Signal Processing
Simplified Matrix-Vector Multiplication
State of the Art:
|
Linear Computation Coding:
|
Compressed Sensing
What: Project source onto lower-dimensional subspace & sample there; then reconstruct lost information via inherent source redundancy
Why: Reduce power consumption & hardware requirements
- Faster image acquisition
- Higher resolution
How: Solve the reconstruction task as optimization problem using prior knowledge of signal
Challenge: Reconstruction algorithms may diverge
Vector-AMP used for MIMO radar imaging
Identifiability of Semi-Blind Parameter Estimation
What: Conditions to uniquely estimate parameters in bilinear systems (estimate H and X from noisy observation Y=HX+W)
Why: None / limited availability of training data -> infeasible parameter estimation (even if noiseless observation)
Where: Geographically distributed transmit and receive nodes
Mapping to a random bipartite geometric graph |
||
TX and RX modeled as independent random point processes |
Study identifiability in terms of statistical graph parameters and properties |