Abstract

Bogdan Roman is at the Dept of Applied Mathematics and Theoretical Physics, University of Cambridge and he is also a Visiting Research Fellow at CL.

[Disclaimer] This will be a non-mathematical talk, accessible to non-experts, and will hopefully also feature some live Matlab demos. Compressed Sensing (CS) took the applied maths and signal processing communities by storm, being by far one of the hottest topics in the last decade. CS states that one can recover signals by sampling at random and much below the known classical Nyquist limit … if some conditions are met. However, the real world is, as usual, cruel and unforgiving. This talk will (i) explain the core concepts and why some of the fundamental conditions in CS are problematic or missing in a large number of practical applications, (ii) introduce a new CS theory based on asymptotic behaviour that real-world signals possess, and (iii) show why the new concepts are a better fit and how they can be exploited in some important practical problems in order to achieve substantial gains even when the traditional CS conditions actually do hold.