Multi-terminal compressed sensing
Description:
In the usual compressed sensing (compressive sampling), we assume that we have a signal x of dimension n and we are taking m < n, possibly noisy, measurements y = Ax+N from the signal using an m×n measurement matrix A. We call m/n the measurement rate. As it is known, under suitable regularity conditions, i.e., sparsity or compressibility on the signal, it is possible to stably recover the signal with asymptotically negligible distortion. This setting is what we intend to call “single terminal” compressed sensing. In the multi-terminal setting (for simplicity two-terminal case), we assume that the signal x is not fully observable through one terminal. In other words, x = [x1;x2] where the xi part can be measured through terminal i = 1,2. Therefore, in this case instead of just one parameter (measurement rate m/n), we have a vector of measurement rates [m1/n ; m2/n] characterizing the performance of the recovery process.
Prerequisites:
A good background in Math, A good programming background in C/C++ and Matlab.
Supervisor:
Saeid Haghighatshoar, LCM (saeid.haghighatshoar@epfl.ch)
Professor:
Prof. Bixio Rimoldi