Auto-calibration of CS system

What is it about ?

Let's say you just built or acquired a new Compressed Sensing device. It could be either an array of microphones, a camera-sensor, a MRI device, etc.

At a stage of the use of your device, you will have to calibrate it. Calibration could occur on the position of the microphones on your array, on the sensitivity of your camera sensor, and can be of any nature : temperature, sound pressure, luminance, etc.

In the following, we present a blind calibration method for compressive sensing systems.

How does it work ?

We define the decalibration of the system as unkwown gains on each mesure.

The challenge here is to perform the calibration in blind context, i.e. the signal measured is already distorted and we can't have access to the clean signal.

Our contribution leads to show that the considered unsupervised calibration problem can actually be exactly expressed as a convex optimization problem.

More details

To come (soon) !