuit (OMP) is the most widely used reconstruction algorithm in Compressive Sensing. The main challenge of the implementation is that it has more time consuming reconstruction process due to its iterative nature and complex matrix inverse finding operation. This paper presents a hybrid approach of implementing OMP on hardware using parallel blocks and optimum memory storage to expedite reconstruction, making it usable for applications where rapid reconstruction is desired. Proposed approach speeds up reconstruction process up to two times as compared to the previous work. The proposed approach uses the concept of adding the multiple matrix inverse blocks and sorting algorithm to find three most weighted columns. Increase in speed of operation comes at cost at hardware (size of hardware on silicon) increasing 3 times in comparison to conventional method.